Merge main into gatekeeper/fix-3530-1

# Conflicts:
#	rust/lancedb/src/table/query.rs

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.38.0"
+current_version = "0.38.0-beta.11"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/dependabot.yml

@@ -44,27 +44,3 @@ updates:
       python-deps:
         patterns:
           - "*"
-
-  # The npm ecosystem covers pnpm lockfiles. There are two separate installs:
-  # the bindings themselves and the examples, which have their own lockfile.
-  # As with cargo and pip above, only bump the lockfile — the version ranges
-  # in package.json are our consumers' constraints, not ours.
-  - package-ecosystem: npm
-    directory: /nodejs
-    schedule:
-      interval: weekly
-    versioning-strategy: lockfile-only
-    groups:
-      nodejs-deps:
-        patterns:
-          - "*"
-
-  - package-ecosystem: npm
-    directory: /nodejs/examples
-    schedule:
-      interval: weekly
-    versioning-strategy: lockfile-only
-    groups:
-      nodejs-examples-deps:
-        patterns:
-          - "*"

.github/workflows/dev.yml

@@ -29,14 +29,12 @@ jobs:
     steps:
       - uses: actions/setup-node@v6
         with:
-          node-version: "24"
-      - uses: pnpm/action-setup@v6
-        with:
-          version: 11.1.1
+          node-version: "18"
       # These rules are disabled because Github will always ensure there
       # is a blank line between the title and the body and Github will
       # word wrap the description field to ensure a reasonable max line
       # length.
+      - run: npm install @commitlint/config-conventional
       - run: >
           echo 'module.exports = {
             "rules": {
@@ -45,11 +43,7 @@ jobs:
               "body-leading-blank": [0, "always"]
             }
           }' > .commitlintrc.js
-      - run: >
-          pnpm dlx
-          --package @commitlint/cli@21.2.2
-          --package @commitlint/config-conventional@21.2.2
-          commitlint --extends @commitlint/config-conventional --verbose <<< $COMMIT_MSG
+      - run: npx commitlint --extends @commitlint/config-conventional --verbose <<< $COMMIT_MSG
         env:
           COMMIT_MSG: >
             ${{ github.event.pull_request.title }}
@@ -60,7 +54,7 @@ jobs:
         with:
           script: |
             const message = `**ACTION NEEDED**
-
+              
               Lance follows the [Conventional Commits specification](https://www.conventionalcommits.org/en/v1.0.0/) for release automation.
 
               The PR title and description are used as the merge commit message.\

.github/workflows/docs-link-check.yml

@@ -56,7 +56,7 @@ jobs:
         uses: lycheeverse/lychee-action@e7477775783ea5526144ba13e8db5eec57747ce8 # v2.9.0
         with:
           # Restricted to http(s) on purpose. Much of docs/src is generated
-          # API reference (the js/ tree comes from `pnpm run docs` in nodejs)
+          # API reference (the js/ tree comes from `npm run docs` in nodejs)
           # and the hand-written pages use mkdocstrings cross-references and
           # nav-relative paths that only resolve in the site mkdocs builds,
           # not in this checkout, so relative links would be reported as

.github/workflows/docs.yml

@@ -55,7 +55,9 @@ jobs:
       - name: Set up node
         uses: actions/setup-node@v6
         with:
-          node-version: 24
+          node-version: 20
+          cache: 'npm'
+          cache-dependency-path: docs/package-lock.json
       - name: Install node dependencies
         working-directory: nodejs
         run: |

.github/workflows/nodejs.yml

@@ -47,8 +47,9 @@ jobs:
         version: 11.1.1
     - uses: actions/setup-node@v6
       with:
-        # Build on a supported LTS; the matrix job below covers every
-        # Node version the library claims to support.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. The library itself still supports Node >= 18
+        # (see test matrix below).
         node-version: 24
         cache: 'pnpm'
         cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -83,7 +84,7 @@ jobs:
     timeout-minutes: 30
     strategy:
       matrix:
-        node-version: [ "22", "24", "26" ]
+        node-version: [ "18", "20" ]
     runs-on: "ubuntu-22.04"
     defaults:
       run:
@@ -100,9 +101,9 @@ jobs:
     - uses: actions/setup-node@v6
       name: Setup Node.js 24 for build
       with:
-        # Build and install once on a fixed version so the generated docs
-        # are identical across matrix legs; the tests below then run on each
-        # supported Node version.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October. Build/install runs on Node 24; tests run on the
+        # matrix version below using direct jest invocation.
         node-version: 24
         cache: 'pnpm'
         cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -151,9 +152,9 @@ jobs:
         S3_TEST: "1"
         # Newer @smithy/core uses dynamic ESM imports.
         NODE_OPTIONS: "--experimental-vm-modules"
-      # Invoke the installed jest binary directly; the pnpm shim is set up
-      # against the build-phase Node, not the version selected above.
-      run: node_modules/.bin/jest --verbose
+      # Invoke jest directly because pnpm 11 itself requires Node 22+
+      # while the matrix tests on older Node versions.
+      run: npx jest --verbose
     - name: Test examples
       working-directory: ./
       env:
@@ -163,7 +164,7 @@ jobs:
       run: |
         python ci/mock_openai.py &
         cd nodejs/examples
-        node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose
+        npx jest --testEnvironment jest-environment-node-single-context --verbose
   macos:
     timeout-minutes: 30
     # macos-15 ships a newer linker; the older macos-14 linker fails to insert
@@ -184,7 +185,8 @@ jobs:
         version: 11.1.1
     - uses: actions/setup-node@v6
       with:
-        # pnpm 11 requires Node >= 22.13.
+        # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+        # in October.
         node-version: 24
         cache: 'pnpm'
         cache-dependency-path: nodejs/pnpm-lock.yaml

.github/workflows/npm-publish.yml

@@ -40,31 +40,40 @@ jobs:
           - target: aarch64-apple-darwin
             host: macos-latest
             features: fp16kernels
-            # Fat LTO was ~111 of this job's ~113 minutes.
-            lto: thin
-            codegen_units: 16
             pre_build: |-
               brew install protobuf
+              # Fat LTO (the workspace default in .cargo/config.toml) is
+              # single-threaded and is the peak-memory step of the build. On
+              # this runner it accounted for ~111 of the job's ~113 minutes,
+              # making it the critical path of the entire publish pipeline.
+              # ThinLTO parallelizes it across the runner's cores, for a few
+              # percent of runtime performance.
+              export CARGO_PROFILE_RELEASE_LTO=thin
+              export CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16
           - target: x86_64-pc-windows-msvc
             host: windows-2025
             features: ","
-            # The lower peak also keeps this on the standard 4-core runner.
-            lto: thin
-            codegen_units: 16
             pre_build: |-
               choco install --no-progress protoc ninja nasm
               tail -n 1000 /c/ProgramData/chocolatey/logs/chocolatey.log
               # There is an issue where choco doesn't add nasm to the path
               export PATH="$PATH:/c/Program Files/NASM"
               nasm -v
+              # See the ThinLTO note on aarch64-apple-darwin above. Keeping
+              # peak memory down is also what lets this run on the standard
+              # 4-core runner: the 8-core larger runner was only needed to
+              # stop fat LTO from OOMing rustc-LLVM.
+              export CARGO_PROFILE_RELEASE_LTO=thin
+              export CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16
           - target: aarch64-pc-windows-msvc
             host: windows-2025
             features: ","
-            lto: thin
-            codegen_units: 16
             pre_build: |-
                 choco install --no-progress protoc
                 rustup target add aarch64-pc-windows-msvc
+                # See the ThinLTO note on aarch64-apple-darwin above.
+                export CARGO_PROFILE_RELEASE_LTO=thin
+                export CARGO_PROFILE_RELEASE_CODEGEN_UNITS=16
           - target: x86_64-unknown-linux-gnu
             host: ubuntu-latest
             features: fp16kernels
@@ -94,14 +103,6 @@ jobs:
             # https://github.com/napi-rs/napi-rs/blob/main/debian-aarch64.Dockerfile
             docker: ghcr.io/napi-rs/napi-rs/nodejs-rust:lts-debian-aarch64
             features: "fp16kernels"
-            # Fat LTO OOM-killed rustc every nightly; even with lld it peaked
-            # at 31391 MiB of the runner's 32 GiB.
-            lto: thin
-            codegen_units: 16
-            # arm64 Linux links through GNU `ld` where x86_64 defaults to
-            # `rust-lld`, which is why only arm64 OOM'd. lld cut the largest
-            # linker process 7.0 -> 4.0 GiB (lancedb/sophon#7313).
-            linker: /tmp/aarch64-lld-clang
             pre_build: |-
               set -e &&
               apt-get update &&
@@ -111,30 +112,9 @@ jobs:
               # AT_HWCAP2 (added in Linux 3.17). Define it for aws-lc-sys.
               export CFLAGS="$CFLAGS -DAT_HWCAP2=26" &&
               rustup target add aarch64-unknown-linux-gnu
-              # Not `&&`-chained: in dash, errexit does not fire for a
-              # non-final command in an `&&` list, so failures were ignored.
-              #
-              # A wrapper rather than `-C link-arg` because the per-target
-              # rustflags variable does not reach every unit that links, while
-              # the linker variable does. `clang` because GCC silently ignores
-              # `-fuse-ld=lld` unless built with lld support. Two echoes
-              # because printf's newline escape gets rewritten to `;` between
-              # here and the container.
-              echo '#!/bin/sh' > /tmp/aarch64-lld-clang
-              echo 'exec clang --target=aarch64-unknown-linux-gnu --sysroot=/usr/aarch64-unknown-linux-gnu/aarch64-unknown-linux-gnu/sysroot --gcc-toolchain=/usr/aarch64-unknown-linux-gnu -fuse-ld=lld "$@"' >> /tmp/aarch64-lld-clang
-              chmod 0755 /tmp/aarch64-lld-clang
-              # Fail now, not at the cdylib link ~30 minutes later. Linking at
-              # all also proves lld resolved; clang errors out when it cannot.
-              echo 'int main(void){return 0;}' > /tmp/probe.c
-              /tmp/aarch64-lld-clang /tmp/probe.c -o /tmp/probe
-              readelf -h /tmp/probe | grep AArch64
           - target: aarch64-unknown-linux-musl
             host: ubuntu-2404-8x-x64
             features: ","
-            # Fat LTO took the whole runner down. lld cannot help: it died
-            # inside rustc's LLVM, before any linker was spawned.
-            lto: thin
-            codegen_units: 16
             pre_build: |-
               set -e &&
               sudo apt-get update &&
@@ -143,19 +123,6 @@ jobs:
               export EXTRA_ARGS="-x"
     name: build - ${{ matrix.settings.target }}
     runs-on: ${{ matrix.settings.host }}
-    # On the job, not exported from `pre_build`: `Swatinem/rust-cache` hashes
-    # `CARGO_*` into its cache key before any step runs, so a step-local export
-    # leaves the key unchanged while cargo still rebuilds cold. The ThinLTO
-    # legs had been doing that every run.
-    #
-    # Not `RUSTFLAGS`: setting it, even to "", discards every config-file
-    # rustflag, silently dropping .cargo/config.toml's `target-cpu` and
-    # `target-feature` from the published binaries.
-    env:
-      CARGO_PROFILE_RELEASE_LTO: ${{ matrix.settings.lto || 'fat' }}
-      CARGO_PROFILE_RELEASE_CODEGEN_UNITS: ${{ matrix.settings.codegen_units || '1' }}
-      # Empty elsewhere: a per-target variable is only read for that triple.
-      CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_LINKER: ${{ matrix.settings.linker }}
     defaults:
       run:
         working-directory: nodejs
@@ -168,7 +135,8 @@ jobs:
       - name: Setup node
         uses: actions/setup-node@v6
         with:
-          # pnpm 11 requires Node >= 22.13.
+          # pnpm 11 requires Node >= 22.13; use 24 since 22 hits EOL
+          # in October.
           node-version: 24
           cache: pnpm
           cache-dependency-path: nodejs/pnpm-lock.yaml
@@ -201,15 +169,19 @@ jobs:
           # creating ref). The nightly cadence also keeps entries inside
           # GitHub's 7-day eviction window, which a tag-only trigger would not.
           save-if: ${{ github.ref == 'refs/heads/main' }}
-      # Docker builds can use rust-cache too: the workspace is bind-mounted, so
-      # `target/` lives on the host and rust-cache's prune keeps the entry
-      # small.
+      # Docker builds can use rust-cache too. `target/` already lives on the
+      # host because the whole workspace is bind-mounted into the container, and
+      # rust-cache's prune and save run host-side, so they can manage it -- which
+      # is what keeps the entry to dependency artifacts rather than a multi-GB
+      # copy of everything.
       #
       # Two differences from the native builds. The container's CARGO_HOME is
-      # bind-mounted from `.cargo-cache` rather than ~/.cargo, so that is cached
-      # explicitly. And the key uses the *host* rustc version, not the compiler
-      # that built these artifacts -- safe, since cargo fingerprints the real
-      # one; a base-image bump just costs one cold build.
+      # bind-mounted from `.cargo-cache` rather than the host's ~/.cargo, so that
+      # has to be cached explicitly. And the key is derived from the *host* rustc
+      # version, which is not the compiler that produced these artifacts; that is
+      # safe because cargo fingerprints the real compiler and rebuilds on a
+      # mismatch, it just means a base-image toolchain bump costs one cold build
+      # instead of invalidating the key.
       - name: Cache cargo (docker builds)
         uses: Swatinem/rust-cache@v2
         if: ${{ matrix.settings.docker }}
@@ -238,19 +210,14 @@ jobs:
           # cache step above saves. Previously the registry mounts pointed at
           # `.cargo/...`, a path nothing cached, so the container re-downloaded
           # the whole crate registry on every run.
-          #
-          # `docker run` inherits nothing; `-e NAME` carries the job's `env:` in.
           options: "--user 0:0 -v ${{ github.workspace }}/.cargo-cache/git/db:/usr/local/cargo/git/db \
                     -v ${{ github.workspace }}/.cargo-cache/registry/cache:/usr/local/cargo/registry/cache \
                     -v ${{ github.workspace }}/.cargo-cache/registry/index:/usr/local/cargo/registry/index \
-                    -e CARGO_PROFILE_RELEASE_LTO \
-                    -e CARGO_PROFILE_RELEASE_CODEGEN_UNITS \
-                    -e CARGO_TARGET_AARCH64_UNKNOWN_LINUX_GNU_LINKER \
                     -v ${{ github.workspace }}:/build -w /build/nodejs"
           run: |
             set -e
             ${{ matrix.settings.pre_build }}
-            node_modules/.bin/napi build --platform --release \
+            npx napi build --platform --release \
               --features ${{ matrix.settings.features }} \
               --target ${{ matrix.settings.target }} \
               --dts ../lancedb/native.d.ts \
@@ -270,7 +237,7 @@ jobs:
       - name: Build
         run: |
           ${{ matrix.settings.pre_build }}
-          node_modules/.bin/napi build --platform --release \
+          npx napi build --platform --release \
               --features ${{ matrix.settings.features }} \
               --target ${{ matrix.settings.target }} \
               --dts ../lancedb/native.d.ts \
@@ -289,18 +256,6 @@ jobs:
         if: always()
         run: df -h
         shell: bash
-      - name: Report peak memory
-        if: always() && runner.os == 'Linux'
-        shell: bash
-        run: |
-          peak=$(find /sys/fs/cgroup -name memory.peak -readable \
-                 -exec cat {} + 2>/dev/null | sort -n | tail -1)
-          if [ -n "$peak" ]; then
-            echo "peak memory: $((peak / 1024 / 1024)) MiB"
-          else
-            echo "peak memory: unavailable (no readable cgroup v2 memory.peak)"
-          fi
-          free -g || true
       - name: Upload artifact
         uses: actions/upload-artifact@v7
         with:
@@ -338,7 +293,7 @@ jobs:
           - target: aarch64-unknown-linux-gnu
             host: ubuntu-2404-8x-arm64
         node:
-          - '22'
+          - '20'
     runs-on: ${{ matrix.settings.host }}
     defaults:
       run:
@@ -384,9 +339,9 @@ jobs:
       - name: Move built files
         run: cp dist/native.d.ts dist/native.js dist/*.node lancedb/
       - name: Test bindings
-        # Invoke the installed jest binary directly; the pnpm shim is set up
-        # against the install-phase Node, not the version selected above.
-        run: node_modules/.bin/jest --verbose
+        # Invoke jest directly because pnpm 11 itself requires Node 22+
+        # while the matrix tests on older Node versions.
+        run: npx jest --verbose
   publish:
     name: Publish
     runs-on: ubuntu-latest

.github/workflows/update_package_lock_run.yml

@@ -0,0 +1,22 @@
+name: Update package-lock.json
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          ref: main
+          persist-credentials: false
+          fetch-depth: 0
+          lfs: true
+      - uses: ./.github/workflows/update_package_lock
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}

.github/workflows/update_package_lock_run_nodejs.yml

@@ -0,0 +1,22 @@
+name: Update NodeJs package-lock.json
+
+on:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          ref: main
+          persist-credentials: false
+          fetch-depth: 0
+          lfs: true
+      - uses: ./.github/workflows/update_package_lock_nodejs
+        with:
+          github_token: ${{ secrets.LANCEDB_RELEASE_TOKEN }}

.pre-commit-config.yaml

@@ -20,10 +20,7 @@ repos:
     hooks:
       - id: local-biome-check
         name: biome check
-        # Use the biome from nodejs/package.json rather than a separately
-        # pinned one: the two drifted apart and disagreed on formatting, so
-        # this hook rejected code that `pnpm lint` accepted.
-        entry: nodejs/node_modules/.bin/biome check --config-path nodejs/biome.json nodejs/
+        entry: npx @biomejs/biome@1.8.3 check --config-path nodejs/biome.json nodejs/
         language: system
         types: [text]
         files: "nodejs/.*"

AGENTS.md

@@ -38,7 +38,7 @@ Before committing changes, run formatting for every language you touched. At min
 * Rust changes: run `cargo fmt --all`.
 * Python changes: run `ruff format .` and `ruff check .` from the repository root,
   and run targeted tests through `cd python && uv run ...`.
-* TypeScript changes: run the relevant `pnpm` lint, format, build, and docs commands in `nodejs`.
+* TypeScript changes: run the relevant `npm`/`pnpm` lint, format, build, and docs commands in `nodejs`.
 
 Before creating a PR, the exact value passed to `gh pr create --title` must follow
 Conventional Commits, such as `fix: support nested field paths in native index creation`
@@ -101,12 +101,12 @@ Python bindings changes:
 TypeScript bindings changes:
 
 1. Add napi-rs method binding on `Table` in `nodejs/src/table.rs`.
-2. Run `pnpm build` to generate TypeScript definitions.
+2. Run `npm run build` to generate TypeScript definitions.
 3. Add typescript method on abstract class `Table` in `nodejs/src/table.ts`.
 4. Add concrete method on `LocalTable` class in `nodejs/src/native_table.ts`.
     * Note: despite the name, this class is also used for remote tables.
 5. Add test in `nodejs/__test__/table.test.ts`.
-6. Run `pnpm run docs` to generate TypeScript documentation.
+6. Run `npm run docs` to generate TypeScript documentation.
 
 ## Python API reference
 

Cargo.lock

@@ -1597,9 +1597,9 @@ checksum = "613afe47fcd5fac7ccf1db93babcb082c5994d996f20b8b159f2ad1658eb5724"
 
 [[package]]
 name = "chacha20"
-version = "0.10.2"
+version = "0.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "65c35e4b699c7e15ccbe7ee35c005e4fc0a278d22238a2857e6ce2dadeda1b06"
+checksum = "6f8d983286843e49675a4b7a2d174efe136dc93a18d69130dd18198a6c167601"
 dependencies = [
  "cfg-if 1.0.4",
  "cpufeatures 0.3.0",
@@ -3455,9 +3455,8 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"
 
 [[package]]
 name = "fsst"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f727719438dfdb74f358a347c91ff81b6e7084a6421f34de3e473ce271f10caa"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "rand 0.9.5",
@@ -4816,9 +4815,8 @@ checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"
 
 [[package]]
 name = "lance"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "be781f40c7a75f9eae2188a2f71174acb7a360dca97163db40b041d0828dea48"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arc-swap",
  "arrow",
@@ -4890,9 +4888,8 @@ dependencies = [
 
 [[package]]
 name = "lance-arrow"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fb97fd9875f3036d7c2561aa5b16eb87b80ccabaa4eeb5e6099b19cc662f1cd8"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4914,8 +4911,7 @@ dependencies = [
 [[package]]
 name = "lance-arrow-scalar"
 version = "58.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "771f68b04b47f3addf781116f65061808de94b05e1e9411c23c18f32d14ebe79"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4929,20 +4925,17 @@ dependencies = [
 [[package]]
 name = "lance-arrow-stats"
 version = "58.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "dd47ec33c90bf29f688fd02118e37d3a5ad5c339caa3163f89e417dc0867001f"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
- "half",
  "lance-arrow-scalar",
 ]
 
 [[package]]
 name = "lance-bitpacking"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6f45658c5b2dc9aada41b66ee44b83af3fa888b7385ae414bae951b12a9f1cd3"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrayref",
  "crunchy",
@@ -4952,9 +4945,8 @@ dependencies = [
 
 [[package]]
 name = "lance-core"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "27af3df3a7d08897efccd04461df31cedf0880c4b86a055ddce48e423d27f967"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4991,9 +4983,8 @@ dependencies = [
 
 [[package]]
 name = "lance-datafusion"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c377f837df5296e92f9fad724c83c1bef4e74d5af6e5a9312e9307e1dead8614"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5022,9 +5013,8 @@ dependencies = [
 
 [[package]]
 name = "lance-datagen"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "778e1a5065fa4bc184e36e32681f10f8f4680ad8cedc9377b4c088dce8c5b8da"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5041,9 +5031,8 @@ dependencies = [
 
 [[package]]
 name = "lance-derive"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "13e5e95e0fd3d74f7938f4bee623041421b323b5c61f242c8622a1f48a202527"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -5052,9 +5041,8 @@ dependencies = [
 
 [[package]]
 name = "lance-encoding"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1625653c55c65f3426e281f6e29b54c603f38a40bd4cebd707bd4f3ea48be6c5"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -5087,9 +5075,8 @@ dependencies = [
 
 [[package]]
 name = "lance-file"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b7e13c9266b478fc98f36ee19347c4658f7a6613fed77778b1a455fe1b88552e"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -5120,9 +5107,8 @@ dependencies = [
 
 [[package]]
 name = "lance-index"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b0e0cb95f2c4f341c4dd04ac60f6a89ea26a6f75e09570225cbda4854c8b088e"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arc-swap",
  "arrow",
@@ -5186,9 +5172,8 @@ dependencies = [
 
 [[package]]
 name = "lance-index-core"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "79ccd371977c1f7168da259d66ad37154f23146f093d46136bc7f79559f00f2c"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5210,9 +5195,8 @@ dependencies = [
 
 [[package]]
 name = "lance-io"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "414d50997391b1ac83dc183c1612fdff88f58b959806078dc4c5e465154566de"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5252,9 +5236,8 @@ dependencies = [
 
 [[package]]
 name = "lance-linalg"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ff55b152ef23a56d7ba7e4d1b2c9cf0cc79aef6ee607c115597557ea4059f41"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5268,9 +5251,8 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "09991c13ab282b731e323619613914e08da9cc82b312f904e58c128b23f2f0e3"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "async-trait",
@@ -5282,9 +5264,8 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace-impls"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0ec0bc005f6bb8f120774eb4a9ba02e10463d8338167a46cbb1391d46680a174"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-ipc",
@@ -5337,9 +5318,8 @@ dependencies = [
 
 [[package]]
 name = "lance-select"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e8f676a2a1837cc85b77feb5326d3296827da964e40d67144f646563302a6ce9"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -5353,9 +5333,8 @@ dependencies = [
 
 [[package]]
 name = "lance-table"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bd33054347395048b1d842dfb85a13f7801392c2da5f39425db62f00a481744b"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -5395,9 +5374,8 @@ dependencies = [
 
 [[package]]
 name = "lance-testing"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ecc9ad9ae24f045dfddd538a39e55a28fa7e1ca6ad9f23e20d4087eaf2bb66f7"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5410,9 +5388,8 @@ dependencies = [
 
 [[package]]
 name = "lance-tokenizer"
-version = "11.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3bfa6f0164c8b7056150f5682ce4d415a335b59b04c479873fda04b200117d27"
+version = "12.0.0-beta.2"
+source = "git+https://github.com/lance-format/lance.git?tag=v12.0.0-beta.2#dafa4642658d996b3e31dde91e02f72db7860d7e"
 dependencies = [
  "frostem",
  "icu_segmenter",
@@ -5425,7 +5402,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb"
-version = "0.38.0-beta.15"
+version = "0.38.0-beta.11"
 dependencies = [
  "ahash",
  "anyhow",
@@ -5513,7 +5490,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-nodejs"
-version = "0.38.0-beta.15"
+version = "0.38.0-beta.11"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -5538,7 +5515,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-python"
-version = "0.38.0-beta.15"
+version = "0.38.0-beta.11"
 dependencies = [
  "arrow",
  "async-trait",

Cargo.toml

@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"
 
 [workspace.dependencies]
-lance = { "version" = "=11.0.0", default-features = false }
-lance-core = "=11.0.0"
-lance-datagen = "=11.0.0"
-lance-file = "=11.0.0"
-lance-io = { "version" = "=11.0.0", default-features = false }
-lance-index = "=11.0.0"
-lance-linalg = "=11.0.0"
-lance-namespace = "=11.0.0"
-lance-namespace-impls = { "version" = "=11.0.0", default-features = false }
-lance-table = "=11.0.0"
-lance-testing = "=11.0.0"
-lance-datafusion = "=11.0.0"
-lance-encoding = "=11.0.0"
-lance-arrow = "=11.0.0"
+lance = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=12.0.0-beta.2", default-features = false, "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=12.0.0-beta.2", "tag" = "v12.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
 lancedb = { path = "rust/lancedb", default-features = false }
 ahash = "0.8"
 # Note that this one does not include pyarrow

Makefile

@@ -5,5 +5,5 @@ licenses:
 	cd python && cargo about generate ../about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c ../about.toml
 	cd python && uv sync --all-extras && uv tool run pip-licenses --python .venv/bin/python --format=markdown --with-urls --output-file=PYTHON_THIRD_PARTY_LICENSES.md
 	cd nodejs && cargo about generate ../about.hbs -o RUST_THIRD_PARTY_LICENSES.html -c ../about.toml
-	cd nodejs && pnpm dlx license-checker@25 --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
+	cd nodejs && npx license-checker --markdown --out NODEJS_THIRD_PARTY_LICENSES.md
 	cd java && ./mvnw license:aggregate-add-third-party -q

ci/update_lockfiles.sh

@@ -12,12 +12,16 @@ done
 # This updates the lockfile without building
 cargo metadata --quiet > /dev/null
 
+pushd nodejs || exit 1
+npm install --package-lock-only --silent
+popd
+
 if git diff --quiet --exit-code; then
   echo "No lockfile changes to commit; skipping amend."
 elif $AMEND; then
-  git add Cargo.lock
+  git add Cargo.lock nodejs/package-lock.json
   git commit --amend --no-edit
 else
-  git add Cargo.lock
+  git add Cargo.lock nodejs/package-lock.json
   git commit -m "Update lockfiles"
 fi

deny.toml

@@ -131,13 +131,18 @@ allow = [
     "BSD-3-Clause",
     "ISC",
     "Unicode-3.0",
+    "Unicode-DFS-2016",
     "Zlib",
     "CC0-1.0",
     "MPL-2.0",
     "BSL-1.0",
+    "OpenSSL",
     # 0BSD ("BSD Zero Clause") is effectively public domain — no attribution
     # required. Pulled in by `mock_instant`.
     "0BSD",
+    # bzip2-1.0.6 is the permissive upstream bzip2 license (BSD-like). Pulled
+    # in by `libbz2-rs-sys`, the pure-Rust bzip2 implementation.
+    "bzip2-1.0.6",
     # CDLA-Permissive-2.0 is a permissive data license used by `webpki-roots`
     # for the Mozilla CA root bundle. Data-only, distribution-compatible.
     "CDLA-Permissive-2.0",
@@ -145,7 +150,12 @@ allow = [
 confidence-threshold = 0.8
 # Per-crate license exceptions: allow a license for a specific crate only,
 # rather than globally via the `allow` list above.
-exceptions = []
+exceptions = [
+    # CDDL-1.0 (copyleft) is pulled in only as a dev/profiling dependency via
+    # `inferno` -> `pprof` -> `lance-testing`; it is a test dependency that we
+    # do not distribute, so scope the allowance to `inferno` alone.
+    { allow = ["CDDL-1.0"], crate = "inferno" },
+]
 # Crates whose license cannot be determined from Cargo metadata but whose
 # license we've manually confirmed from upstream. Keep this list minimal.
 [[licenses.clarify]]

docs/README.md

@@ -47,24 +47,22 @@ pytest -vv python/tests/docs
 
 ### Checking typescript examples
 
-The examples depend on `@lancedb/lancedb` at `file:../dist`, so the package must be
-built before running the tests. This uses pnpm; see the
-[Typescript contributing guide](../nodejs/CONTRIBUTING.md) for the toolchain setup.
+The `@lancedb/lancedb` package must be built before running the tests:
 
 ```shell
 pushd nodejs
-pnpm install
-pnpm build
+npm ci
+npm run build
 popd
 ```
 
-Then you can run the examples by going to the `nodejs/examples` directory, which is a
-separate pnpm package with its own lockfile:
+Then you can run the examples by going to the `nodejs/examples` directory and
+running the tests like a normal npm package:
 
 ```shell
 pushd nodejs/examples
-pnpm install
-pnpm test
+npm ci
+npm test
 popd
 ```
 
@@ -86,7 +84,6 @@ The new files should be checked into the repository.
 
 ```shell
 pushd nodejs
-# `pnpm docs` would invoke pnpm's built-in `docs` command, not the script.
-pnpm run docs
+npm run docs
 popd
 ```

docs/package-lock.json

@@ -0,0 +1,135 @@
+{
+  "name": "lancedb-docs-test",
+  "version": "1.0.0",
+  "lockfileVersion": 3,
+  "requires": true,
+  "packages": {
+    "": {
+      "name": "lancedb-docs-test",
+      "version": "1.0.0",
+      "license": "Apache 2",
+      "dependencies": {
+        "apache-arrow": "file:../node/node_modules/apache-arrow",
+        "vectordb": "file:../node"
+      },
+      "devDependencies": {
+        "@types/node": "^20.11.8",
+        "typescript": "^5.3.3"
+      }
+    },
+    "../node": {
+      "name": "vectordb",
+      "version": "0.21.2-beta.0",
+      "cpu": [
+        "x64",
+        "arm64"
+      ],
+      "license": "Apache-2.0",
+      "os": [
+        "darwin",
+        "linux",
+        "win32"
+      ],
+      "dependencies": {
+        "@neon-rs/load": "^0.0.74",
+        "axios": "^1.4.0"
+      },
+      "devDependencies": {
+        "@neon-rs/cli": "^0.0.160",
+        "@types/chai": "^4.3.4",
+        "@types/chai-as-promised": "^7.1.5",
+        "@types/mocha": "^10.0.1",
+        "@types/node": "^18.16.2",
+        "@types/sinon": "^10.0.15",
+        "@types/temp": "^0.9.1",
+        "@types/uuid": "^9.0.3",
+        "@typescript-eslint/eslint-plugin": "^5.59.1",
+        "apache-arrow-old": "npm:apache-arrow@13.0.0",
+        "cargo-cp-artifact": "^0.1",
+        "chai": "^4.3.7",
+        "chai-as-promised": "^7.1.1",
+        "eslint": "^8.39.0",
+        "eslint-config-standard-with-typescript": "^34.0.1",
+        "eslint-plugin-import": "^2.26.0",
+        "eslint-plugin-n": "^15.7.0",
+        "eslint-plugin-promise": "^6.1.1",
+        "mocha": "^10.2.0",
+        "openai": "^4.24.1",
+        "sinon": "^15.1.0",
+        "temp": "^0.9.4",
+        "ts-node": "^10.9.1",
+        "ts-node-dev": "^2.0.0",
+        "typedoc": "^0.24.7",
+        "typedoc-plugin-markdown": "^3.15.3",
+        "typescript": "^5.1.0",
+        "uuid": "^9.0.0"
+      },
+      "optionalDependencies": {
+        "@lancedb/vectordb-darwin-arm64": "0.21.2-beta.0",
+        "@lancedb/vectordb-darwin-x64": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-arm64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-linux-x64-gnu": "0.21.2-beta.0",
+        "@lancedb/vectordb-win32-x64-msvc": "0.21.2-beta.0"
+      },
+      "peerDependencies": {
+        "@apache-arrow/ts": "^14.0.2",
+        "apache-arrow": "^14.0.2"
+      }
+    },
+    "../node/node_modules/apache-arrow": {
+      "version": "14.0.2",
+      "license": "Apache-2.0",
+      "dependencies": {
+        "@types/command-line-args": "5.2.0",
+        "@types/command-line-usage": "5.0.2",
+        "@types/node": "20.3.0",
+        "@types/pad-left": "2.1.1",
+        "command-line-args": "5.2.1",
+        "command-line-usage": "7.0.1",
+        "flatbuffers": "23.5.26",
+        "json-bignum": "^0.0.3",
+        "pad-left": "^2.1.0",
+        "tslib": "^2.5.3"
+      },
+      "bin": {
+        "arrow2csv": "bin/arrow2csv.js"
+      }
+    },
+    "node_modules/@types/node": {
+      "version": "20.11.8",
+      "resolved": "https://registry.npmjs.org/@types/node/-/node-20.11.8.tgz",
+      "integrity": "sha512-i7omyekpPTNdv4Jb/Rgqg0RU8YqLcNsI12quKSDkRXNfx7Wxdm6HhK1awT3xTgEkgxPn3bvnSpiEAc7a7Lpyow==",
+      "dev": true,
+      "dependencies": {
+        "undici-types": "~5.26.4"
+      }
+    },
+    "node_modules/apache-arrow": {
+      "resolved": "../node/node_modules/apache-arrow",
+      "link": true
+    },
+    "node_modules/typescript": {
+      "version": "5.3.3",
+      "resolved": "https://registry.npmjs.org/typescript/-/typescript-5.3.3.tgz",
+      "integrity": "sha512-pXWcraxM0uxAS+tN0AG/BF2TyqmHO014Z070UsJ+pFvYuRSq8KH8DmWpnbXe0pEPDHXZV3FcAbJkijJ5oNEnWw==",
+      "dev": true,
+      "bin": {
+        "tsc": "bin/tsc",
+        "tsserver": "bin/tsserver"
+      },
+      "engines": {
+        "node": ">=14.17"
+      }
+    },
+    "node_modules/undici-types": {
+      "version": "5.26.5",
+      "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-5.26.5.tgz",
+      "integrity": "sha512-JlCMO+ehdEIKqlFxk6IfVoAUVmgz7cU7zD/h9XZ0qzeosSHmUJVOzSQvvYSYWXkFXC+IfLKSIffhv0sVZup6pA==",
+      "dev": true
+    },
+    "node_modules/vectordb": {
+      "resolved": "../node",
+      "link": true
+    }
+  }
+}

docs/package.json

@@ -0,0 +1,20 @@
+{
+  "name": "lancedb-docs-test",
+  "version": "1.0.0",
+  "description": "auto-generated tests from doc",
+  "author": "dev@lancedb.com",
+  "license": "Apache 2",
+  "dependencies": {
+    "apache-arrow": "file:../node/node_modules/apache-arrow",
+    "vectordb": "file:../node"
+  },
+  "scripts": {
+    "build": "tsc -b && cd ../node && npm run build-release",
+    "example": "npm run build && node",
+    "test": "npm run build && ls dist/*.js | xargs -n 1 node"
+  },
+  "devDependencies": {
+    "@types/node": "^20.11.8",
+    "typescript": "^5.3.3"
+  }
+}

docs/src/java/java.md

@@ -14,7 +14,7 @@ Add the following dependency to your `pom.xml`:
 <dependency>
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-core</artifactId>
-    <version>0.38.0</version>
+    <version>0.38.0-beta.11</version>
 </dependency>
 ```
 

docs/src/python/python.md

@@ -223,14 +223,10 @@ tokens = list(
 Blob columns store large binary values out of line so they can be read lazily
 instead of being materialized with the rest of the row.
 
-`lancedb.BlobType` is `lance.blob.BlobType` when pylance is installed. Without
-pylance, LanceDB uses a matching `lance.blob.v2` extension type so blob columns
-still work. Queries return descriptors. Call
-[`fetch_blob_files`][lancedb.table.Table.fetch_blob_files] for lazy reads or
-[`fetch_blobs`][lancedb.table.Table.fetch_blobs] for eager bytes.
-
 ::: lancedb.blob
 
+::: lancedb.BlobType
+
 ::: lancedb._blob.BlobFile
     options:
       show_root_full_path: false

docs/tsconfig.json

@@ -0,0 +1,17 @@
+{
+  "include": [
+    "src/*.ts",
+  ],
+  "compilerOptions": {
+    "target": "es2022",
+    "module": "nodenext",
+    "declaration": true,
+    "outDir": "./dist",
+    "strict": true,
+    "allowJs": true,
+    "resolveJsonModule": true,
+  },
+  "exclude": [
+    "./dist/*",
+  ]
+}

java/lancedb-core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
       <groupId>com.lancedb</groupId>
       <artifactId>lancedb-parent</artifactId>
-      <version>0.38.0-final.0</version>
+      <version>0.38.0-beta.11</version>
       <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.38.0-final.0</version>
+    <version>0.38.0-beta.11</version>
     <packaging>pom</packaging>
     <name>${project.artifactId}</name>
     <description>LanceDB Java SDK Parent POM</description>
@@ -28,7 +28,7 @@
     <properties>
         <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
         <arrow.version>15.0.0</arrow.version>
-        <lance-core.version>11.0.0</lance-core.version>
+        <lance-core.version>12.0.0-beta.2</lance-core.version>
         <spotless.skip>false</spotless.skip>
         <spotless.version>2.30.0</spotless.version>
         <spotless.java.googlejavaformat.version>1.7</spotless.java.googlejavaformat.version>

nodejs/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
 name = "lancedb-nodejs"
 edition.workspace = true
-version = "0.38.0"
+version = "0.38.0-beta.11"
 publish = false
 license.workspace = true
 description.workspace = true

nodejs/__test__/arrow.test.ts

@@ -1,7 +1,5 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
-import * as fs from "node:fs";
-import * as vm from "node:vm";
 import * as arrow15 from "apache-arrow-15";
 import * as arrow16 from "apache-arrow-16";
 import * as arrow17 from "apache-arrow-17";
@@ -42,41 +40,6 @@ function sampleRecords(): Array<Record<string, any>> {
   ];
 }
 
-it("serializes an Arrow Table created in another JavaScript realm", async () => {
-  const context = vm.createContext({
-    TextDecoder,
-    TextEncoder,
-    console,
-    setTimeout,
-    clearTimeout,
-  });
-  vm.runInContext(
-    fs.readFileSync(
-      require.resolve("apache-arrow-15/Arrow.es2015.min"),
-      "utf8",
-    ),
-    context,
-  );
-  const foreignTable: unknown = vm.runInContext(
-    "Arrow.tableFromArrays({ id: new Int32Array([1, 2, 3]), text: ['foo', 'bar', 'baz'] })",
-    context,
-  );
-
-  const foreignMetadata = (
-    foreignTable as { schema: { metadata: Map<string, string> } }
-  ).schema.metadata;
-  expect(foreignMetadata).not.toBeInstanceOf(Map);
-
-  const buf = await fromDataToBuffer(
-    foreignTable as Parameters<typeof fromDataToBuffer>[0],
-  );
-  const actual = currentTableFromIPC(buf);
-
-  expect(actual.numRows).toBe(3);
-  expect(actual.getChild("id")?.toJSON()).toEqual([1, 2, 3]);
-  expect(actual.getChild("text")?.toJSON()).toEqual(["foo", "bar", "baz"]);
-});
-
 it("preserves field metadata from a provided schema", async function () {
   const jsonMetadata = new Map([["ARROW:extension:name", "lance.json"]]);
   const schema = new CurrentSchema([

nodejs/__test__/package.test.ts

@@ -5,8 +5,8 @@ import packageJson = require("../package.json");
 
 describe("package metadata", () => {
   it("requires Node.js type declarations compatible with the runtime", () => {
-    expect(packageJson.engines.node).toBe(">= 22");
-    expect(packageJson.peerDependencies["@types/node"]).toBe(">=22");
+    expect(packageJson.engines.node).toBe(">= 18");
+    expect(packageJson.peerDependencies["@types/node"]).toBe(">=18");
     expect(packageJson.peerDependenciesMeta["@types/node"]).toEqual({
       optional: true,
     });

nodejs/__test__/remote.test.ts

@@ -3,7 +3,6 @@
 
 import * as http from "http";
 import { RequestListener } from "http";
-import packageJson = require("../package.json");
 import {
   ClientConfig,
   Connection,
@@ -71,13 +70,7 @@ async function withMockDatabase(
   try {
     await callback(db);
   } finally {
-    // `close()` alone leaves the port bound until keep-alive sockets drain, so
-    // a single failing test would cascade into EADDRINUSE for every test after
-    // it. Destroy the connections and wait for the port to actually be free.
-    await new Promise<void>((resolve) => {
-      server.closeAllConnections();
-      server.close(() => resolve());
-    });
+    server.close();
   }
 }
 
@@ -138,7 +131,7 @@ describe("remote connection", () => {
       (req, res) => {
         expect(req.headers["x-api-key"]).toEqual("fake");
         expect(req.headers["user-agent"]).toEqual(
-          `LanceDB-Node-Client/${packageJson.version}`,
+          `LanceDB-Node-Client/${process.env.npm_package_version}`,
         );
 
         const body = JSON.stringify({ tables: [] });

nodejs/examples/package.json

@@ -8,8 +8,7 @@
     "//1": "--experimental-vm-modules is needed to run jest with sentence-transformers",
     "//2": "--testEnvironment is needed to run jest with sentence-transformers",
     "//3": "See: https://github.com/huggingface/transformers.js/issues/57",
-    "//4": "jest is invoked by its JS entry, not node_modules/.bin/jest: under pnpm that path is a shell shim, which `node` cannot execute",
-    "test": "node --experimental-vm-modules node_modules/jest/bin/jest.js --testEnvironment jest-environment-node-single-context --verbose",
+    "test": "node --experimental-vm-modules node_modules/.bin/jest --testEnvironment jest-environment-node-single-context --verbose",
     "lint": "biome check *.ts && biome format *.ts",
     "lint-ci": "biome ci .",
     "lint-fix": "biome check --write *.ts && pnpm format",

nodejs/lancedb/arrow.ts

@@ -72,7 +72,8 @@ export type FieldLike =
     };
 
 export type DataLike =
-  | import("apache-arrow").Data
+  // biome-ignore lint/suspicious/noExplicitAny: <explanation>
+  | import("apache-arrow").Data<Struct<any>>
   | {
       // biome-ignore lint/suspicious/noExplicitAny: <explanation>
       type: any;
@@ -81,7 +82,6 @@ export type DataLike =
       stride: number;
       nullable: boolean;
       children: DataLike[];
-      dictionary?: { data: readonly DataLike[] };
       get nullCount(): number;
       // biome-ignore lint/suspicious/noExplicitAny: <explanation>
       values: Buffers<any>[BufferType.DATA];

nodejs/lancedb/sanitize.ts

@@ -94,24 +94,17 @@ export function sanitizeMetadata(
   if (metadataLike === undefined || metadataLike === null) {
     return undefined;
   }
-
-  let entries: IterableIterator<[unknown, unknown]>;
-  try {
-    entries = Map.prototype.entries.call(metadataLike);
-  } catch {
+  if (!(metadataLike instanceof Map)) {
     throw Error("Expected metadata, if present, to be a Map<string, string>");
   }
-
-  const metadata = new Map<string, string>();
-  for (const [key, value] of entries) {
-    if (typeof key !== "string" || typeof value !== "string") {
+  for (const item of metadataLike) {
+    if (typeof item[0] !== "string" || typeof item[1] !== "string") {
       throw Error(
         "Expected metadata, if present, to be a Map<string, string> but it had non-string keys or values",
       );
     }
-    metadata.set(key, value);
   }
-  return metadata;
+  return metadataLike as Map<string, string>;
 }
 
 export function sanitizeInt(typeLike: object) {

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.38.0",
+	"version": "0.38.0-beta.11",
 	"os": ["darwin"],
 	"cpu": ["arm64"],
 	"main": "lancedb.darwin-arm64.node",

nodejs/npm/linux-arm64-gnu/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-gnu",
-	"version": "0.38.0",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-gnu.node",

nodejs/npm/linux-arm64-musl/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-arm64-musl",
-	"version": "0.38.0",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["arm64"],
 	"main": "lancedb.linux-arm64-musl.node",

nodejs/npm/linux-x64-gnu/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-gnu",
-	"version": "0.38.0",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-gnu.node",

nodejs/npm/linux-x64-musl/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-linux-x64-musl",
-	"version": "0.38.0",
+	"version": "0.38.0-beta.11",
 	"os": ["linux"],
 	"cpu": ["x64"],
 	"main": "lancedb.linux-x64-musl.node",

nodejs/npm/win32-arm64-msvc/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lancedb/lancedb-win32-arm64-msvc",
-  "version": "0.38.0",
+  "version": "0.38.0-beta.11",
   "os": [
     "win32"
   ],

nodejs/npm/win32-x64-msvc/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.38.0",
+	"version": "0.38.0-beta.11",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.38.0",
+  "version": "0.38.0-beta.11",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",
@@ -67,7 +67,7 @@
     "timeout": "3m"
   },
   "engines": {
-    "node": ">= 22"
+    "node": ">= 18"
   },
   "packageManager": "pnpm@11.1.1",
   "cpu": ["x64", "arm64"],
@@ -101,7 +101,7 @@
     "openai": "4.29.2"
   },
   "peerDependencies": {
-    "@types/node": ">=22",
+    "@types/node": ">=18",
     "apache-arrow": ">=15.0.0 <=18.1.0"
   },
   "peerDependenciesMeta": {

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.38.0"
+version = "0.38.0-beta.11"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"

python/python/lancedb/__init__.py

@@ -6,7 +6,7 @@ import importlib.metadata
 import os
 from concurrent.futures import ThreadPoolExecutor
 from datetime import timedelta
-from typing import Dict, Optional, Union, Any, List, Iterable, TYPE_CHECKING
+from typing import Dict, Optional, Union, Any, List, Iterable
 
 __version__ = importlib.metadata.version("lancedb")
 
@@ -20,7 +20,7 @@ from .db import AsyncConnection, DBConnection, LanceDBConnection
 from .remote import ClientConfig
 from .remote.db import RemoteDBConnection
 from .expr import Expr, col, lit, func
-from .schema import blob, vector
+from .schema import blob, vector, BlobType
 from .job import AsyncJob, Job
 from .functions import (
     FunctionArtifactRequest as FunctionArtifactRequest,
@@ -49,19 +49,6 @@ from .namespace import (
 )
 
 
-if TYPE_CHECKING:
-    from lance.blob import BlobType as BlobType
-
-
-def __getattr__(name: str):
-    if name == "BlobType":
-        from .schema import BlobType
-
-        globals()["BlobType"] = BlobType
-        return BlobType
-    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
-
-
 def _check_s3_bucket_with_dots(
     uri: str, storage_options: Optional[Dict[str, str]]
 ) -> None:

python/python/lancedb/_blob.py

@@ -12,7 +12,7 @@ from typing import TYPE_CHECKING, Optional, Union
 import pyarrow as pa
 
 from .expr import Expr
-from .schema import row_addressable_blob_v2_paths
+from .schema import blob_v2_column_paths
 from .types import BlobMode, QueryProjection, QueryProjectionSpec
 
 if TYPE_CHECKING:
@@ -119,7 +119,7 @@ def blob_v2_projection_sources(
     schema: pa.Schema,
     projection: QueryProjection,
 ) -> dict[str, str]:
-    blob_columns = row_addressable_blob_v2_paths(schema)
+    blob_columns = blob_v2_column_paths(schema)
     if not blob_columns:
         return {}
     columns = set(blob_columns)
@@ -140,9 +140,7 @@ def v2_projection_needs_row_id(
 ) -> bool:
     if with_row_id:
         return False
-    return projection_includes_blob_column(
-        projection, row_addressable_blob_v2_paths(schema)
-    )
+    return projection_includes_blob_column(projection, blob_v2_column_paths(schema))
 
 
 def blob_auto_row_id_for_scan(
@@ -272,8 +270,7 @@ def _iter_projection_pairs(
             if isinstance(expr, str):
                 yield name, expr
             elif isinstance(expr, Expr):
-                source = expr._column_name()
-                yield name, source if source is not None else expr.to_sql()
+                yield name, expr.to_sql()
         return
     for column in projection:
         if isinstance(column, str):
@@ -283,8 +280,7 @@ def _iter_projection_pairs(
             if isinstance(expr, str):
                 yield name, expr
             elif isinstance(expr, Expr):
-                source = expr._column_name()
-                yield name, source if source is not None else expr.to_sql()
+                yield name, expr.to_sql()
 
 
 def _set_blob_column(tbl: pa.Table, output_name: str, blobs: pa.Array) -> pa.Table:

python/python/lancedb/_lancedb.pyi

@@ -87,7 +87,6 @@ class PyExpr:
     def contains(self, substr: "PyExpr") -> "PyExpr": ...
     def isin(self, values: List["PyExpr"]) -> "PyExpr": ...
     def cast(self, data_type: pa.DataType) -> "PyExpr": ...
-    def column_name(self) -> Optional[str]: ...
     def to_sql(self) -> str: ...
 
 def expr_col(name: str) -> PyExpr: ...
@@ -609,7 +608,6 @@ class PyQueryRequest:
     filter: Optional[Union[str, bytes]]
     full_text_search: Optional[FullTextQuery]
     select: Optional[Union[str, List[str]]]
-    select_source_columns: Optional[Dict[str, str]]
     fast_search: Optional[bool]
     with_row_id: Optional[bool]
     use_lsm: Optional[bool]

python/python/lancedb/expr.py

@@ -249,10 +249,6 @@ class Expr:
 
     # ── utilities ────────────────────────────────────────────────────────────
 
-    def _column_name(self) -> str | None:
-        """Return the source name when this is a bare column expression."""
-        return self._inner.column_name()
-
     def to_sql(self) -> str:
         """Render the expression as a SQL string (useful for debugging)."""
         return self._inner.to_sql()
@@ -316,7 +312,7 @@ def func(name: str, *args: ExprLike) -> Expr:
     --------
     >>> from lancedb.expr import col, func
     >>> func("lower", col("name"))
-    Expr(lower(`name`))
+    Expr(lower(name))
     """
     inner_args = [_coerce(a)._inner for a in args]
     return Expr(expr_func(name, inner_args))

python/python/lancedb/functions.py

@@ -49,25 +49,11 @@ from pydantic import (
     model_validator,
 )
 
-from .schema import is_blob_v2_field as _is_blob_v2_field
-
 _Int32 = conint(strict=True, ge=-(2**31), le=2**31 - 1)
 _UInt32 = conint(strict=True, ge=0, le=2**32 - 1)
 _UInt64 = conint(strict=True, ge=0, le=2**64 - 1)
 
 
-def _validate_gpu_wire_marker(value: Any) -> bool:
-    if value is not True:
-        raise ValueError("runtime.gpu must be true")
-    return True
-
-
-def _normalize_gpu_marker(value: bool) -> Optional[bool]:
-    if not isinstance(value, bool):
-        raise ValueError("gpu must be a boolean")
-    return True if value else None
-
-
 class _FrozenDict(dict):
     def _immutable(self, *args, **kwargs):
         raise TypeError("remote canonical values are immutable")
@@ -253,23 +239,6 @@ class PythonRuntimeSpec(_RemoteValue):
     python_version: Optional[str] = None
     environment: Optional[PythonEnvironmentSpec] = None
     env: Optional[Mapping[str, str]] = None
-    gpu: Optional[bool] = None
-
-    @model_validator(mode="before")
-    @classmethod
-    def _discard_unknown_runtime_payload(cls, value):
-        if isinstance(value, Mapping):
-            kind = value.get("kind")
-            if isinstance(kind, str) and kind not in {"python", "python_v2"}:
-                return {"kind": kind}
-        return value
-
-    @field_validator("gpu", mode="before")
-    @classmethod
-    def _validate_gpu_marker(cls, value):
-        if value is None:
-            return None
-        return _validate_gpu_wire_marker(value)
 
     @model_validator(mode="after")
     def _validate_runtime_kind(self):
@@ -278,28 +247,18 @@ class PythonRuntimeSpec(_RemoteValue):
                 raise ValueError("python runtime requires python_version")
             if self.environment is None:
                 raise ValueError("python runtime requires environment")
-            if self.gpu is not None:
-                raise ValueError("python runtime with gpu requires kind='python_v2'")
-        elif self.kind == "python_v2":
-            if self.python_version is None:
-                raise ValueError("python_v2 runtime requires python_version")
-            if self.environment is None:
-                raise ValueError("python_v2 runtime requires environment")
-            if self.gpu is None:
-                raise ValueError("python_v2 runtime requires gpu")
         else:
             object.__setattr__(self, "python_version", None)
             object.__setattr__(self, "environment", None)
             object.__setattr__(self, "env", None)
-            object.__setattr__(self, "gpu", None)
         return self
 
 
 class FunctionVersion(_RemoteValue):
     """An exact immutable Function version returned by Enterprise.
 
-    The GPU execution requirement is part of this identity. CPU and memory sizing,
-    priority, concurrency, and retry policy belong to the execution platform.
+    Scheduling resources, priority, concurrency, and retry policy belong to
+    the submitting Job and are not part of this identity.
     """
 
     name: str
@@ -520,7 +479,6 @@ class RefreshColumnResult(_RemoteValue):
 
 
 _FUNCTION_NAME = re.compile(r"^[A-Za-z_][A-Za-z0-9_.-]*$")
-_FUNCTION_BLOB_V2_TYPE = "blob_v2"
 
 
 _GRAMMAR_PRIMITIVES = (
@@ -537,7 +495,6 @@ _GRAMMAR_PRIMITIVES = (
     (pa.float32(), "float32"),
     (pa.float64(), "float64"),
     (pa.string(), "utf8"),
-    (pa.large_string(), "large_utf8"),
     (pa.binary(), "binary"),
     (pa.date32(), "date32"),
     (pa.date64(), "date64"),
@@ -545,177 +502,31 @@ _GRAMMAR_PRIMITIVES = (
 
 
 def _canonical_arrow_type(data_type: pa.DataType) -> str:
-    """The compact Function grammar, or canonical exact JSON for nested types."""
-    grammar = _grammar_arrow_type(data_type)
-    if grammar is not None:
-        return grammar
-    exact = _exact_arrow_type(data_type)
-    return json.dumps(exact, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
-
-
-def _grammar_arrow_type(data_type: pa.DataType) -> Optional[str]:
+    """The server's V1 Function type grammar. Anything outside it is rejected
+    here rather than at registration."""
     for candidate, name in _GRAMMAR_PRIMITIVES:
         if data_type == candidate:
             return name
     if pa.types.is_list(data_type) or pa.types.is_large_list(data_type):
-        item = _grammar_list_item(data_type)
-        if item is None:
-            return None
         prefix = "list" if pa.types.is_list(data_type) else "large_list"
-        return f"{prefix}<{item}>"
+        return f"{prefix}<{_canonical_list_item(data_type)}>"
     if pa.types.is_fixed_size_list(data_type) and data_type.list_size > 0:
-        item = _grammar_list_item(data_type)
-        if item is not None:
-            return f"fixed_size_list<{item}, {data_type.list_size}>"
-    return None
+        return (
+            f"fixed_size_list<{_canonical_list_item(data_type)}, {data_type.list_size}>"
+        )
+    raise TypeError(f"unsupported Arrow type for Function signature: {data_type}")
 
 
-def _grammar_list_item(data_type: pa.DataType) -> Optional[str]:
+def _canonical_list_item(data_type: pa.DataType) -> str:
     """The grammar names only the item type; it always means a non-nullable
-    child called `item`, so other child properties require exact JSON."""
+    child called `item`, so any other child metadata cannot be represented."""
     child = data_type.value_field
     if child.name != "item" or child.nullable or child.metadata:
-        return None
-    return _grammar_arrow_type(child.type)
-
-
-def _validate_exact_arrow_field(field: pa.Field) -> None:
-    if not field.name:
         raise TypeError(
-            "unsupported Arrow type for Function signature: field names "
-            "must not be empty"
+            "unsupported Arrow type for Function signature: list items must be a "
+            f"non-nullable field named 'item', got {child}"
         )
-    if _is_blob_v2_field(field):
-        if not _has_supported_blob_v2_layout(field):
-            raise TypeError(
-                "unsupported Arrow type for Function signature: lance.blob.v2 "
-                f"requires a supported Blob storage layout, got {field}"
-            )
-    elif field.metadata:
-        raise TypeError(
-            "unsupported Arrow type for Function signature: field metadata "
-            f"is not supported, got {field}"
-        )
-
-
-def _has_supported_blob_v2_layout(field: pa.Field) -> bool:
-    data_type = field.type
-    if isinstance(data_type, pa.ExtensionType):
-        data_type = data_type.storage_type
-    if not pa.types.is_struct(data_type):
-        return False
-
-    fields = tuple(data_type)
-
-    def matches(spec, compare_nullable) -> bool:
-        return len(fields) == len(spec) and all(
-            actual.name == name
-            and actual.type == expected_type
-            and (not check_nullable or actual.nullable == nullable)
-            for actual, (name, expected_type, nullable), check_nullable in zip(
-                fields, spec, compare_nullable
-            )
-        )
-
-    logical_minimal = (
-        ("data", pa.large_binary(), True),
-        ("uri", pa.utf8(), True),
-    )
-    logical_full = logical_minimal + (
-        ("position", pa.uint64(), True),
-        ("size", pa.uint64(), True),
-    )
-    prepared = (
-        ("kind", pa.uint8(), True),
-        ("data", pa.large_binary(), True),
-        ("uri", pa.utf8(), True),
-        ("blob_id", pa.uint32(), True),
-        ("blob_size", pa.uint64(), True),
-        ("position", pa.uint64(), True),
-    )
-    descriptor = (
-        ("kind", pa.uint8(), False),
-        ("position", pa.uint64(), False),
-        ("size", pa.uint64(), False),
-        ("blob_id", pa.uint32(), False),
-        ("blob_uri", pa.utf8(), False),
-    )
-    return (
-        matches(logical_minimal, (True, True))
-        or matches(logical_full, (True, True, False, False))
-        or matches(prepared, (True,) * len(prepared))
-        or matches(descriptor, (False,) * len(descriptor))
-    )
-
-
-def _canonical_arrow_field(field: pa.Field) -> str:
-    _validate_exact_arrow_field(field)
-    if _is_blob_v2_field(field):
-        return _FUNCTION_BLOB_V2_TYPE
-    return _canonical_arrow_type(field.type)
-
-
-def _exact_arrow_field(field: pa.Field) -> dict[str, Any]:
-    _validate_exact_arrow_field(field)
-    if _is_blob_v2_field(field):
-        raise TypeError(
-            "unsupported Arrow type for Function signature: nested Blob v2 "
-            "fields are not supported; declare Blob parameters or named result "
-            "fields directly"
-        )
-    value = {
-        "name": field.name,
-        "nullable": field.nullable,
-        "type": _exact_arrow_type(field.type),
-    }
-    return value
-
-
-def _exact_arrow_type(data_type: pa.DataType) -> dict[str, Any]:
-    for candidate, name in _GRAMMAR_PRIMITIVES:
-        if data_type == candidate:
-            return {"type": name}
-    if pa.types.is_struct(data_type):
-        fields = list(data_type)
-        names = [field.name for field in fields]
-        if not fields or len(set(names)) != len(names):
-            raise TypeError(
-                "unsupported Arrow type for Function signature: structs must have "
-                "non-empty, uniquely named fields"
-            )
-        return {
-            "type": "struct",
-            "fields": [_exact_arrow_field(field) for field in fields],
-        }
-    if (
-        pa.types.is_list(data_type)
-        or pa.types.is_large_list(data_type)
-        or pa.types.is_fixed_size_list(data_type)
-    ):
-        if pa.types.is_fixed_size_list(data_type):
-            if data_type.value_field.name != "item":
-                raise TypeError(
-                    "unsupported Arrow type for Function signature: fixed-size list "
-                    "items must be named 'item'"
-                )
-            if data_type.list_size <= 0:
-                raise TypeError(
-                    f"unsupported Arrow type for Function signature: {data_type}"
-                )
-        value: dict[str, Any] = {
-            "type": (
-                "list"
-                if pa.types.is_list(data_type)
-                else "large_list"
-                if pa.types.is_large_list(data_type)
-                else "fixed_size_list"
-            ),
-            "fields": [_exact_arrow_field(data_type.value_field)],
-        }
-        if pa.types.is_fixed_size_list(data_type):
-            value["length"] = data_type.list_size
-        return value
-    raise TypeError(f"unsupported Arrow type for Function signature: {data_type}")
+    return _canonical_arrow_type(child.type)
 
 
 def _list_of(item: pa.DataType) -> pa.DataType:
@@ -789,15 +600,8 @@ def _callable_parameters(function: Callable[..., Any]) -> tuple[inspect.Paramete
 
 def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutput:
     if isinstance(output, pa.Schema):
-        if output.metadata:
-            raise TypeError("Function output schema metadata is not supported")
         fields = tuple(output)
-    elif (
-        isinstance(output, pa.Field)
-        and not _is_blob_v2_field(output)
-        and pa.types.is_struct(output.type)
-    ):
-        _validate_exact_arrow_field(output)
+    elif isinstance(output, pa.Field) and pa.types.is_struct(output.type):
         if output.nullable:
             raise ValueError("Function output must be non-nullable")
         fields = tuple(output.type)
@@ -813,12 +617,11 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
             raise TypeError(
                 "output_schema must be a PyArrow DataType, Field, or Schema"
             )
-        _validate_exact_arrow_field(field)
         if field.nullable:
             raise ValueError("Function output must be non-nullable")
         return FunctionOutput(
             kind="scalar",
-            arrow_type=_canonical_arrow_field(field),
+            arrow_type=_canonical_arrow_type(field.type),
             nullable=False,
         )
 
@@ -826,8 +629,6 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
         raise ValueError("named-struct Function output must contain at least one field")
     if any(field.nullable for field in fields):
         raise ValueError("Function output fields must be non-nullable")
-    for field in fields:
-        _validate_exact_arrow_field(field)
     names = [field.name for field in fields]
     if len(set(names)) != len(names):
         raise ValueError("Function output field names must be unique")
@@ -836,7 +637,7 @@ def _function_output(output: pa.DataType | pa.Field | pa.Schema) -> FunctionOutp
         fields=tuple(
             FunctionResultField(
                 name=field.name,
-                arrow_type=_canonical_arrow_field(field),
+                arrow_type=_canonical_arrow_type(field.type),
                 nullable=False,
             )
             for field in fields
@@ -856,10 +657,6 @@ def _infer_signature(
     if input_schema is not None:
         if not isinstance(input_schema, pa.Schema):
             raise TypeError("input_schema must be a PyArrow Schema")
-        if input_schema.metadata:
-            raise TypeError("Function input schema metadata is not supported")
-        for field in input_schema:
-            _validate_exact_arrow_field(field)
         expected = tuple(parameter.name for parameter in parameters)
         actual = tuple(input_schema.names)
         if actual != expected:
@@ -870,7 +667,7 @@ def _infer_signature(
         inputs = tuple(
             FunctionParameter(
                 name=field.name,
-                arrow_type=_canonical_arrow_field(field),
+                arrow_type=_canonical_arrow_type(field.type),
                 nullable=field.nullable,
             )
             for field in input_schema
@@ -893,9 +690,7 @@ def _infer_signature(
         inputs.append(
             FunctionParameter(
                 name=parameter.name,
-                arrow_type=_canonical_arrow_field(
-                    pa.field(parameter.name, data_type, nullable=nullable)
-                ),
+                arrow_type=_canonical_arrow_type(data_type),
                 nullable=nullable,
             )
         )
@@ -1115,7 +910,6 @@ class UdfDefinition:
         pip: tuple[str, ...],
         env: Mapping[str, str],
         python_version: Optional[str],
-        gpu: bool = False,
         conda: tuple[str, ...] = (),
         conda_channels: tuple[str, ...] = (),
     ):
@@ -1144,14 +938,12 @@ class UdfDefinition:
         signature = _infer_signature(function, input_schema, output_schema)
         source = _package_source(function)
         digest = f"sha256:{hashlib.sha256(source).hexdigest()}"
-        gpu_marker = _normalize_gpu_marker(gpu)
         runtime = PythonRuntimeSpec(
-            kind="python_v2" if gpu_marker is not None else "python",
+            kind="python",
             python_version=python_version
             or f"{sys.version_info.major}.{sys.version_info.minor}",
             environment=environment_spec,
             env=environment,
-            gpu=gpu_marker,
         )
         self._function = function
         self._request = FunctionRegistrationRequest(
@@ -1197,7 +989,6 @@ def udf(
     pip: tuple[str, ...] | list[str] = (),
     env: Optional[Mapping[str, str]] = None,
     python_version: Optional[str] = None,
-    gpu: bool = False,
     conda: tuple[str, ...] | list[str] = (),
     conda_channels: tuple[str, ...] | list[str] = (),
 ) -> Callable[[Callable[..., Any]], UdfDefinition]: ...
@@ -1212,7 +1003,6 @@ def udf(
     pip: tuple[str, ...] | list[str] = (),
     env: Optional[Mapping[str, str]] = None,
     python_version: Optional[str] = None,
-    gpu: bool = False,
     conda: tuple[str, ...] | list[str] = (),
     conda_channels: tuple[str, ...] | list[str] = (),
 ):
@@ -1245,10 +1035,6 @@ def udf(
         Environment variables included in the Function definition.
     python_version : str, optional
         Remote Python major/minor version. Defaults to the client version.
-    gpu : bool, default False
-        Whether every remote execution requires a GPU. The execution platform
-        selects one compatible GPU for each worker. The requirement is part of
-        the immutable Function version.
 
     The packaged artifact is a snapshot: the function source plus exactly
     the module-level names it references (modules as imports, importable
@@ -1273,11 +1059,6 @@ def udf(
     ...     return value * 2
     >>> score(1.5)
     3.0
-    >>> @udf(pip=["cupy-cuda12x"], gpu=True)
-    ... def gpu_score(value: int) -> int:
-    ...     return value * 2
-    >>> gpu_score.registration_request.runtime.gpu
-    True
     """
 
     def decorate(target: Callable[..., Any]) -> UdfDefinition:
@@ -1289,7 +1070,6 @@ def udf(
             pip=tuple(pip),
             env={} if env is None else env,
             python_version=python_version,
-            gpu=gpu,
             conda=tuple(conda),
             conda_channels=tuple(conda_channels),
         )

python/python/lancedb/query.py

@@ -167,12 +167,6 @@ def _projection_to_scanner_kwargs(columns: QueryProjection) -> Dict[str, Any]:
     return {"columns": projection}
 
 
-def _query_request_projection(req: "PyQueryRequest") -> QueryProjection:
-    if req.select_source_columns is not None:
-        return req.select_source_columns
-    return req.select
-
-
 def _scanner_kwargs_for_query(
     query: Query,
     blob_mode: BlobMode,
@@ -2805,16 +2799,15 @@ class AsyncQueryBase(object):
 
         req = self._inner.to_query_request()
         schema = await self._table.schema()
-        projection = _query_request_projection(req)
         self._blob_auto_row_id = blob_auto_row_id_for_scan(
             schema,
-            projection,
+            req.select,
             with_row_id=self._with_row_id,
         )
         if not self._blob_auto_row_id:
             self._blob_paths = ()
             return
-        self._blob_paths = tuple(blob_v2_projection_sources(schema, projection).keys())
+        self._blob_paths = tuple(blob_v2_projection_sources(schema, req.select).keys())
         self._inner.with_row_id()
 
     def select(self, columns: Union[List[str], dict[str, str]]) -> Self:
@@ -3901,15 +3894,14 @@ class AsyncHybridQuery(AsyncStandardQuery, AsyncVectorQueryBase):
         blob_paths: tuple[str, ...] = ()
         if self._table is not None:
             schema = await self._table.schema()
-            projection = _query_request_projection(req)
             blob_auto_row_id = blob_auto_row_id_for_scan(
                 schema,
-                projection,
+                req.select,
                 with_row_id=self._with_row_id,
             )
             if blob_auto_row_id:
                 blob_paths = tuple(
-                    blob_v2_projection_sources(schema, projection).keys()
+                    blob_v2_projection_sources(schema, req.select).keys()
                 )
         self._blob_auto_row_id = blob_auto_row_id
         self._blob_paths = blob_paths

python/python/lancedb/remote/table.py

@@ -36,7 +36,6 @@ from lancedb._lancedb import (
     UpdateResult,
 )
 from lancedb.embeddings.base import EmbeddingFunctionConfig
-from lancedb.expr import Expr
 from lancedb.index import (
     FTS,
     BTree,
@@ -67,15 +66,7 @@ from ..query import (
     LanceTakeQueryBuilder,
     LanceVectorQueryBuilder,
 )
-from ..table import (
-    AsyncTable,
-    BlobMode,
-    Branches,
-    IndexStatistics,
-    Query,
-    Table,
-    Tags,
-)
+from ..table import AsyncTable, BlobMode, Branches, IndexStatistics, Query, Table, Tags
 from ..types import BaseTokenizerType
 
 
@@ -872,7 +863,7 @@ class RemoteTable(Table):
 
     def update(
         self,
-        where: Optional[Union[str, Expr]] = None,
+        where: Optional[str] = None,
         values: Optional[dict] = None,
         *,
         values_sql: Optional[Dict[str, str]] = None,
@@ -883,11 +874,9 @@ class RemoteTable(Table):
 
         Parameters
         ----------
-        where: str or [Expr][lancedb.expr.Expr], optional
-            The filter condition. Can be a SQL string or a type-safe
-            [Expr][lancedb.expr.Expr] built with [col][lancedb.expr.col] and
-            [lit][lancedb.expr.lit]. The filter must not be empty, or it will
-            error.
+        where: str, optional
+            The SQL where clause to use when updating rows. For example, 'x = 2'
+            or 'x IN (1, 2, 3)'. The filter must not be empty, or it will error.
         values: dict, optional
             The values to update. The keys are the column names and the values
             are the values to set.

python/python/lancedb/schema.py

@@ -4,34 +4,30 @@
 
 """Schema helpers for Lance blob columns."""
 
-import importlib
-from typing import TYPE_CHECKING
-
 import pyarrow as pa
-import pyarrow.ipc
-
-if TYPE_CHECKING:
-    from lance.blob import BlobType as BlobType
 
 _BLOB_EXTENSION_NAME = "lance.blob.v2"
 _BLOB_V1_KEY = "lance-encoding:blob"
 _ARROW_EXT_NAME_KEY = "ARROW:extension:name"
-_BLOB_V2_STORAGE_TYPE = pa.struct(
-    [
-        pa.field("data", pa.large_binary(), nullable=True),
-        pa.field("uri", pa.utf8(), nullable=True),
-        pa.field("position", pa.uint64(), nullable=True),
-        pa.field("size", pa.uint64(), nullable=True),
-    ]
-)
-_resolved_blob_type = None
 
 
-class _FallbackBlobType(pa.ExtensionType):
-    """lance.blob.v2 extension type used when pylance is not installed."""
+class BlobType(pa.ExtensionType):
+    """PyArrow extension type for a Lance blob v2 column.
+
+    Queries return descriptors; call :meth:`~lancedb.table.Table.fetch_blob_files`
+    for lazy reads or :meth:`~lancedb.table.Table.fetch_blobs` for eager bytes.
+    """
 
     def __init__(self) -> None:
-        pa.ExtensionType.__init__(self, _BLOB_V2_STORAGE_TYPE, _BLOB_EXTENSION_NAME)
+        storage_type = pa.struct(
+            [
+                pa.field("data", pa.large_binary(), nullable=True),
+                pa.field("uri", pa.utf8(), nullable=True),
+                pa.field("position", pa.uint64(), nullable=True),
+                pa.field("size", pa.uint64(), nullable=True),
+            ]
+        )
+        super().__init__(storage_type, _BLOB_EXTENSION_NAME)
 
     def __arrow_ext_serialize__(self) -> bytes:
         return b""
@@ -39,16 +35,23 @@ class _FallbackBlobType(pa.ExtensionType):
     @classmethod
     def __arrow_ext_deserialize__(
         cls, storage_type: pa.DataType, serialized: bytes
-    ) -> "_FallbackBlobType":
+    ) -> "BlobType":
         return cls()
 
     def __reduce__(self):
+        # Ensure pickle round-trips on older pyarrow (apache/arrow#35599).
         return type(self).__arrow_ext_deserialize__, (
             self.storage_type,
             self.__arrow_ext_serialize__(),
         )
 
 
+try:
+    pa.register_extension_type(BlobType())  # type: ignore[arg-type]
+except pa.ArrowKeyError:
+    pass
+
+
 def _metadata_value(metadata: dict, key: str):
     return metadata.get(key.encode()) or metadata.get(key)
 
@@ -89,105 +92,43 @@ def is_blob_like_field(field: pa.Field) -> bool:
     return is_blob_v2_field(field) or _metadata_marks_legacy_blob(field.metadata or {})
 
 
-def _collect_blob_paths(schema: pa.Schema, is_blob) -> list[tuple[str, bool]]:
-    """Walk the schema and return (path, has_list_ancestor) for each blob field."""
-    paths: list[tuple[str, bool]] = []
+def _collect_blob_paths(schema: pa.Schema, is_blob) -> list[str]:
+    paths: list[str] = []
 
-    def walk(fields, prefix: str, has_list_ancestor: bool) -> None:
+    def walk(fields, prefix: str) -> None:
         for field in fields:
             path = f"{prefix}.{field.name}" if prefix else field.name
             if is_blob(field):
-                paths.append((path, has_list_ancestor))
+                paths.append(path)
             elif pa.types.is_struct(field.type):
-                walk(field.type, path, has_list_ancestor)
+                walk(field.type, path)
             elif (
                 pa.types.is_list(field.type)
                 or pa.types.is_large_list(field.type)
                 or pa.types.is_fixed_size_list(field.type)
             ):
-                walk([field.type.value_field], path, True)
+                walk([field.type.value_field], path)
 
-    walk(schema, "", False)
+    walk(schema, "")
     return paths
 
 
 def blob_column_paths(schema: pa.Schema) -> list[str]:
     """Dotted paths of blob-like columns (v2 extension or legacy metadata)."""
-    return [path for path, _ in _collect_blob_paths(schema, is_blob_like_field)]
+    return _collect_blob_paths(schema, is_blob_like_field)
 
 
 def blob_v2_column_paths(schema: pa.Schema) -> list[str]:
-    return [path for path, _ in _collect_blob_paths(schema, is_blob_v2_field)]
-
-
-def row_addressable_blob_v2_paths(schema: pa.Schema) -> list[str]:
-    """Blob v2 paths with one blob addressable by table row id.
-
-    ``fetch_blobs`` and the descriptor row-id ride-along address one blob per
-    row, so a blob inside a list container has no row-id slot and no fetch
-    path. Those columns still store and query as raw descriptors.
-    """
-    return [
-        path
-        for path, has_list_ancestor in _collect_blob_paths(schema, is_blob_v2_field)
-        if not has_list_ancestor
-    ]
+    return _collect_blob_paths(schema, is_blob_v2_field)
 
 
 def schema_has_blob_field(schema: pa.Schema) -> bool:
     return bool(blob_column_paths(schema))
 
 
-def _deserialize_registered_type(extension_type: pa.ExtensionType) -> pa.DataType:
-    """Return the type Arrow reconstructs for this extension name."""
-    schema = pa.schema([pa.field("value", extension_type)])
-    restored = pa.ipc.read_schema(schema.serialize())
-    return restored.field("value").type
-
-
-def _resolve_blob_type():
-    """Return the BlobType class this process should use.
-
-    pylance's class when it owns the lance.blob.v2 registry entry,
-    otherwise LanceDB's fallback. A different registered class is an error.
-    """
-    global _resolved_blob_type
-    if _resolved_blob_type is not None:
-        return _resolved_blob_type
-    try:
-        blob_module = importlib.import_module("lance.blob")
-    except ModuleNotFoundError as err:
-        if err.name not in ("lance", "lance.blob"):
-            raise
-    else:
-        blob_type = getattr(blob_module, "BlobType", None)
-        if blob_type is not None:
-            registered_type = _deserialize_registered_type(blob_type())
-            if type(registered_type) is not blob_type:
-                registered_cls = type(registered_type)
-                raise ValueError(
-                    "lance.blob.v2 is already registered by "
-                    f"{registered_cls.__module__}.{registered_cls.__qualname__}"
-                )
-            _resolved_blob_type = blob_type
-            return blob_type
-    try:
-        pa.register_extension_type(_FallbackBlobType())  # type: ignore[arg-type]
-    except pa.ArrowKeyError as err:
-        raise ValueError(
-            "lance.blob.v2 is already registered by another extension class"
-        ) from err
-    _resolved_blob_type = _FallbackBlobType
-    return _resolved_blob_type
-
-
 def blob(name: str, nullable: bool = True) -> pa.Field:
-    """Create a Lance blob v2 column field.
-
-    When pylance is installed this is ``lance.blob.BlobType``.
-    """
-    blob_type = _resolve_blob_type()
-    return pa.field(name, blob_type(), nullable=nullable)
+    """Create a Lance blob v2 column field."""
+    return pa.field(name, BlobType(), nullable=nullable)
 
 
 def vector(dimension: int, value_type: pa.DataType = pa.float32()) -> pa.DataType:
@@ -214,11 +155,3 @@ def vector(dimension: int, value_type: pa.DataType = pa.float32()) -> pa.DataTyp
     ... ])
     """
     return pa.list_(value_type, dimension)
-
-
-def __getattr__(name: str):
-    if name == "BlobType":
-        blob_type = _resolve_blob_type()
-        globals()["BlobType"] = blob_type
-        return blob_type
-    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

python/python/lancedb/table.py

@@ -104,12 +104,7 @@ from .util import (
     value_to_sql,
 )
 from .index import lang_mapping
-from .schema import (
-    blob_v2_column_paths,
-    is_blob_v2_field,
-    row_addressable_blob_v2_paths,
-    schema_has_blob_field,
-)
+from .schema import blob_v2_column_paths, schema_has_blob_field
 
 
 def _should_push_down_query_table(
@@ -431,7 +426,6 @@ def _cast_to_target_schema(
 
     def gen():
         for batch in reader:
-            batch = _coerce_blob_write_columns(batch, reordered_schema)
             # Table but not RecordBatch has cast.
             cast_batches = (
                 pa.Table.from_batches([batch]).cast(reordered_schema).to_batches()
@@ -444,166 +438,6 @@ def _cast_to_target_schema(
     return pa.RecordBatchReader.from_batches(reordered_schema, gen())
 
 
-def _coerce_blob_write_columns(
-    batch: pa.RecordBatch, target_schema: pa.Schema
-) -> pa.RecordBatch:
-    """Materialize blob storage structs before the stream leaves Python.
-
-    merge_insert requires its source reader to already match the table's
-    physical schema. Unlike add and insert, it does not pass through
-    LanceDB's Rust blob coercion, so preserving binary input here would
-    reach Lance as binary and fail the schema check.
-    """
-    columns = []
-    fields = []
-    changed = False
-    for field, column in zip(batch.schema, batch.columns):
-        target_field = target_schema.field(field.name)
-        coerced = _coerce_blob_value(column, target_field)
-        if coerced is not column:
-            column = coerced
-            field = pa.field(
-                field.name,
-                coerced.type,
-                field.nullable,
-                target_field.metadata,
-            )
-            changed = True
-        columns.append(column)
-        fields.append(field)
-    if not changed:
-        return batch
-    return pa.RecordBatch.from_arrays(
-        columns, schema=pa.schema(fields, metadata=batch.schema.metadata)
-    )
-
-
-def _coerce_blob_value(column: pa.Array, target_field: pa.Field) -> pa.Array:
-    if is_blob_v2_field(target_field) and _can_coerce_to_blob(column.type):
-        return _coerce_value_to_blob(column, target_field)
-
-    target_type = target_field.type
-    if pa.types.is_struct(target_type) and pa.types.is_struct(column.type):
-        children = []
-        fields = []
-        changed = False
-        for source_field in column.type:
-            source_column = column.field(source_field.name)
-            nested_target = next(
-                (field for field in target_type if field.name == source_field.name),
-                None,
-            )
-            if nested_target is None:
-                children.append(source_column)
-                fields.append(source_field)
-                continue
-            coerced = _coerce_blob_value(source_column, nested_target)
-            if coerced is not source_column:
-                changed = True
-            child_array, child_type = _physical_array_and_type(coerced)
-            children.append(child_array)
-            fields.append(
-                pa.field(
-                    source_field.name,
-                    child_type,
-                    source_field.nullable,
-                    nested_target.metadata,
-                )
-            )
-        if not changed:
-            return column
-        return pa.StructArray.from_arrays(
-            children,
-            fields=fields,
-            mask=column.is_null() if column.null_count else None,
-        )
-
-    if _is_list_like(target_type) and _is_list_like(column.type):
-        return _coerce_blob_list_values(column, target_type.value_field)
-
-    return column
-
-
-def _coerce_blob_list_values(
-    column: pa.Array, target_value_field: pa.Field
-) -> pa.Array:
-    """Coerce blob values inside a list column, preserving offsets and nulls.
-
-    Works on the raw child values window instead of ``pc.list_flatten`` because
-    flatten drops values spanned by null slots, which would misalign offsets.
-    """
-    mask = column.is_null() if column.null_count else None
-    if pa.types.is_fixed_size_list(column.type):
-        list_size = column.type.list_size
-        values = column.values.slice(column.offset * list_size, len(column) * list_size)
-        coerced = _coerce_blob_value(values, target_value_field)
-        if coerced is values:
-            return column
-        physical_values, _ = _physical_array_and_type(coerced)
-        return pa.FixedSizeListArray.from_arrays(physical_values, list_size, mask=mask)
-    offsets = column.offsets
-    first_offset = offsets[0].as_py()
-    values = column.values.slice(
-        first_offset,
-        offsets[-1].as_py() - first_offset,
-    )
-    coerced = _coerce_blob_value(values, target_value_field)
-    if coerced is values:
-        return column
-    physical_values, _ = _physical_array_and_type(coerced)
-    if first_offset:
-        offsets = pc.subtract(offsets, pa.scalar(first_offset, offsets.type))
-    if pa.types.is_large_list(column.type):
-        return pa.LargeListArray.from_arrays(offsets, physical_values, mask=mask)
-    return pa.ListArray.from_arrays(offsets, physical_values, mask=mask)
-
-
-def _coerce_value_to_blob(values: pa.Array, target_field: pa.Field) -> pa.Array:
-    if pa.types.is_null(values.type):
-        data = pa.nulls(len(values), type=pa.large_binary())
-    elif pa.types.is_large_binary(values.type):
-        data = values
-    else:
-        data = values.cast(pa.large_binary())
-    length = len(values)
-    storage_type = target_field.type
-    if isinstance(storage_type, pa.ExtensionType):
-        storage_type = storage_type.storage_type
-    storage_fields = list(storage_type)
-    children = []
-    for storage_field in storage_fields:
-        if storage_field.name == "data":
-            children.append(data)
-        else:
-            children.append(pa.nulls(length, type=storage_field.type))
-    storage = pa.StructArray.from_arrays(
-        children,
-        fields=storage_fields,
-        mask=values.is_null() if values.null_count else None,
-    )
-    if isinstance(target_field.type, pa.ExtensionType):
-        return pa.ExtensionArray.from_storage(target_field.type, storage)
-    return storage
-
-
-def _physical_array_and_type(array: pa.Array) -> tuple[pa.Array, pa.DataType]:
-    if isinstance(array.type, pa.ExtensionType):
-        return array.storage, array.type.storage_type
-    return array, array.type
-
-
-def _can_coerce_to_blob(data_type: pa.DataType) -> bool:
-    return _is_binary_like(data_type) or pa.types.is_null(data_type)
-
-
-def _is_binary_like(data_type: pa.DataType) -> bool:
-    return (
-        pa.types.is_binary(data_type)
-        or pa.types.is_large_binary(data_type)
-        or pa.types.is_binary_view(data_type)
-    )
-
-
 def _field_extension_name(field: pa.Field) -> Optional[str]:
     extension_name = getattr(field.type, "extension_name", None)
     if extension_name is not None:
@@ -630,73 +464,65 @@ def _align_field_types(
         target_field = next((f for f in target_fields if f.name == field.name), None)
         if target_field is None:
             raise ValueError(f"Field '{field.name}' not found in target schema")
-        new_fields.append(_align_field(field, target_field))
+        # Preserve arrow.json input until it reaches Lance. LanceDB exposes stored
+        # JSON columns as lance.json (JSONB-backed LargeBinary), but casting the
+        # input to that storage type here merely relabels the raw JSON bytes as
+        # JSONB. Lance must see arrow.json so it can perform the JSONB encoding.
+        if (
+            _field_extension_name(field) == "arrow.json"
+            and _field_extension_name(target_field) == "lance.json"
+        ):
+            new_fields.append(field)
+            continue
+        if pa.types.is_struct(target_field.type):
+            if pa.types.is_struct(field.type):
+                new_type = pa.struct(
+                    _align_field_types(
+                        field.type.fields,
+                        target_field.type.fields,
+                    )
+                )
+            else:
+                new_type = target_field.type
+        elif pa.types.is_list(target_field.type):
+            if _is_list_like(field.type):
+                new_type = pa.list_(
+                    _align_field_types(
+                        [field.type.value_field],
+                        [target_field.type.value_field],
+                    )[0]
+                )
+            else:
+                new_type = target_field.type
+        elif pa.types.is_large_list(target_field.type):
+            if _is_list_like(field.type):
+                new_type = pa.large_list(
+                    _align_field_types(
+                        [field.type.value_field],
+                        [target_field.type.value_field],
+                    )[0]
+                )
+            else:
+                new_type = target_field.type
+        elif pa.types.is_fixed_size_list(target_field.type):
+            if _is_list_like(field.type):
+                new_type = pa.list_(
+                    _align_field_types(
+                        [field.type.value_field],
+                        [target_field.type.value_field],
+                    )[0],
+                    target_field.type.list_size,
+                )
+            else:
+                new_type = target_field.type
+        else:
+            new_type = target_field.type
+        new_fields.append(
+            pa.field(field.name, new_type, field.nullable, target_field.metadata)
+        )
     return new_fields
 
 
-def _align_list_value_field(
-    value_field: pa.Field, target_value_field: pa.Field
-) -> pa.Field:
-    # A list has exactly one child, so the inferred child name ("item") aligns
-    # positionally and adopts the table's child name; pa.Table.cast renames it.
-    return _align_field(value_field, target_value_field).with_name(
-        target_value_field.name
-    )
-
-
-def _align_field(field: pa.Field, target_field: pa.Field) -> pa.Field:
-    # Preserve arrow.json input until it reaches Lance. LanceDB exposes stored
-    # JSON columns as lance.json (JSONB-backed LargeBinary), but casting the
-    # input to that storage type here merely relabels the raw JSON bytes as
-    # JSONB. Lance must see arrow.json so it can perform the JSONB encoding.
-    if (
-        _field_extension_name(field) == "arrow.json"
-        and _field_extension_name(target_field) == "lance.json"
-    ):
-        return field
-    if pa.types.is_struct(target_field.type):
-        if pa.types.is_struct(field.type):
-            new_type = pa.struct(
-                _align_field_types(
-                    field.type.fields,
-                    target_field.type.fields,
-                )
-            )
-        else:
-            new_type = target_field.type
-    elif pa.types.is_list(target_field.type):
-        if _is_list_like(field.type):
-            new_type = pa.list_(
-                _align_list_value_field(
-                    field.type.value_field, target_field.type.value_field
-                )
-            )
-        else:
-            new_type = target_field.type
-    elif pa.types.is_large_list(target_field.type):
-        if _is_list_like(field.type):
-            new_type = pa.large_list(
-                _align_list_value_field(
-                    field.type.value_field, target_field.type.value_field
-                )
-            )
-        else:
-            new_type = target_field.type
-    elif pa.types.is_fixed_size_list(target_field.type):
-        if _is_list_like(field.type):
-            new_type = pa.list_(
-                _align_list_value_field(
-                    field.type.value_field, target_field.type.value_field
-                ),
-                target_field.type.list_size,
-            )
-        else:
-            new_type = target_field.type
-    else:
-        new_type = target_field.type
-    return pa.field(field.name, new_type, field.nullable, target_field.metadata)
-
-
 def _infer_subschema(
     schema: List[pa.Field],
     reference_fields: List[pa.Field],
@@ -763,7 +589,7 @@ def sanitize_create_table(
         schema = data.schema
     else:
         if schema is not None:
-            data = pa.Table.from_batches([], schema=schema)
+            data = pa.Table.from_pylist([], schema)
     if schema is None:
         if data is None:
             raise ValueError("Either data or schema must be provided")
@@ -1918,7 +1744,7 @@ class Table(ABC):
     @abstractmethod
     def update(
         self,
-        where: Optional[Union[str, Expr]] = None,
+        where: Optional[str] = None,
         values: Optional[dict] = None,
         *,
         values_sql: Optional[Dict[str, str]] = None,
@@ -1933,11 +1759,9 @@ class Table(ABC):
 
         Parameters
         ----------
-        where: str or [Expr][lancedb.expr.Expr], optional
-            The filter condition. Can be a SQL string or a type-safe
-            [Expr][lancedb.expr.Expr] built with [col][lancedb.expr.col] and
-            [lit][lancedb.expr.lit]. The filter must not be empty, or it will
-            error.
+        where: str, optional
+            The SQL where clause to use when updating rows. For example, 'x = 2'
+            or 'x IN (1, 2, 3)'. The filter must not be empty, or it will error.
         values: dict, optional
             The values to update. The keys are the column names and the values
             are the values to set.
@@ -1955,7 +1779,6 @@ class Table(ABC):
         Examples
         --------
         >>> import lancedb
-        >>> from lancedb.expr import col
         >>> import pandas as pd
         >>> data = pd.DataFrame({"x": [1, 2, 3], "vector": [[1.0, 2], [3, 4], [5, 6]]})
         >>> db = lancedb.connect("./.lancedb")
@@ -1965,7 +1788,7 @@ class Table(ABC):
         0  1  [1.0, 2.0]
         1  2  [3.0, 4.0]
         2  3  [5.0, 6.0]
-        >>> table.update(where=col("x") == 2, values={"vector": [10.0, 10]})
+        >>> table.update(where="x = 2", values={"vector": [10.0, 10]})
         UpdateResult(rows_updated=1, version=2)
         >>> table.to_pandas()
            x        vector
@@ -2165,11 +1988,9 @@ class Table(ABC):
             Function columns are supported only on LanceDB Cloud and
             Enterprise.
         computed: Dict[str, str], optional
-            A mapping from output column names to SQL expressions derives each
-            output field from its expression. A direct projection of a Blob v2
-            field inherits Blob v2 semantics; other expressions derive their
-            ordinary Arrow type. Mapping order is declaration and dependency
-            order.
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression, so no
+            data type is supplied.
 
             Unlike ``transforms``, the expression is stored rather than
             evaluated now: the column is committed with no values, and rows get
@@ -2874,7 +2695,7 @@ class LanceTable(Table):
             arrow_tbl = self.to_arrow()
             if blob_mode == "descriptions":
                 arrow_tbl = strip_auto_row_ids(
-                    arrow_tbl, row_addressable_blob_v2_paths(self.schema)
+                    arrow_tbl, blob_v2_column_paths(self.schema)
                 )
             return arrow_tbl.to_pandas(**kwargs)
 
@@ -4020,7 +3841,7 @@ class LanceTable(Table):
 
     def update(
         self,
-        where: Optional[Union[str, Expr]] = None,
+        where: Optional[str] = None,
         values: Optional[dict] = None,
         *,
         values_sql: Optional[Dict[str, str]] = None,
@@ -4031,11 +3852,9 @@ class LanceTable(Table):
 
         Parameters
         ----------
-        where: str or [Expr][lancedb.expr.Expr], optional
-            The filter condition. Can be a SQL string or a type-safe
-            [Expr][lancedb.expr.Expr] built with [col][lancedb.expr.col] and
-            [lit][lancedb.expr.lit]. The filter must not be empty, or it will
-            error.
+        where: str, optional
+            The SQL where clause to use when updating rows. For example, 'x = 2'
+            or 'x IN (1, 2, 3)'. The filter must not be empty, or it will error.
         values: dict, optional
             The values to update. The keys are the column names and the values
             are the values to set.
@@ -4053,7 +3872,6 @@ class LanceTable(Table):
         Examples
         --------
         >>> import lancedb
-        >>> from lancedb.expr import col
         >>> import pandas as pd
         >>> data = pd.DataFrame({"x": [1, 2, 3], "vector": [[1.0, 2], [3, 4], [5, 6]]})
         >>> db = lancedb.connect("./.lancedb")
@@ -4063,7 +3881,7 @@ class LanceTable(Table):
         0  1  [1.0, 2.0]
         1  2  [3.0, 4.0]
         2  3  [5.0, 6.0]
-        >>> table.update(where=col("x") == 2, values={"vector": [10.0, 10]})
+        >>> table.update(where="x = 2", values={"vector": [10.0, 10]})
         UpdateResult(rows_updated=1, version=2)
         >>> table.to_pandas()
            x        vector
@@ -5278,9 +5096,7 @@ class AsyncTable:
         if blob_mode == "descriptions" or not schema_has_blob_field(schema):
             arrow_tbl = await self.to_arrow()
             if blob_mode == "descriptions":
-                arrow_tbl = strip_auto_row_ids(
-                    arrow_tbl, row_addressable_blob_v2_paths(schema)
-                )
+                arrow_tbl = strip_auto_row_ids(arrow_tbl, blob_v2_column_paths(schema))
             return arrow_tbl.to_pandas(**kwargs)
 
         if blob_mode == "lazy" and get_uri_scheme(await self.uri()) == "memory":
@@ -6179,7 +5995,7 @@ class AsyncTable:
         self,
         updates: Optional[Dict[str, Any]] = None,
         *,
-        where: Optional[Union[str, Expr]] = None,
+        where: Optional[str] = None,
         updates_sql: Optional[Dict[str, str]] = None,
     ) -> UpdateResult:
         """
@@ -6194,11 +6010,9 @@ class AsyncTable:
             The updates to apply.  The keys should be the name of the column to
             update.  The values should be the new values to assign.  This is
             required unless updates_sql is supplied.
-        where: str or [Expr][lancedb.expr.Expr], optional
-            The filter condition. Can be a SQL string or a type-safe
-            [Expr][lancedb.expr.Expr] built with [col][lancedb.expr.col] and
-            [lit][lancedb.expr.lit]. Only rows that satisfy this filter will
-            be updated.
+        where: str, optional
+            An SQL filter that controls which rows are updated. For example, 'x = 2'
+            or 'x IN (1, 2, 3)'.  Only rows that satisfy this filter will be udpated.
         updates_sql: dict, optional
             The updates to apply, expressed as SQL expression strings.  The keys should
             be column names. The values should be SQL expressions.  These can be SQL
@@ -6216,14 +6030,13 @@ class AsyncTable:
         --------
         >>> import asyncio
         >>> import lancedb
-        >>> from lancedb.expr import col
         >>> import pandas as pd
         >>> async def demo_update():
         ...     data = pd.DataFrame({"x": [1, 2], "vector": [[1, 2], [3, 4]]})
         ...     db = await lancedb.connect_async("./.lancedb")
         ...     table = await db.create_table("my_table", data)
         ...     # x is [1, 2], vector is [[1, 2], [3, 4]]
-        ...     await table.update({"vector": [10, 10]}, where=col("x") == 2)
+        ...     await table.update({"vector": [10, 10]}, where="x = 2")
         ...     # x is [1, 2], vector is [[1, 2], [10, 10]]
         ...     await table.update(updates_sql={"x": "x + 1"})
         ...     # x is [2, 3], vector is [[1, 2], [10, 10]]
@@ -6237,8 +6050,7 @@ class AsyncTable:
         if updates is not None:
             updates_sql = {k: value_to_sql(v) for k, v in updates.items()}
 
-        predicate = where.to_sql() if isinstance(where, Expr) else where
-        return await self._inner.update(updates_sql, predicate)
+        return await self._inner.update(updates_sql, where)
 
     async def add_columns(
         self,
@@ -6270,11 +6082,8 @@ class AsyncTable:
             Function columns are supported only on LanceDB Cloud and
             Enterprise.
         computed: Dict[str, str], optional
-            A mapping from output column names to SQL expressions derives each
-            output field from its expression. A direct projection of a Blob v2
-            field inherits Blob v2 semantics; other expressions derive their
-            ordinary Arrow type. Mapping order is declaration and dependency
-            order.
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression.
 
             Unlike ``transforms``, the expression is stored rather than
             evaluated now: the column is committed with no values, and rows get

python/python/tests/test_blob.py

@@ -2,41 +2,17 @@
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 import io
-import subprocess
-import sys
-import textwrap
 
-import lance
 import pyarrow as pa
 import pyarrow.compute as pc
 import pytest
-from lance.blob import BlobType as LanceBlobType
 
 import lancedb
-from lancedb._blob import (
-    blob_v2_projection_sources,
-    read_row_ids_from_hits,
-    stash_auto_row_ids,
-)
-from lancedb.expr import col
+from lancedb._blob import read_row_ids_from_hits, stash_auto_row_ids
 from lancedb.index import FTS
 from lancedb.schema import blob_column_paths, blob_v2_column_paths
 
 
-_HIDE_LANCE_BLOB = """\
-import importlib.abc
-import sys
-
-class _MissingLanceBlob(importlib.abc.MetaPathFinder):
-    def find_spec(self, fullname, path, target=None):
-        if fullname == "lance.blob" or fullname.startswith("lance.blob."):
-            raise ModuleNotFoundError(fullname, name="lance.blob")
-
-sys.modules.pop("lance.blob", None)
-sys.meta_path.insert(0, _MissingLanceBlob())
-"""
-
-
 def _blob_table(name, rows):
     db = lancedb.connect("memory:///")
     schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
@@ -70,181 +46,6 @@ def test_blob_factory_declares_v2_field():
     field = lancedb.blob("image")
     assert isinstance(field.type, pa.ExtensionType)
     assert field.type.extension_name == "lance.blob.v2"
-    assert lancedb.BlobType is LanceBlobType
-    assert type(field.type) is LanceBlobType
-
-
-def test_blob_type_works_without_pylance():
-    script = _HIDE_LANCE_BLOB + textwrap.dedent(
-        """\
-        import lancedb
-        import pyarrow as pa
-
-        field = lancedb.blob("image")
-        if not isinstance(field.type, pa.ExtensionType):
-            raise SystemExit("expected an extension type")
-        if field.type.extension_name != "lance.blob.v2":
-            raise SystemExit(field.type.extension_name)
-        if lancedb.BlobType is not type(field.type):
-            raise SystemExit("BlobType is not the field type class")
-        if lancedb.BlobType.__module__ != "lancedb.schema":
-            raise SystemExit(lancedb.BlobType.__module__)
-
-        db = lancedb.connect("memory:///")
-        table = db.create_table(
-            "images",
-            schema=pa.schema([pa.field("id", pa.int64()), field]),
-        )
-        table.add([{"id": 1, "image": b"hello"}])
-        result = (
-            table.merge_insert("id")
-            .when_matched_update_all()
-            .when_not_matched_insert_all()
-            .execute([{"id": 1, "image": b"updated"}, {"id": 2, "image": b"inserted"}])
-        )
-        if result.num_updated_rows != 1 or result.num_inserted_rows != 1:
-            raise SystemExit(
-                f"merge_insert rows updated={result.num_updated_rows} "
-                f"inserted={result.num_inserted_rows}"
-            )
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
-
-
-def test_blob_resolves_pylance_type_without_eager_import():
-    script = textwrap.dedent(
-        """\
-        import sys
-        import lancedb
-
-        if "lance.blob" in sys.modules:
-            raise SystemExit("import lancedb imported lance.blob")
-        field = lancedb.blob("image")
-        from lance.blob import BlobType
-
-        if type(field.type) is not BlobType:
-            raise SystemExit(f"{type(field.type)} is not {BlobType}")
-        import lance
-
-        image = lance.blob_array([b"x"])
-        if type(image.type) is not BlobType:
-            raise SystemExit("blob_array used a different class")
-        if type(image.type) is not type(field.type):
-            raise SystemExit("field and array classes differ")
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
-
-
-def test_blob_fallback_fails_if_name_already_registered():
-    script = _HIDE_LANCE_BLOB + textwrap.dedent(
-        """\
-        import pyarrow as pa
-
-        class OtherBlobType(pa.ExtensionType):
-            def __init__(self):
-                super().__init__(
-                    pa.struct([pa.field("data", pa.large_binary())]),
-                    "lance.blob.v2",
-                )
-
-            def __arrow_ext_serialize__(self):
-                return b""
-
-            @classmethod
-            def __arrow_ext_deserialize__(cls, storage_type, serialized):
-                return cls()
-
-        pa.register_extension_type(OtherBlobType())
-        import lancedb
-
-        try:
-            lancedb.blob("image")
-        except ValueError as err:
-            if "already registered" not in str(err):
-                raise SystemExit(err)
-        else:
-            raise SystemExit("expected ValueError")
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
-
-
-def test_blob_type_rejects_competing_registration_with_pylance():
-    script = textwrap.dedent(
-        """\
-        import pyarrow as pa
-        import pyarrow.ipc
-
-        class OtherBlobType(pa.ExtensionType):
-            def __init__(self):
-                super().__init__(
-                    pa.struct(
-                        [
-                            pa.field("data", pa.large_binary()),
-                            pa.field("uri", pa.utf8()),
-                            pa.field("position", pa.uint64()),
-                            pa.field("size", pa.uint64()),
-                        ]
-                    ),
-                    "lance.blob.v2",
-                )
-
-            def __arrow_ext_serialize__(self):
-                return b""
-
-            @classmethod
-            def __arrow_ext_deserialize__(cls, storage_type, serialized):
-                return cls()
-
-        pa.register_extension_type(OtherBlobType())
-
-        from lance.blob import BlobType
-
-        if BlobType is OtherBlobType:
-            raise SystemExit("pylance BlobType was replaced")
-        schema = pa.schema([pa.field("value", BlobType())])
-        restored = pa.ipc.read_schema(schema.serialize())
-        if type(restored.field("value").type) is not OtherBlobType:
-            raise SystemExit(type(restored.field("value").type))
-
-        import lancedb
-
-        try:
-            lancedb.blob("image")
-        except ValueError as err:
-            if "__main__.OtherBlobType" not in str(err):
-                raise SystemExit(err)
-        else:
-            raise SystemExit("expected ValueError")
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
 
 
 def test_blob_v2_column_paths_include_list_children():
@@ -269,14 +70,6 @@ def test_blob_v2_column_paths_include_list_children():
     ]
 
 
-def test_blob_v2_projection_sources_use_typed_column_name():
-    schema = pa.schema([lancedb.blob("blob")])
-
-    assert blob_v2_projection_sources(schema, {"blob_alias": col("blob")}) == {
-        "blob_alias": "blob"
-    }
-
-
 def _legacy_v1_table(name):
     db = lancedb.connect("memory:///")
     schema = pa.schema(
@@ -373,20 +166,6 @@ async def test_async_table_to_pandas_descriptions_mode_omits_row_id():
     assert set(descriptor.keys()) == {"kind", "position", "size", "blob_id", "blob_uri"}
 
 
-@pytest.mark.asyncio
-async def test_async_typed_blob_projection_preserves_source_column():
-    db = await lancedb.connect_async("memory:///typed_blob_projection")
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("blob")])
-    table = await db.create_table("typed_blob_projection", schema=schema)
-    await table.add([{"id": 1, "blob": b"alpha"}])
-
-    hits = await table.query().select({"blob_alias": col("blob")}).to_arrow()
-
-    assert "_lance_row_id" in hits.schema.field("blob_alias").type.names
-    blobs = await table.fetch_blobs("blob", hits)
-    assert blobs.to_pylist() == [b"alpha"]
-
-
 def test_fetch_blobs_round_trip():
     table = _blob_table(
         "round_trip",
@@ -397,292 +176,6 @@ def test_fetch_blobs_round_trip():
     assert [blobs[0].as_py(), blobs[1].as_py()] == [b"alpha", b"beta"]
 
 
-def test_merge_insert_writes_python_bytes():
-    table = _blob_table("merge_bytes", [{"id": 1, "image": b"before"}])
-    result = (
-        table.merge_insert("id")
-        .when_matched_update_all()
-        .when_not_matched_insert_all()
-        .execute([{"id": 1, "image": b"updated"}, {"id": 2, "image": b"inserted"}])
-    )
-    assert result.num_updated_rows == 1
-    assert result.num_inserted_rows == 1
-    by_id = _row_ids_by_id(table)
-    blobs = table.fetch_blobs("image", [by_id[1], by_id[2]])
-    assert blobs.to_pylist() == [b"updated", b"inserted"]
-
-
-def test_merge_insert_bytes_after_reopen_without_touching_blob_type(tmp_path):
-    db = lancedb.connect(tmp_path)
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
-    table = db.create_table("images", schema=schema)
-    table.add([{"id": 1, "image": b"hello"}])
-
-    script = textwrap.dedent(
-        f"""\
-        import lancedb
-
-        db = lancedb.connect({str(tmp_path)!r})
-        table = db.open_table("images")
-        image_type = table.schema.field("image").type
-        if type(image_type).__name__ != "StructType":
-            raise SystemExit(f"expected StructType, got {{type(image_type)}}")
-        result = (
-            table.merge_insert("id")
-            .when_matched_update_all()
-            .when_not_matched_insert_all()
-            .execute(
-                [{{"id": 1, "image": b"updated"}}, {{"id": 2, "image": b"inserted"}}]
-            )
-        )
-        if result.num_updated_rows != 1 or result.num_inserted_rows != 1:
-            raise SystemExit(
-                f"rows updated={{result.num_updated_rows}} "
-                f"inserted={{result.num_inserted_rows}}"
-            )
-        hits = table.search().with_row_id(True).limit(10).to_arrow()
-        by_id = dict(zip(hits["id"].to_pylist(), hits["_rowid"].to_pylist()))
-        blobs = table.fetch_blobs("image", [by_id[1], by_id[2]])
-        if blobs.to_pylist() != [b"updated", b"inserted"]:
-            raise SystemExit(blobs.to_pylist())
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
-
-
-def test_merge_insert_bytes_after_reopen_without_pylance(tmp_path):
-    db = lancedb.connect(tmp_path)
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
-    table = db.create_table("images", schema=schema)
-    table.add([{"id": 1, "image": b"hello"}])
-
-    script = _HIDE_LANCE_BLOB + textwrap.dedent(
-        f"""\
-        import lancedb
-
-        db = lancedb.connect({str(tmp_path)!r})
-        table = db.open_table("images")
-        image_type = table.schema.field("image").type
-        if type(image_type).__name__ != "StructType":
-            raise SystemExit(f"expected StructType, got {{type(image_type)}}")
-        result = (
-            table.merge_insert("id")
-            .when_matched_update_all()
-            .when_not_matched_insert_all()
-            .execute(
-                [{{"id": 1, "image": b"updated"}}, {{"id": 2, "image": b"inserted"}}]
-            )
-        )
-        if result.num_updated_rows != 1 or result.num_inserted_rows != 1:
-            raise SystemExit(
-                f"rows updated={{result.num_updated_rows}} "
-                f"inserted={{result.num_inserted_rows}}"
-            )
-        hits = table.search().with_row_id(True).limit(10).to_arrow()
-        by_id = dict(zip(hits["id"].to_pylist(), hits["_rowid"].to_pylist()))
-        blobs = table.fetch_blobs("image", [by_id[1], by_id[2]])
-        if blobs.to_pylist() != [b"updated", b"inserted"]:
-            raise SystemExit(blobs.to_pylist())
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
-
-
-def test_merge_insert_blob_array_into_reopened_unregistered_table(tmp_path):
-    db = lancedb.connect(tmp_path)
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
-    table = db.create_table("images", schema=schema)
-    table.add([{"id": 1, "image": b"before"}])
-
-    script = textwrap.dedent(
-        f"""\
-        import pyarrow as pa
-        import lancedb
-
-        db = lancedb.connect({str(tmp_path)!r})
-        table = db.open_table("images")
-        image_type = table.schema.field("image").type
-        if type(image_type).__name__ != "StructType":
-            raise SystemExit(
-                f"expected StructType before lance import, got {{type(image_type)}}"
-            )
-
-        import lance
-
-        updates = pa.Table.from_arrays(
-            [
-                pa.array([1, 2], type=pa.int64()),
-                lance.blob_array([b"updated", b"inserted"]),
-            ],
-            names=["id", "image"],
-        )
-        result = (
-            table.merge_insert("id")
-            .when_matched_update_all()
-            .when_not_matched_insert_all()
-            .execute(updates)
-        )
-        if result.num_updated_rows != 1 or result.num_inserted_rows != 1:
-            raise SystemExit(
-                f"rows updated={{result.num_updated_rows}} "
-                f"inserted={{result.num_inserted_rows}}"
-            )
-        hits = table.search().with_row_id(True).limit(10).to_arrow()
-        by_id = dict(zip(hits["id"].to_pylist(), hits["_rowid"].to_pylist()))
-        blobs = table.fetch_blobs("image", [by_id[1], by_id[2]])
-        if blobs.to_pylist() != [b"updated", b"inserted"]:
-            raise SystemExit(blobs.to_pylist())
-        """
-    )
-    result = subprocess.run(
-        [sys.executable, "-c", script],
-        capture_output=True,
-        text=True,
-        check=False,
-    )
-    assert result.returncode == 0, result.stderr
-
-
-def test_add_all_null_blob_column():
-    db = lancedb.connect("memory:///")
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
-    table = db.create_table("all_null", schema=schema)
-    table.add([{"id": 1, "image": None}, {"id": 2, "image": None}])
-    by_id = _row_ids_by_id(table)
-    blobs = table.fetch_blobs("image", [by_id[1], by_id[2]])
-    assert blobs.to_pylist() == [None, None]
-
-
-def test_create_table_nested_blob_schema_without_rows():
-    db = lancedb.connect("memory:///")
-    schema = pa.schema(
-        [
-            pa.field("id", pa.int64()),
-            pa.field("info", pa.struct([lancedb.blob("blob")])),
-            pa.field("images", pa.list_(lancedb.blob("image"))),
-        ]
-    )
-    table = db.create_table("nested_empty", schema=schema)
-    assert table.count_rows() == 0
-
-
-def test_merge_insert_nested_blob_dicts():
-    db = lancedb.connect("memory:///")
-    info = pa.StructArray.from_arrays(
-        [
-            pa.array(["first"], type=pa.string()),
-            _blob_array("blob", [b"before"]),
-        ],
-        names=["name", "blob"],
-    )
-    data = pa.Table.from_arrays(
-        [pa.array([1], type=pa.int64()), info],
-        names=["id", "info"],
-    )
-    table = db.create_table("nested_merge", data=data)
-    result = (
-        table.merge_insert("id")
-        .when_matched_update_all()
-        .execute([{"id": 1, "info": {"name": "first", "blob": b"after"}}])
-    )
-    assert result.num_updated_rows == 1
-    by_id = _row_ids_by_id(table)
-    blobs = table.fetch_blobs("info.blob", [by_id[1]])
-    assert blobs.to_pylist() == [b"after"]
-
-
-def _list_blob_table(name):
-    db = lancedb.connect("memory:///")
-    blob_field = lancedb.blob("image")
-    images = pa.ListArray.from_arrays(
-        pa.array([0, 1], type=pa.int32()), _blob_array("image", [b"before"])
-    )
-    data = pa.Table.from_arrays(
-        [pa.array([1], type=pa.int64()), images],
-        schema=pa.schema(
-            [pa.field("id", pa.int64()), pa.field("images", pa.list_(blob_field))]
-        ),
-    )
-    return db.create_table(name, data=data)
-
-
-def test_merge_insert_list_blob_dicts():
-    table = _list_blob_table("list_merge")
-    result = (
-        table.merge_insert("id")
-        .when_matched_update_all()
-        .when_not_matched_insert_all()
-        .execute([{"id": 1, "images": [b"one", b"two"]}, {"id": 2, "images": None}])
-    )
-    assert result.num_updated_rows == 1
-    assert result.num_inserted_rows == 1
-    hits = table.search().limit(10).to_arrow()
-    sizes = {
-        row["id"]: None if row["images"] is None else [d["size"] for d in row["images"]]
-        for row in hits.to_pylist()
-    }
-    assert sizes == {1: [3, 3], 2: None}
-
-
-def test_list_blob_column_queries_as_raw_descriptors():
-    table = _list_blob_table("list_query")
-    hits = table.search().limit(10).to_arrow()
-    element = hits.schema.field("images").type.value_type
-    assert pa.types.is_struct(element)
-    assert "_lance_row_id" not in element.names
-    with pytest.raises(ValueError, match="expected struct before segment"):
-        table.fetch_blobs("images.image", [0])
-
-
-def test_row_addressable_paths_exclude_list_children():
-    from lancedb.schema import row_addressable_blob_v2_paths
-
-    schema = pa.schema(
-        [
-            pa.field("id", pa.int64()),
-            pa.field("info", pa.struct([lancedb.blob("blob")])),
-            pa.field("images", pa.list_(lancedb.blob("image"))),
-        ]
-    )
-    assert blob_v2_column_paths(schema) == ["info.blob", "images.image"]
-    assert row_addressable_blob_v2_paths(schema) == ["info.blob"]
-
-
-def test_merge_insert_writes_pylance_blob_array():
-    table = _blob_table("merge_pylance", [{"id": 1, "image": b"before"}])
-    image = lance.blob_array([b"updated", b"inserted"])
-    assert type(image.type) is LanceBlobType
-    assert type(image.type) is type(lancedb.BlobType())
-    updates = pa.Table.from_arrays(
-        [pa.array([1, 2], type=pa.int64()), image], names=["id", "image"]
-    )
-
-    result = (
-        table.merge_insert("id")
-        .when_matched_update_all()
-        .when_not_matched_insert_all()
-        .execute(updates)
-    )
-
-    assert result.num_updated_rows == 1
-    assert result.num_inserted_rows == 1
-    by_id = _row_ids_by_id(table)
-    blobs = table.fetch_blobs("image", [by_id[1], by_id[2]])
-    assert blobs.to_pylist() == [b"updated", b"inserted"]
-
-
 def test_fetch_blobs_accepts_query_result():
     table = _blob_table("from_result", [{"id": 1, "image": b"gamma"}])
     hits = table.search().limit(10).to_arrow()
@@ -910,50 +403,6 @@ async def test_blob_v2_hybrid_fetch_blobs_async():
     assert {blobs[i].as_py() for i in range(len(blobs))} == {b"alpha", b"beta"}
 
 
-@pytest.mark.asyncio
-async def test_async_hybrid_typed_blob_projection_preserves_source_column():
-    db = await lancedb.connect_async("memory:///hybrid_typed_blob")
-    schema = pa.schema(
-        [
-            pa.field("id", pa.int64()),
-            pa.field("text", pa.utf8()),
-            pa.field("vector", pa.list_(pa.float32(), list_size=2)),
-            lancedb.blob("blob"),
-        ]
-    )
-    table = await db.create_table("hybrid_typed_blob", schema=schema)
-    await table.add(
-        [
-            {
-                "id": 1,
-                "text": "hello alpha",
-                "vector": [1.0, 0.0],
-                "blob": b"alpha",
-            },
-            {
-                "id": 2,
-                "text": "hello beta",
-                "vector": [0.9, 0.1],
-                "blob": b"beta",
-            },
-        ]
-    )
-    await table.create_index("text", config=FTS(with_position=False))
-
-    hits = await (
-        table.query()
-        .nearest_to([1.0, 0.0])
-        .nearest_to_text("hello")
-        .select({"blob_alias": col("blob")})
-        .limit(2)
-        .to_arrow()
-    )
-
-    assert "_lance_row_id" in hits.schema.field("blob_alias").type.names
-    blobs = await table.fetch_blobs("blob", hits)
-    assert {blobs[i].as_py() for i in range(len(blobs))} == {b"alpha", b"beta"}
-
-
 def test_blob_file_seek_read_and_read_range():
     payload = _identifiable_payload(1024)
     table = _blob_table("seek_read", [{"id": 1, "image": payload}])

python/python/tests/test_expr.py

@@ -52,7 +52,7 @@ class TestExprConstruction:
     def test_func(self):
         e = func("lower", col("name"))
         assert isinstance(e, Expr)
-        assert e.to_sql() == "lower(`name`)"
+        assert e.to_sql() == "lower(name)"
 
     def test_func_unknown_raises(self):
         with pytest.raises(Exception):
@@ -115,7 +115,7 @@ class TestExprOperators:
     def test_and_operator(self):
         e = (col("age") > lit(18)) & (col("status") == lit("active"))
         assert isinstance(e, Expr)
-        assert e.to_sql() == "((age > 18) AND (`status` = 'active'))"
+        assert e.to_sql() == "((age > 18) AND (status = 'active'))"
 
     def test_or_operator(self):
         e = (col("a") == lit(1)) | (col("b") == lit(2))
@@ -166,7 +166,7 @@ class TestExprOperators:
     def test_coerce_plain_str(self):
         e = col("name") == "alice"
         assert isinstance(e, Expr)
-        assert e.to_sql() == "(`name` = 'alice')"
+        assert e.to_sql() == "(name = 'alice')"
 
     def test_reflexive_comparisons(self):
         # 10 < col("age") swaps to col("age") > 10
@@ -198,85 +198,85 @@ class TestExprBytesLiteral:
 
     def test_bytes_equality_expr_sql(self):
         e = col("data") == lit(b"\xca\xfe")
-        assert e.to_sql() == "(`data` = X'CAFE')"
+        assert e.to_sql() == "(data = X'CAFE')"
 
     def test_bytes_ne_expr_sql(self):
         e = col("data") != lit(b"\xff")
-        assert e.to_sql() == "(`data` <> X'FF')"
+        assert e.to_sql() == "(data <> X'FF')"
 
     def test_bytes_compound_expr_sql(self):
         e = (col("data") == lit(b"\x01")) & (col("id") > lit(5))
-        assert e.to_sql() == "((`data` = X'01') AND (id > 5))"
+        assert e.to_sql() == "((data = X'01') AND (id > 5))"
 
     def test_bytes_in_function_call(self):
         # Regression test: binary literals inside scalar function calls
         # used to fail because DataFusion's unparser does not support Binary
         # scalars.  Now handled via a placeholder-substitution rewrite.
         e = func("contains", col("data"), lit(b"\xff"))
-        assert e.to_sql() == "contains(`data`, X'FF')"
+        assert e.to_sql() == "contains(data, X'FF')"
 
     def test_bytes_in_not(self):
         e = ~(col("data") == lit(b"\xff"))
-        assert e.to_sql() == "NOT (`data` = X'FF')"
+        assert e.to_sql() == "NOT (data = X'FF')"
 
 
 class TestExprStringMethods:
     def test_lower(self):
         e = col("name").lower()
         assert isinstance(e, Expr)
-        assert e.to_sql() == "lower(`name`)"
+        assert e.to_sql() == "lower(name)"
 
     def test_upper(self):
         e = col("name").upper()
         assert isinstance(e, Expr)
-        assert e.to_sql() == "upper(`name`)"
+        assert e.to_sql() == "upper(name)"
 
     def test_contains(self):
         e = col("text").contains(lit("hello"))
         assert isinstance(e, Expr)
-        assert e.to_sql() == "contains(`text`, 'hello')"
+        assert e.to_sql() == "contains(text, 'hello')"
 
     def test_contains_with_str_coerce(self):
         e = col("text").contains("hello")
         assert isinstance(e, Expr)
-        assert e.to_sql() == "contains(`text`, 'hello')"
+        assert e.to_sql() == "contains(text, 'hello')"
 
     def test_chained_lower_eq(self):
         e = col("name").lower() == lit("alice")
         assert isinstance(e, Expr)
-        assert e.to_sql() == "(lower(`name`) = 'alice')"
+        assert e.to_sql() == "(lower(name) = 'alice')"
 
 
 class TestExprCast:
     def test_cast_string(self):
         e = col("id").cast("string")
         assert isinstance(e, Expr)
-        assert e.to_sql() == "arrow_cast(id, 'Utf8')"
+        assert e.to_sql() == "CAST(id AS VARCHAR)"
 
     def test_cast_int32(self):
         e = col("score").cast("int32")
         assert isinstance(e, Expr)
-        assert e.to_sql() == "arrow_cast(score, 'Int32')"
+        assert e.to_sql() == "CAST(score AS INTEGER)"
 
     def test_cast_float64(self):
         e = col("val").cast("float64")
         assert isinstance(e, Expr)
-        assert e.to_sql() == "arrow_cast(val, 'Float64')"
+        assert e.to_sql() == "CAST(val AS DOUBLE)"
 
     def test_cast_pyarrow_type(self):
         e = col("score").cast(pa.int32())
         assert isinstance(e, Expr)
-        assert e.to_sql() == "arrow_cast(score, 'Int32')"
+        assert e.to_sql() == "CAST(score AS INTEGER)"
 
     def test_cast_pyarrow_float64(self):
         e = col("val").cast(pa.float64())
         assert isinstance(e, Expr)
-        assert e.to_sql() == "arrow_cast(val, 'Float64')"
+        assert e.to_sql() == "CAST(val AS DOUBLE)"
 
     def test_cast_pyarrow_string(self):
         e = col("id").cast(pa.string())
         assert isinstance(e, Expr)
-        assert e.to_sql() == "arrow_cast(id, 'Utf8')"
+        assert e.to_sql() == "CAST(id AS VARCHAR)"
 
     def test_cast_pyarrow_and_string_equivalent(self):
         # pa.int32() and "int32" should produce equivalent SQL
@@ -597,14 +597,14 @@ class TestExprIsin:
     def test_isin_strs(self):
         assert (
             col("status").isin(["active", "pending"]).to_sql()
-            == "`status` IN ('active', 'pending')"
+            == "status IN ('active', 'pending')"
         )
 
     def test_isin_coerces_and_mixes(self):
         assert col("id").isin([lit(1), 2]).to_sql() == "id IN (1, 2)"
 
     def test_isin_empty(self):
-        assert col("id").isin([]).to_sql() == "false"
+        assert col("id").isin([]).to_sql() == "id IN ()"
 
     def test_isin_filter(self, simple_table):
         result = simple_table.search().where(col("id").isin([1, 3, 5])).to_arrow()

python/python/tests/test_first_class_function_slice2.py

@@ -19,13 +19,7 @@ import pyarrow as pa
 import pytest
 
 import lancedb
-from lancedb.functions import (
-    PythonRuntimeSpec,
-    UdfDefinition,
-    _canonical_arrow_type,
-    _GRAMMAR_PRIMITIVES,
-    udf,
-)
+from lancedb.functions import UdfDefinition, udf
 
 THRESHOLD = 20
 _CACHE = None
@@ -95,58 +89,6 @@ def test_udf_conda_environment():
         udf(name="channels", conda_channels=["conda-forge"])(lambda value: value)
 
 
-def test_udf_gpu_marker_uses_gpu_runtime():
-    @udf(pip=["cupy-cuda12x"], gpu=True)
-    def double_on_gpu(value: int) -> int:
-        return value * 2
-
-    request = json.loads(double_on_gpu.registration_request.to_canonical_json())
-    assert request["runtime"]["kind"] == "python_v2"
-    assert request["runtime"]["gpu"] is True
-
-    @udf(pip=["pyarrow"])
-    def cpu_function(value: int) -> int:
-        return value
-
-    cpu_runtime = json.loads(cpu_function.registration_request.to_canonical_json())[
-        "runtime"
-    ]
-    assert cpu_runtime["kind"] == "python"
-    assert "gpu" not in cpu_runtime
-
-    def identity(value: int) -> int:
-        return value
-
-    for invalid in [None, 0, 1, -1, 1.5, "", "true", "1", "H100"]:
-        with pytest.raises(ValueError, match="gpu must be a boolean"):
-            udf(name="invalid_gpu", gpu=invalid)(identity)
-
-    base_runtime = {
-        "kind": "python_v2",
-        "python_version": "3.12",
-        "environment": {"kind": "pip"},
-    }
-    runtime = PythonRuntimeSpec.model_validate({**base_runtime, "gpu": True})
-    assert runtime.gpu is True
-    for invalid in [False, 1, 0, "", "true", "1", "H100"]:
-        with pytest.raises(ValueError, match="runtime.gpu must be true"):
-            PythonRuntimeSpec.model_validate({**base_runtime, "gpu": invalid})
-
-
-def test_unknown_runtime_discards_payload_before_known_field_validation():
-    for payload in [
-        {"kind": "python_v3", "gpu": {"model": "H100"}},
-        {"kind": "python_v3", "resources": []},
-        {
-            "kind": "python_v3",
-            "environment": {"kind": []},
-            "python_version": 3.15,
-        },
-    ]:
-        runtime = PythonRuntimeSpec.model_validate(payload)
-        assert runtime.to_canonical_json() == '{"kind":"python_v3"}'
-
-
 def test_udf_packages_attribute_access_and_body_imports():
     @udf
     def word_norm(body: str) -> float:
@@ -226,7 +168,9 @@ def test_udf_resolves_module_globals_before_builtins(tmp_path):
         udf(module.uses_callable_shadow)
 
 
-def test_canonical_arrow_type_prefers_the_compact_grammar():
+def test_canonical_arrow_type_is_exactly_the_grammar():
+    from lancedb.functions import _GRAMMAR_PRIMITIVES, _canonical_arrow_type
+
     golden = json.loads(
         (
             Path(__file__).parents[3]
@@ -237,19 +181,14 @@ def test_canonical_arrow_type_prefers_the_compact_grammar():
         case["arrow_type"] for case in golden["valid"] if "<" not in case["arrow_type"]
     ]
     assert [name for _, name in _GRAMMAR_PRIMITIVES] == primitives
-    assert _canonical_arrow_type(pa.list_(pa.field("item", pa.float32(), False))) == (
-        "list<float32>"
-    )
-    assert (
-        _canonical_arrow_type(pa.large_list(pa.field("item", pa.float32(), False)))
-        == "large_list<float32>"
-    )
     for outside in [
         pa.timestamp("us"),
         pa.decimal128(10, 2),
+        pa.large_string(),
         pa.large_binary(),
         pa.binary(4),
         pa.duration("s"),
+        pa.struct([pa.field("a", pa.int32())]),
         pa.list_(pa.float32(), 0),
         pa.list_(pa.timestamp("us")),
     ]:
@@ -439,27 +378,14 @@ def test_udf_recursion_versus_a_rebound_module_name(tmp_path):
         udf(raw_fact)
 
 
-def test_canonical_arrow_type_uses_exact_json_for_list_child_properties():
-    nullable = pa.list_(pa.float32())
-    assert json.loads(_canonical_arrow_type(nullable)) == {
-        "type": "list",
-        "fields": [
-            {
-                "name": "item",
-                "nullable": True,
-                "type": {"type": "float32"},
-            }
-        ],
-    }
-    named = pa.list_(pa.field("custom", pa.float32(), nullable=False))
-    assert json.loads(_canonical_arrow_type(named))["fields"][0]["name"] == "custom"
+def test_canonical_arrow_type_rejects_unrepresentable_list_children():
+    from lancedb.functions import _canonical_arrow_type
+
     for outside in [
+        pa.list_(pa.float32()),  # pyarrow default: nullable child
+        pa.list_(pa.field("custom", pa.float32(), nullable=False)),
         pa.list_(pa.field("item", pa.float32(), nullable=False, metadata={"k": "v"})),
         pa.list_(pa.field("item", pa.float32(), nullable=False), 0),
-        pa.list_(
-            pa.field("item", pa.float32(), nullable=False, metadata={"k": "v"}), 3
-        ),
-        pa.list_(pa.field("custom", pa.float32(), nullable=False), 3),
     ]:
         with pytest.raises(TypeError, match="unsupported Arrow type"):
             _canonical_arrow_type(outside)
@@ -469,29 +395,6 @@ def test_canonical_arrow_type_uses_exact_json_for_list_child_properties():
         )
         == "fixed_size_list<float32, 3>"
     )
-    fixed = json.loads(_canonical_arrow_type(pa.list_(pa.float32(), 3)))
-    assert fixed == {
-        "type": "fixed_size_list",
-        "fields": [
-            {
-                "name": "item",
-                "nullable": True,
-                "type": {"type": "float32"},
-            }
-        ],
-        "length": 3,
-    }
-    large = json.loads(_canonical_arrow_type(pa.large_list(pa.float32())))
-    assert large["type"] == "large_list"
-    assert large["fields"][0]["nullable"] is True
-
-    for invalid_struct in [
-        pa.struct([]),
-        pa.struct([pa.field("a", pa.int32()), pa.field("a", pa.int64())]),
-        pa.struct([pa.field("", pa.int32())]),
-    ]:
-        with pytest.raises(TypeError, match="unsupported Arrow type"):
-            _canonical_arrow_type(invalid_struct)
 
 
 def _calls_missing(value: int) -> int:
@@ -529,7 +432,6 @@ def _arrow_type_from_golden(spec: dict) -> pa.DataType:
         "null": pa.null(),
         "bool": pa.bool_(),
         "utf8": pa.string(),
-        "large_utf8": pa.large_string(),
         "binary": pa.binary(),
         "float16": pa.float16(),
         "float32": pa.float32(),
@@ -546,6 +448,8 @@ def test_arrow_type_grammar_matches_the_shared_golden():
             / "rust/lancedb/tests/fixtures/first_class_functions/v1/arrow_types.json"
         ).read_text()
     )
+    from lancedb.functions import _canonical_arrow_type
+
     emitted = {
         case["arrow_type"]: _canonical_arrow_type(_arrow_type_from_golden(case["json"]))
         for case in golden["valid"]
@@ -578,250 +482,6 @@ def test_explicit_arrow_schema_is_deterministic():
     assert signature.output.nullable is False
 
 
-def test_blob_fields_use_the_scalar_function_semantic_type():
-    @udf(
-        input_schema=pa.schema([lancedb.blob("image", nullable=False)]),
-        output_schema=lancedb.blob("result", nullable=False),
-    )
-    def copy_blob(image):
-        return image
-
-    signature = copy_blob.registration_request.signature
-    assert signature.inputs[0].arrow_type == "blob_v2"
-    assert signature.output.kind == "scalar"
-    assert signature.output.arrow_type == "blob_v2"
-
-
-def test_named_struct_function_can_include_a_blob_result_field():
-    @udf(
-        input_schema=pa.schema([lancedb.blob("image", nullable=False)]),
-        output_schema=pa.schema(
-            [
-                lancedb.blob("thumbnail", nullable=False),
-                pa.field("width", pa.int32(), nullable=False),
-            ]
-        ),
-    )
-    def inspect_blob(image):
-        return {"thumbnail": image, "width": 1}
-
-    output = inspect_blob.registration_request.signature.output
-    assert output.kind == "named_struct"
-    assert [(field.name, field.arrow_type) for field in output.fields] == [
-        ("thumbnail", "blob_v2"),
-        ("width", "int32"),
-    ]
-
-
-def test_metadata_marked_blob_field_uses_the_semantic_type():
-    extension = lancedb.blob("image", nullable=False).type
-    storage = (
-        extension.storage_type if isinstance(extension, pa.ExtensionType) else extension
-    )
-    metadata_blob = pa.field(
-        "image",
-        storage,
-        nullable=False,
-        metadata={"ARROW:extension:name": "lance.blob.v2"},
-    )
-
-    @udf(
-        input_schema=pa.schema([metadata_blob]),
-        output_schema=pa.field("size", pa.int64(), nullable=False),
-    )
-    def blob_size(image):
-        return len(image)
-
-    assert blob_size.registration_request.signature.inputs[0].arrow_type == "blob_v2"
-
-
-def test_blob_marker_rejects_invalid_storage_layout():
-    malformed = pa.field(
-        "image",
-        pa.int64(),
-        nullable=False,
-        metadata={"ARROW:extension:name": "lance.blob.v2"},
-    )
-
-    with pytest.raises(TypeError, match="requires a supported Blob storage layout"):
-
-        @udf(
-            input_schema=pa.schema([malformed]),
-            output_schema=pa.field("size", pa.int64(), nullable=False),
-        )
-        def blob_size(image):
-            return len(image)
-
-
-def test_nested_blob_signature_field_has_a_clear_error():
-    nested = pa.field(
-        "value",
-        pa.struct([lancedb.blob("image", nullable=False)]),
-        nullable=False,
-    )
-    with pytest.raises(TypeError, match="nested Blob v2 fields are not supported"):
-
-        @udf(
-            input_schema=pa.schema([nested]),
-            output_schema=pa.field("size", pa.int64(), nullable=False),
-        )
-        def blob_size(value):
-            return len(value["image"])
-
-
-def test_nested_non_blob_extension_is_not_silently_unwrapped():
-    class TestExtension(pa.ExtensionType):
-        def __init__(self):
-            super().__init__(pa.int64(), "test.function.extension")
-
-        def __arrow_ext_serialize__(self):
-            return b""
-
-        @classmethod
-        def __arrow_ext_deserialize__(cls, storage_type, serialized):
-            return cls()
-
-    nested = pa.field(
-        "value",
-        pa.struct([pa.field("extended", TestExtension(), nullable=False)]),
-        nullable=False,
-    )
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema([nested]),
-            output_schema=pa.field("result", pa.int64(), nullable=False),
-        )
-        def extension_value(value):
-            return value["extended"]
-
-
-def test_explicit_large_utf8_schemas_use_the_canonical_function_name():
-    input_schema = pa.schema([pa.field("text", pa.large_string(), nullable=True)])
-    output_schema = pa.field("result", pa.large_string(), nullable=False)
-
-    @udf(input_schema=input_schema, output_schema=output_schema)
-    def preserve(text):
-        return text
-
-    signature = preserve.registration_request.signature
-    assert signature.inputs[0].arrow_type == "large_utf8"
-    assert signature.inputs[0].nullable is True
-    assert signature.output.arrow_type == "large_utf8"
-    assert signature.output.nullable is False
-
-    nested = pa.struct([pa.field("text", pa.large_string(), nullable=True)])
-    assert json.loads(_canonical_arrow_type(nested)) == {
-        "type": "struct",
-        "fields": [
-            {
-                "name": "text",
-                "nullable": True,
-                "type": {"type": "large_utf8"},
-            }
-        ],
-    }
-
-
-def test_nested_struct_output_uses_canonical_exact_json():
-    token = pa.struct(
-        [
-            pa.field("position", pa.int32(), nullable=False),
-            pa.field("value", pa.string(), nullable=False),
-            pa.field("length", pa.int32(), nullable=False),
-        ]
-    )
-    analysis = pa.struct(
-        [
-            pa.field("normalized_text", pa.string(), nullable=False),
-            pa.field("has_content", pa.bool_(), nullable=False),
-            pa.field(
-                "metrics",
-                pa.struct(
-                    [
-                        pa.field("character_count", pa.int64(), nullable=False),
-                        pa.field("word_count", pa.int32(), nullable=False),
-                        pa.field("average_word_length", pa.float64(), nullable=False),
-                    ]
-                ),
-                nullable=False,
-            ),
-            pa.field(
-                "diagnostics",
-                pa.struct(
-                    [
-                        pa.field("status", pa.string(), nullable=False),
-                        pa.field(
-                            "normalization",
-                            pa.struct(
-                                [
-                                    pa.field("changed", pa.bool_(), nullable=False),
-                                    pa.field(
-                                        "original_length", pa.int64(), nullable=False
-                                    ),
-                                ]
-                            ),
-                            nullable=False,
-                        ),
-                    ]
-                ),
-                nullable=False,
-            ),
-            pa.field(
-                "token_preview",
-                pa.list_(pa.field("item", token, nullable=False)),
-                nullable=False,
-            ),
-        ]
-    )
-
-    @udf(
-        input_schema=pa.schema([pa.field("text", pa.string(), nullable=False)]),
-        output_schema=pa.field("analysis", analysis, nullable=False),
-    )
-    def analyze(text):
-        return {"normalized_text": text}
-
-    output = analyze.registration_request.signature.output
-    assert output.kind == "named_struct"
-    assert [field.name for field in output.fields] == [
-        "normalized_text",
-        "has_content",
-        "metrics",
-        "diagnostics",
-        "token_preview",
-    ]
-    metrics = json.loads(output.fields[2].arrow_type)
-    assert metrics == {
-        "type": "struct",
-        "fields": [
-            {
-                "name": "character_count",
-                "nullable": False,
-                "type": {"type": "int64"},
-            },
-            {
-                "name": "word_count",
-                "nullable": False,
-                "type": {"type": "int32"},
-            },
-            {
-                "name": "average_word_length",
-                "nullable": False,
-                "type": {"type": "float64"},
-            },
-        ],
-    }
-    preview = json.loads(output.fields[4].arrow_type)
-    assert preview["type"] == "list"
-    assert preview["fields"][0]["type"]["type"] == "struct"
-    assert [field["name"] for field in preview["fields"][0]["type"]["fields"]] == [
-        "position",
-        "value",
-        "length",
-    ]
-
-
 def test_annotation_and_explicit_schema_validation_fail_closed():
     with pytest.raises(TypeError, match="missing Function annotations"):
 
@@ -865,72 +525,6 @@ def test_annotation_and_explicit_schema_validation_fail_closed():
         def nullable_explicit(value):
             return value
 
-    for invalid_field in [
-        pa.field("", pa.int32(), nullable=False),
-        pa.field("result", pa.int32(), nullable=False, metadata={"k": "v"}),
-    ]:
-        with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-            @udf(
-                input_schema=pa.schema([pa.field("value", pa.int64())]),
-                output_schema=pa.schema([invalid_field]),
-            )
-            def invalid_explicit_field(value):
-                return value
-
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema(
-                [pa.field("value", pa.int64(), metadata={"k": "v"})]
-            ),
-            output_schema=pa.int64(),
-        )
-        def input_field_metadata(value):
-            return value
-
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema([pa.field("value", pa.int64())]),
-            output_schema=pa.field(
-                "result", pa.int64(), nullable=False, metadata={"k": "v"}
-            ),
-        )
-        def scalar_output_field_metadata(value):
-            return value
-
-    struct_type = pa.struct([pa.field("value", pa.int64(), nullable=False)])
-    with pytest.raises(TypeError, match="unsupported Arrow type"):
-
-        @udf(
-            input_schema=pa.schema([pa.field("value", pa.int64())]),
-            output_schema=pa.field(
-                "result", struct_type, nullable=False, metadata={"k": "v"}
-            ),
-        )
-        def struct_output_field_metadata(value):
-            return {"value": value}
-
-    for input_schema, output_schema in [
-        (
-            pa.schema([pa.field("value", pa.int64())], metadata={"k": "v"}),
-            pa.int64(),
-        ),
-        (
-            pa.schema([pa.field("value", pa.int64())]),
-            pa.schema(
-                [pa.field("result", pa.int64(), nullable=False)],
-                metadata={"k": "v"},
-            ),
-        ),
-    ]:
-        with pytest.raises(TypeError, match="schema metadata"):
-
-            @udf(input_schema=input_schema, output_schema=output_schema)
-            def schema_metadata(value):
-                return value
-
 
 def test_local_function_catalog_operations_are_not_supported(tmp_path):
     db = lancedb.connect(tmp_path)

python/python/tests/test_query.py

@@ -675,21 +675,6 @@ def test_distance_range(table: lancedb.table.Table):
     assert res["_distance"].to_pylist() == [min_dist, max_dist]
 
 
-@pytest.mark.parametrize("expression", ["1 - _distance", "1.0 - _distance"])
-def test_select_arithmetic_with_distance(table, expression):
-    result = (
-        table.search([10, 10])
-        .select({"similarity": expression, "_distance": "_distance"})
-        .distance_type("cosine")
-        .to_arrow()
-    )
-
-    assert result.schema.field("similarity").type == pa.float32()
-    assert result["similarity"].to_pylist() == pytest.approx(
-        [1 - distance for distance in result["_distance"].to_pylist()]
-    )
-
-
 @pytest.mark.asyncio
 async def test_distance_range_async(table_async: AsyncTable):
     q = [0, 0]

python/python/tests/test_table.py

@@ -11,7 +11,6 @@ import warnings
 import weakref
 from concurrent.futures import ThreadPoolExecutor
 from datetime import date, datetime, timedelta
-from decimal import Decimal
 from time import sleep
 from typing import List
 from unittest.mock import patch
@@ -337,21 +336,6 @@ async def test_update_async(mem_db_async: AsyncConnection):
     assert await table.count_rows("id == 10") == 1
 
 
-@pytest.mark.asyncio
-async def test_update_expr_filter_literals_async(mem_db_async: AsyncConnection):
-    values = ["5", "4.66e-84", "it's"]
-    table = await mem_db_async.create_table(
-        "update_expr_literals",
-        data=[{"field": value, "result": "original"} for value in values],
-    )
-
-    for value in values:
-        update_res = await table.update({"result": value}, where=col("field") == value)
-        assert update_res.rows_updated == 1
-
-    assert (await table.to_arrow())["result"].to_pylist() == values
-
-
 def test_create_table(mem_db: DBConnection):
     schema = pa.schema(
         {
@@ -2359,148 +2343,6 @@ def test_update(mem_db: DBConnection):
     assert np.allclose(v, np.array([[1.2, 1.9], [1.1, 1.1]]))
 
 
-def test_update_expr_filter_literals(mem_db: DBConnection):
-    values = ["5", "4.66e-84", "it's"]
-    table = mem_db.create_table(
-        "update_expr_literals",
-        data=[{"field": value, "result": "original"} for value in values],
-    )
-
-    for value in values:
-        update_res = table.update(where=col("field") == value, values={"result": value})
-        assert update_res.rows_updated == 1
-
-    assert table.to_arrow()["result"].to_pylist() == values
-
-
-def test_update_expr_filter_preserves_typed_semantics(mem_db: DBConnection):
-    low = Decimal("1.234567890123456789")
-    high = Decimal("1.234567890123456790")
-    decimal_schema = pa.schema(
-        [("val", pa.decimal128(19, 18)), ("result", pa.string())]
-    )
-    decimal_table = mem_db.create_table(
-        "update_expr_decimal",
-        pa.table(
-            {"val": [low, high], "result": ["old", "old"]},
-            schema=decimal_schema,
-        ),
-    )
-    predicate = col("val") < lit(high)
-    assert decimal_table.search().where(predicate).to_arrow().num_rows == 1
-    result = decimal_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 1
-
-    keyword_table = mem_db.create_table(
-        "update_expr_keyword", [{"null": 1, "result": "old"}]
-    )
-    predicate = col("null") == 1
-    assert keyword_table.search().where(predicate).to_arrow().num_rows == 1
-    result = keyword_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 1
-
-    empty_in_table = mem_db.create_table(
-        "update_expr_empty_in", [{"id": 1, "result": "old"}]
-    )
-    predicate = col("id").isin([])
-    assert empty_in_table.search().where(predicate).to_arrow().num_rows == 0
-    result = empty_in_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 0
-
-    marker = "__lancedb_binary_placeholder_0__"
-    binary_schema = pa.schema(
-        [("payload", pa.binary()), ("text", pa.string()), ("result", pa.string())]
-    )
-    binary_table = mem_db.create_table(
-        "update_expr_binary",
-        pa.table(
-            {
-                "payload": [b"\x01", b"\x02"],
-                "text": ["other", marker],
-                "result": ["old", "old"],
-            },
-            schema=binary_schema,
-        ),
-    )
-    predicate = (col("payload") == lit(b"\x01")) | (col("text") == marker)
-    assert binary_table.search().where(predicate).to_arrow().num_rows == 2
-    result = binary_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 2
-
-    nonfinite_table = mem_db.create_table(
-        "update_expr_nonfinite",
-        [{"x": 1.0, "result": "old"}, {"x": 2.0, "result": "old"}],
-    )
-    predicate = col("x") < float("inf")
-    assert nonfinite_table.search().where(predicate).to_arrow().num_rows == 2
-    result = nonfinite_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 2
-
-    float16_table = mem_db.create_table(
-        "update_expr_float16",
-        [{"x": 1.0, "result": "old"}, {"x": 3.0, "result": "old"}],
-    )
-    predicate = col("x").cast(pa.float16()) < 2.0
-    assert float16_table.search().where(predicate).to_arrow().num_rows == 1
-    result = float16_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 1
-
-    string_cast_table = mem_db.create_table(
-        "update_expr_string_cast",
-        [{"x": 1, "result": "old"}, {"x": 2, "result": "old"}],
-    )
-    predicate = col("x").cast("string") == "1"
-    assert string_cast_table.search().where(predicate).to_arrow().num_rows == 1
-    result = string_cast_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 1
-
-    quoted_identifier_schema = pa.schema(
-        [("payload", pa.binary()), ("odd'name", pa.int64()), ("result", pa.string())]
-    )
-    quoted_identifier_table = mem_db.create_table(
-        "update_expr_quoted_identifier",
-        pa.table(
-            {"payload": [b"\x01"], "odd'name": [1], "result": ["old"]},
-            schema=quoted_identifier_schema,
-        ),
-    )
-    predicate = (col("payload") == lit(b"\x01")) & (col("odd'name") == 1)
-    assert quoted_identifier_table.search().where(predicate).to_arrow().num_rows == 1
-    result = quoted_identifier_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 1
-
-    decimal256_schema = pa.schema(
-        [("val", pa.decimal256(40, 2)), ("result", pa.string())]
-    )
-    decimal256_table = mem_db.create_table(
-        "update_expr_decimal256",
-        pa.table(
-            {
-                "val": [Decimal("1.00"), Decimal("3.00")],
-                "result": ["old", "old"],
-            },
-            schema=decimal256_schema,
-        ),
-    )
-    predicate = col("val") < lit(Decimal("2.00")).cast(pa.decimal256(40, 2))
-    assert decimal256_table.search().where(predicate).to_arrow().num_rows == 1
-    result = decimal256_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 1
-
-    binary_empty_table = mem_db.create_table(
-        "update_expr_binary_empty",
-        pa.table(
-            {"payload": [b"\x01", b"\x02"], "result": ["old", "old"]},
-            schema=pa.schema([("payload", pa.binary()), ("result", pa.string())]),
-        ),
-    )
-    predicate = (col("payload") == lit(b"\x01")).isin([])
-    assert binary_empty_table.search().where(predicate).to_arrow().num_rows == 0
-    assert predicate.to_sql() == "false"
-    result = binary_empty_table.update(where=predicate, values={"result": "new"})
-    assert result.rows_updated == 0
-
-
 def test_update_with_arrow_scalar(mem_db: DBConnection):
     schema = pa.schema({"id": pa.int64(), "vector": pa.list_(pa.float32(), 4)})
     table = mem_db.create_table("my_table", schema=schema)
@@ -4087,29 +3929,6 @@ def test_computed_column_rejects_transforms_and_computed_together(tmp_path):
         table.add_columns({"a": "x + 1"}, computed={"b": "x * 2"})
 
 
-def test_computed_column_blob_projection_inherits_semantics(tmp_path):
-    schema = pa.schema([pa.field("id", pa.int64()), lancedb.blob("image")])
-    db = lancedb.connect(tmp_path)
-    table = db.create_table("computed_column_blob", schema=schema)
-    table.add(
-        [
-            {"id": 1, "image": b"hello"},
-            {"id": 2, "image": b""},
-            {"id": 3, "image": None},
-        ]
-    )
-
-    table.add_columns(computed={"image_copy": "image", "second_copy": "image_copy"})
-    assert table.refresh_column("image_copy").rows_filled == 2
-    assert table.refresh_column("second_copy").rows_filled == 2
-    assert table.blob_columns() == ["image", "image_copy", "second_copy"]
-
-    hits = table.search().with_row_id(True).limit(10).to_arrow()
-    rows = sorted(zip(hits["id"].to_pylist(), hits["_rowid"].to_pylist()))
-    copied = table.fetch_blobs("second_copy", [row_id for _, row_id in rows])
-    assert copied.to_pylist() == [b"hello", b"", None]
-
-
 @pytest.mark.asyncio
 async def test_computed_column_async(tmp_path):
     db = await lancedb.connect_async(tmp_path)

python/python/tests/test_util.py

@@ -7,7 +7,6 @@ import pathlib
 from typing import Optional
 
 import lance
-from lance.blob import BlobType as LanceBlobType
 from lancedb.conftest import MockTextEmbeddingFunction
 from lancedb.embeddings.base import EmbeddingFunctionConfig
 from lancedb.embeddings.registry import EmbeddingFunctionRegistry
@@ -908,165 +907,6 @@ def test_cast_to_target_schema():
     assert output == expected
 
 
-def test_cast_to_target_schema_coerces_binary_to_blob_v2():
-    data = pa.table({"image": pa.array([b"hello", None], type=pa.binary())})
-    target = pa.schema([lancedb.blob("image")])
-
-    output = _cast_to_target_schema(data.to_reader(), target).read_all()
-
-    image = output["image"].chunk(0)
-    assert type(image.type) is lancedb.BlobType
-    assert image.storage.to_pylist() == [
-        {"data": b"hello", "uri": None, "position": None, "size": None},
-        None,
-    ]
-
-
-def test_cast_to_target_schema_coerces_binary_to_metadata_blob_struct():
-    storage = lancedb.blob("image").type.storage_type
-    target = pa.schema(
-        [
-            pa.field(
-                "image",
-                storage,
-                metadata={
-                    b"ARROW:extension:name": b"lance.blob.v2",
-                    b"ARROW:extension:metadata": b"",
-                },
-            )
-        ]
-    )
-    data = pa.table({"image": pa.array([b"hello", None], type=pa.binary())})
-
-    output = _cast_to_target_schema(data.to_reader(), target).read_all()
-
-    image = output["image"].chunk(0)
-    assert not isinstance(image.type, pa.ExtensionType)
-    assert image.to_pylist() == [
-        {"data": b"hello", "uri": None, "position": None, "size": None},
-        None,
-    ]
-
-
-def test_cast_to_target_schema_coerces_nested_binary_blob():
-    data = pa.table(
-        {
-            "info": pa.array(
-                [{"blob": b"hello"}, {"blob": None}],
-                type=pa.struct([pa.field("blob", pa.binary())]),
-            )
-        }
-    )
-    target = pa.schema([pa.field("info", pa.struct([lancedb.blob("blob")]))])
-
-    output = _cast_to_target_schema(data.to_reader(), target).read_all()
-
-    blob = output["info"].chunk(0).field("blob")
-    assert type(blob.type) is lancedb.BlobType
-    assert blob.storage.to_pylist() == [
-        {"data": b"hello", "uri": None, "position": None, "size": None},
-        None,
-    ]
-
-
-def test_cast_to_target_schema_coerces_list_binary_blob_with_inferred_child_name():
-    data = pa.table(
-        {"images": pa.array([[b"a", b"b"], None], type=pa.list_(pa.binary()))}
-    )
-    target = pa.schema([pa.field("images", pa.list_(lancedb.blob("image")))])
-
-    output = _cast_to_target_schema(data.to_reader(), target).read_all()
-
-    images = output["images"].chunk(0)
-    assert images.type.value_field.name == "image"
-    assert type(images.type.value_type) is lancedb.BlobType
-    assert images.to_pylist()[1] is None
-    assert images.values.storage.to_pylist() == [
-        {"data": b"a", "uri": None, "position": None, "size": None},
-        {"data": b"b", "uri": None, "position": None, "size": None},
-    ]
-
-
-def test_list_blob_coercion_preserves_null_slots_with_nonzero_extent():
-    child = pa.field("image", pa.binary())
-    source = pa.ListArray.from_arrays(
-        pa.array([0, 2, 4], type=pa.int32()),
-        pa.array([b"a", b"b", b"dead", b"beef"], type=pa.binary()),
-        mask=pa.array([False, True]),
-    ).cast(pa.list_(child))
-    target = pa.schema([pa.field("images", pa.list_(lancedb.blob("image")))])
-
-    output = _cast_to_target_schema(
-        pa.table({"images": source}).to_reader(), target
-    ).read_all()
-
-    images = output["images"].chunk(0)
-    assert images.to_pylist()[1] is None
-    assert [b["data"] for b in images.to_pylist()[0]] == [b"a", b"b"]
-
-
-def test_fixed_size_list_blob_coercion_keeps_null_rows():
-    child = pa.field("frame", pa.binary())
-    source = (
-        pa.FixedSizeListArray.from_arrays(
-            pa.array([b"a", b"b", b"c", b"d"], type=pa.binary()), 2
-        )
-        .take(pa.array([0, None], type=pa.int32()))
-        .cast(pa.list_(child, 2))
-    )
-    target = pa.schema([pa.field("frames", pa.list_(lancedb.blob("frame"), 2))])
-
-    output = _cast_to_target_schema(
-        pa.table({"frames": source}).to_reader(), target
-    ).read_all()
-
-    frames = output["frames"].chunk(0)
-    assert frames.to_pylist()[1] is None
-    assert [b["data"] for b in frames.to_pylist()[0]] == [b"a", b"b"]
-
-
-def test_cast_to_target_schema_accepts_pylance_blob_v2():
-    target_type = lancedb.BlobType()
-    source = lance.blob_array([b"hello", None])
-    assert type(source.type) is LanceBlobType
-    assert type(source.type) is type(target_type)
-    data = pa.table({"image": source})
-    target = pa.schema([pa.field("image", target_type)])
-
-    output = _cast_to_target_schema(data.to_reader(), target).read_all()
-
-    image = output["image"].chunk(0)
-    assert type(image.type) is LanceBlobType
-    assert image.type == target_type
-    assert image.storage.to_pylist() == [
-        {"data": b"hello", "uri": None, "position": None, "size": None},
-        None,
-    ]
-
-
-def test_cast_to_target_schema_rejects_different_blob_v2_class():
-    class OtherBlobType(pa.ExtensionType):
-        def __init__(self):
-            super().__init__(lancedb.BlobType().storage_type, "lance.blob.v2")
-
-        def __arrow_ext_serialize__(self) -> bytes:
-            return b""
-
-        @classmethod
-        def __arrow_ext_deserialize__(
-            cls, storage_type: pa.DataType, serialized: bytes
-        ) -> "OtherBlobType":
-            return cls()
-
-    storage = lance.blob_array([b"hello"]).storage
-    source = pa.ExtensionArray.from_storage(OtherBlobType(), storage)
-    data = pa.table({"image": source})
-    target = pa.schema([lancedb.blob("image")])
-
-    with pytest.raises(pa.ArrowTypeError, match="different extension type"):
-        _cast_to_target_schema(data.to_reader(), target).read_all()
-
-
 def test_sanitize_data_stream():
     # Make sure we don't collect the whole stream when running sanitize_data
     schema = pa.schema({"a": pa.int32()})

python/src/expr.rs

@@ -130,14 +130,6 @@ impl PyExpr {
 
     // ── utilities ────────────────────────────────────────────────────────────
 
-    /// Return the referenced column name for a bare column expression.
-    fn column_name(&self) -> Option<String> {
-        match &self.0 {
-            DfExpr::Column(column) if column.relation.is_none() => Some(column.name.clone()),
-            _ => None,
-        }
-    }
-
     /// Render the expression as a SQL string (useful for debugging).
     fn to_sql(&self) -> PyResult<String> {
         lancedb::expr::expr_to_sql_string(&self.0).map_err(|e| PyValueError::new_err(e.to_string()))

python/src/query.rs

@@ -1,7 +1,6 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use std::collections::HashMap;
 use std::sync::Arc;
 use std::time::Duration;
 
@@ -326,7 +325,6 @@ pub struct PyQueryRequest {
     pub filter: Option<PyQueryFilter>,
     pub full_text_search: Option<PyLanceDB<FtsQuery>>,
     pub select: PySelect,
-    pub select_source_columns: Option<HashMap<String, String>>,
     pub fast_search: Option<bool>,
     pub with_row_id: Option<bool>,
     pub use_lsm: Option<bool>,
@@ -357,7 +355,6 @@ impl From<AnyQuery> for PyQueryRequest {
                 full_text_search: query_request
                     .full_text_search
                     .map(|fts| PyLanceDB(fts.query)),
-                select_source_columns: PySelect::source_columns(&query_request.select),
                 select: PySelect(query_request.select),
                 fast_search: Some(query_request.fast_search),
                 with_row_id: Some(query_request.with_row_id),
@@ -383,7 +380,6 @@ impl From<AnyQuery> for PyQueryRequest {
                 offset: vector_query.base.offset,
                 filter: vector_query.base.filter.map(PyQueryFilter),
                 full_text_search: None,
-                select_source_columns: PySelect::source_columns(&vector_query.base.select),
                 select: PySelect(vector_query.base.select),
                 fast_search: Some(vector_query.base.fast_search),
                 with_row_id: Some(vector_query.base.with_row_id),
@@ -416,25 +412,6 @@ impl From<AnyQuery> for PyQueryRequest {
 #[derive(Clone)]
 pub struct PySelect(Select);
 
-impl PySelect {
-    fn source_columns(select: &Select) -> Option<HashMap<String, String>> {
-        match select {
-            Select::Expr(pairs) => Some(
-                pairs
-                    .iter()
-                    .filter_map(|(output, expr)| match expr {
-                        lancedb::expr::DfExpr::Column(column) if column.relation.is_none() => {
-                            Some((output.clone(), column.name.clone()))
-                        }
-                        _ => None,
-                    })
-                    .collect(),
-            ),
-            _ => None,
-        }
-    }
-}
-
 impl<'py> IntoPyObject<'py> for PySelect {
     type Target = PyAny;
     type Output = Bound<'py, Self::Target>;

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.38.0"
+version = "0.38.0-beta.11"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true

rust/lancedb/src/database/listing.rs

@@ -281,88 +281,6 @@ impl std::fmt::Display for ListingDatabase {
 }
 
 const LANCE_EXTENSION: &str = "lance";
-
-/// The table a listed child directory holds, or `None` if it is not a table at all.
-///
-/// A table is the directory `<name>.lance`; a loose file or any other directory under the
-/// database prefix belongs to something else. `dir_suffix` is `.lance`, built once by the
-/// caller rather than per child.
-fn table_name(location: &object_store::path::Path, dir_suffix: &str) -> Option<String> {
-    location
-        .filename()?
-        .strip_suffix(dir_suffix)
-        .map(String::from)
-        .filter(|name| !name.is_empty())
-}
-
-/// One page of the table directories under the database directory, in key order.
-struct DirPage {
-    /// The table directories the page holds, as the store lists them.
-    common_prefixes: Vec<object_store::path::Path>,
-    /// Resumes after this page, or `None` when the page reached the end of the level.
-    page_token: Option<String>,
-}
-
-/// Where a listed location sits inside the database directory — the space page tokens live
-/// in — or `None` if it is not a child of that directory at all. Matching both halves of the
-/// prefix drops a location that merely starts with the directory's name (`dbx/y` against
-/// `db/`) as well as the marker object some stores keep for the directory itself.
-fn relative_key<'a>(prefix: Option<&str>, location: &'a str) -> Option<&'a str> {
-    let relative = match prefix {
-        Some(prefix) => location.strip_prefix(prefix)?,
-        None => location,
-    };
-    (!relative.is_empty()).then_some(relative)
-}
-
-/// One page of the table directories under `base_path`, one directory level deep.
-///
-/// Lance 11 exposes no paginated directory listing, so the level is listed in full and paged
-/// locally: table directories go into key order (a directory's key keeps its trailing `/`,
-/// so a token is never a table name), the page is the smallest `limit` of them past
-/// `page_token`, and the token handed back is the key of the last directory the page took —
-/// so a page that took nothing ends the listing rather than resuming from a position no page
-/// ever reached. Only `<name>.lance/` directories enter the page: loose objects, other
-/// directories, and a bare `.lance/` never take a page slot or name a token, which keeps a
-/// page to exactly one listing of the level. Correct on every store, at the cost of that one
-/// full-level listing per page.
-async fn read_dir_page(
-    object_store: &ObjectStore,
-    base_path: &object_store::path::Path,
-    page_token: Option<String>,
-    limit: Option<usize>,
-) -> Result<DirPage> {
-    let listed = object_store.list_with_delimiter(Some(base_path)).await?;
-    let prefix = {
-        let base = base_path.as_ref();
-        (!base.is_empty()).then(|| format!("{base}/"))
-    };
-    let table_dir_suffix = format!(".{LANCE_EXTENSION}/");
-    let mut children: Vec<(String, object_store::path::Path)> = listed
-        .common_prefixes
-        .into_iter()
-        .filter_map(|location| {
-            let key = format!("{}/", relative_key(prefix.as_deref(), location.as_ref())?);
-            (key.len() > table_dir_suffix.len() && key.ends_with(&table_dir_suffix))
-                .then_some((key, location))
-        })
-        .collect();
-    children.sort_unstable_by(|(left, _), (right, _)| left.cmp(right));
-    if let Some(resume) = &page_token {
-        children.retain(|(key, _)| key > resume);
-    }
-    let total = children.len();
-    children.truncate(limit.unwrap_or(total).min(total));
-    let page_token = match children.last() {
-        Some((last, _)) if children.len() < total => Some(last.clone()),
-        _ => None,
-    };
-    Ok(DirPage {
-        common_prefixes: children.into_iter().map(|(_, location)| location).collect(),
-        page_token,
-    })
-}
-
 const ENGINE: &str = "engine";
 const MIRRORED_STORE: &str = "mirroredStore";
 
@@ -1026,57 +944,51 @@ impl Database for ListingDatabase {
         Ok(f)
     }
 
-    /// List the tables in the database, a page at a time.
-    ///
-    /// The page_token is opaque, unlike the `start_after` parameter of [`Self::table_names()`].
-    ///
-    /// When there are no more results, the returned page_token will be None.
-    ///
-    /// `limit` is the maximum number of tables to return in the response. But it is possible
-    /// for the response to contain fewer than `limit` tables, even when there are more tables
-    /// to return. Clients should check the returned page_token to determine if there are
-    /// more results, rather than relying on the number of tables returned.
-    ///
-    /// The order that results are returned in not guaranteed to be stable across calls,
-    /// so clients should not rely on it.
     async fn list_tables(&self, request: ListTablesRequest) -> Result<ListTablesResponse> {
         if request.id.as_ref().map(|v| !v.is_empty()).unwrap_or(false) {
             return self.namespace_database().list_tables(request).await;
         }
-        let limit = request.limit.map(|limit| limit.max(0) as usize);
-        let dir_suffix = format!(".{LANCE_EXTENSION}");
-        let mut tables = Vec::new();
-        let mut page_token = request.page_token.filter(|token| !token.is_empty());
+        let mut f = self
+            .object_store
+            .read_dir(self.base_path.clone())
+            .await?
+            .iter()
+            .map(Path::new)
+            .filter(|path| {
+                let is_lance = path
+                    .extension()
+                    .and_then(|e| e.to_str())
+                    .map(|e| e == LANCE_EXTENSION);
+                is_lance.unwrap_or(false)
+            })
+            .filter_map(|p| p.file_stem().and_then(|s| s.to_str().map(String::from)))
+            .collect::<Vec<String>>();
+        f.sort();
 
-        // A page of nothing: no table was handed over for a token to resume after.
-        if limit == Some(0) {
-            return Ok(ListTablesResponse {
-                context: None,
-                tables,
-                page_token: None,
-            });
+        // Handle pagination with page_token
+        if let Some(ref page_token) = request.page_token {
+            let index = f
+                .iter()
+                .position(|name| name.as_str() > page_token.as_str())
+                .unwrap_or(f.len());
+            f.drain(0..index);
         }
 
-        // The page holds only table directories, so one call — and the one full-level
-        // listing behind it — fills it.
-        let page = read_dir_page(
-            &self.object_store,
-            &self.base_path,
-            page_token.take(),
-            limit,
-        )
-        .await?;
-        page_token = page.page_token;
-        tables.extend(
-            page.common_prefixes
-                .iter()
-                .filter_map(|location| table_name(location, &dir_suffix)),
-        );
+        // Determine if there's a next page. The token is the last name of this page,
+        // not the first of the next one: the next page resumes strictly after the
+        // token, so naming the next page's first entry would skip it.
+        let next_page_token = match request.limit {
+            Some(limit) if f.len() > limit as usize => {
+                f.truncate(limit as usize);
+                f.last().cloned()
+            }
+            _ => None,
+        };
 
         Ok(ListTablesResponse {
             context: None,
-            tables,
-            page_token,
+            tables: f,
+            page_token: next_page_token,
         })
     }
 
@@ -1572,213 +1484,6 @@ mod tests {
     use tokio::sync::Barrier;
     use tokio::time::timeout;
 
-    async fn create_tables(db: &ListingDatabase, names: &[&str]) {
-        let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int32, false)]));
-        for name in names {
-            db.create_table(CreateTableRequest {
-                name: name.to_string(),
-                namespace_path: vec![],
-                data: Box::new(RecordBatch::new_empty(schema.clone())) as Box<dyn Scannable>,
-                mode: CreateTableMode::Create,
-                write_options: Default::default(),
-                location: None,
-                namespace_client: None,
-            })
-            .await
-            .unwrap();
-        }
-    }
-
-    /// Every table in the database, taken `limit` at a time, which is how a caller walks a
-    /// listing: the token ends the walk, never a short page.
-    async fn walk(db: &ListingDatabase, limit: Option<i32>) -> Vec<String> {
-        let mut seen = Vec::new();
-        let mut page_token = None;
-        loop {
-            let page = db
-                .list_tables(ListTablesRequest {
-                    limit,
-                    page_token,
-                    ..Default::default()
-                })
-                .await
-                .unwrap();
-            seen.extend(page.tables);
-            page_token = page.page_token;
-            if page_token.is_none() {
-                return seen;
-            }
-            assert!(
-                seen.len() < 100,
-                "the walk is serving tables more than once"
-            );
-        }
-    }
-
-    /// Paging with the returned token has to visit every table exactly once, whatever the
-    /// page size, with nothing lost or repeated at a boundary.
-    #[rstest::rstest]
-    #[tokio::test]
-    async fn test_list_tables_pages_over_every_table_once(#[values(1, 2, 3, 5, 10)] limit: i32) {
-        let (_tempdir, db) = setup_database().await;
-        create_tables(&db, &["a", "b", "c", "d", "e"]).await;
-
-        assert_eq!(walk(&db, Some(limit)).await, vec!["a", "b", "c", "d", "e"]);
-    }
-
-    /// The token is opaque: it is whatever resumes the store the database sits on, not a
-    /// table name. Callers hand it back and nothing else.
-    ///
-    /// Nothing validates a token, so one invented by a caller is read as a position rather
-    /// than refused — which is why the token has to come back from a previous page.
-    #[tokio::test]
-    async fn test_the_page_token_is_not_a_table_name() {
-        let (_tempdir, db) = setup_database().await;
-        create_tables(&db, &["a", "b", "c"]).await;
-
-        let page = db
-            .list_tables(ListTablesRequest {
-                limit: Some(1),
-                ..Default::default()
-            })
-            .await
-            .unwrap();
-
-        assert_eq!(page.tables, vec!["a"]);
-        let token = page.page_token.expect("two tables are still to come");
-        assert_ne!(token, "a");
-
-        // Handing it back is the only thing a caller does with it, and it resumes.
-        let rest = db
-            .list_tables(ListTablesRequest {
-                page_token: Some(token),
-                ..Default::default()
-            })
-            .await
-            .unwrap();
-        assert_eq!(rest.tables, vec!["b", "c"]);
-    }
-
-    /// A limit the listing does not fill leaves no token behind, so a caller paging by token
-    /// stops without asking for an empty page.
-    #[tokio::test]
-    async fn test_a_listing_that_runs_out_has_no_token() {
-        let (_tempdir, db) = setup_database().await;
-        create_tables(&db, &["a", "b"]).await;
-
-        let page = db
-            .list_tables(ListTablesRequest {
-                limit: Some(10),
-                ..Default::default()
-            })
-            .await
-            .unwrap();
-
-        assert_eq!(page.tables, vec!["a", "b"]);
-        assert_eq!(page.page_token, None);
-    }
-
-    /// An empty page token means "from the start", which is how a client looping on a token
-    /// spells its first request.
-    #[tokio::test]
-    async fn test_an_empty_page_token_lists_from_the_start() {
-        let (_tempdir, db) = setup_database().await;
-        create_tables(&db, &["a", "b"]).await;
-
-        let page = db
-            .list_tables(ListTablesRequest {
-                page_token: Some(String::new()),
-                ..Default::default()
-            })
-            .await
-            .unwrap();
-
-        assert_eq!(page.tables, vec!["a", "b"]);
-    }
-
-    /// Listing follows the order the object store lists directories in, so a name that
-    /// extends another comes first: the `-` of `users-archive.lance` sorts below the `.` of
-    /// `users.lance`. Pagination pushes its cursor into the list request, so it cannot report
-    /// an order other than the one it resumes in.
-    #[tokio::test]
-    async fn test_listing_order_follows_the_store_not_the_table_name() {
-        let (_tempdir, db) = setup_database().await;
-        create_tables(&db, &["users", "users-archive", "users.old"]).await;
-
-        assert_eq!(
-            walk(&db, None).await,
-            vec!["users-archive", "users", "users.old"]
-        );
-        // And paging reports the same order, so a walk sees each table once.
-        assert_eq!(
-            walk(&db, Some(1)).await,
-            vec!["users-archive", "users", "users.old"]
-        );
-    }
-
-    /// Only directories named `<name>.lance` are tables; loose files and other directories
-    /// under the database prefix are not. They never take a page slot, so even a `limit`
-    /// smaller than the clutter ahead of the first table returns that table.
-    #[tokio::test]
-    async fn test_listing_ignores_non_table_children() {
-        let (tempdir, db) = setup_database().await;
-        create_tables(&db, &["real"]).await;
-        std::fs::write(tempdir.path().join("aaa-loose.lance"), b"not a table").unwrap();
-        create_dir_all(tempdir.path().join("aaa-scratch")).unwrap();
-
-        let page = db
-            .list_tables(ListTablesRequest {
-                limit: Some(1),
-                ..Default::default()
-            })
-            .await
-            .unwrap();
-
-        assert_eq!(page.tables, vec!["real"]);
-    }
-
-    /// The Lance 11 fallback pages locally over one full-level listing, so a bounded page
-    /// costs exactly one listing call — clutter ahead of the first table must not buy extra
-    /// round trips.
-    #[tokio::test]
-    async fn test_one_full_listing_per_public_page() {
-        use crate::io::object_store::io_tracking::IoStatsHolder;
-        use lance_io::object_store::WrappingObjectStore;
-
-        let (tempdir, mut db) = setup_database().await;
-        create_tables(&db, &["real"]).await;
-        std::fs::write(tempdir.path().join("aaa-loose.lance"), b"not a table").unwrap();
-        create_dir_all(tempdir.path().join("aaa-scratch")).unwrap();
-
-        let io_stats = IoStatsHolder::default();
-        let mut tracked_store = (*db.object_store).clone();
-        tracked_store.inner =
-            io_stats.wrap(&tracked_store.store_prefix, tracked_store.inner.clone());
-        db.object_store = Arc::new(tracked_store);
-
-        let page = db
-            .list_tables(ListTablesRequest {
-                limit: Some(1),
-                ..Default::default()
-            })
-            .await
-            .unwrap();
-
-        assert_eq!(page.tables, vec!["real"]);
-        assert_eq!(io_stats.incremental_stats().read_iops, 1);
-    }
-
-    #[tokio::test]
-    async fn listing_ignores_empty_table_name() {
-        let (tempdir, db) = setup_database().await;
-        create_dir_all(tempdir.path().join(".lance")).unwrap();
-        let page = db.list_tables(ListTablesRequest::default()).await.unwrap();
-        assert!(
-            page.tables.is_empty(),
-            "invalid empty table name was listed"
-        );
-    }
-
     async fn setup_database() -> (tempfile::TempDir, ListingDatabase) {
         let tempdir = tempdir().unwrap();
         let uri = tempdir.path().to_str().unwrap();

rust/lancedb/src/expr.rs

@@ -157,7 +157,7 @@ mod tests {
         use datafusion_common::ScalarValue;
         let expr = col("data").eq(lit(ScalarValue::Binary(Some(vec![0xca, 0xfe]))));
         let sql = expr_to_sql_string(&expr).unwrap();
-        assert_eq!(sql, "(`data` = X'CAFE')");
+        assert_eq!(sql, "(data = X'CAFE')");
     }
 
     #[test]
@@ -167,7 +167,7 @@ mod tests {
         let int_expr = col("id").gt(lit(5i64));
         let combined = bin_expr.and(int_expr);
         let sql = expr_to_sql_string(&combined).unwrap();
-        assert_eq!(sql, "((`data` = X'01') AND (id > 5))");
+        assert_eq!(sql, "((data = X'01') AND (id > 5))");
     }
 
     #[test]
@@ -185,7 +185,7 @@ mod tests {
         // serialized correctly (regression test for placeholder rewrite path).
         let expr = contains(col("data"), lit(ScalarValue::Binary(Some(vec![0xff]))));
         let sql = expr_to_sql_string(&expr).unwrap();
-        assert_eq!(sql, "contains(`data`, X'FF')");
+        assert_eq!(sql, "contains(data, X'FF')");
     }
 
     #[test]
@@ -196,7 +196,7 @@ mod tests {
             .eq(lit(ScalarValue::Binary(Some(vec![0xab, 0xcd]))))
             .not();
         let sql = expr_to_sql_string(&expr).unwrap();
-        assert_eq!(sql, "NOT (`data` = X'ABCD')");
+        assert_eq!(sql, "NOT (data = X'ABCD')");
     }
 
     #[test]
@@ -206,122 +206,6 @@ mod tests {
         assert!(sql.contains("IN"), "expected IN in: {}", sql);
     }
 
-    #[test]
-    fn test_empty_is_in() {
-        let expr = is_in(col("id"), vec![]);
-        assert_eq!(expr_to_sql_string(&expr).unwrap(), "false");
-    }
-
-    #[test]
-    fn test_empty_is_in_discards_binary_children() {
-        use datafusion_common::ScalarValue;
-
-        let expr = is_in(
-            col("payload").eq(lit(ScalarValue::Binary(Some(vec![0x01])))),
-            vec![],
-        );
-        assert_eq!(expr_to_sql_string(&expr).unwrap(), "false");
-    }
-
-    #[test]
-    fn test_keyword_identifier() {
-        let expr = col("null").eq(lit(1i64));
-        assert_eq!(expr_to_sql_string(&expr).unwrap(), "(`null` = 1)");
-    }
-
-    #[test]
-    fn test_decimal_literal_preserves_type() {
-        use datafusion_common::ScalarValue;
-
-        let expr = col("val").lt(lit(ScalarValue::Decimal128(
-            Some(1_234_567_890_123_456_790),
-            19,
-            18,
-        )));
-        let sql = expr_to_sql_string(&expr).unwrap();
-        assert_eq!(
-            sql,
-            "(val < arrow_cast('1.234567890123456790', 'Decimal128(19, 18)'))"
-        );
-    }
-
-    #[test]
-    fn test_non_finite_float_literal_preserves_type() {
-        let expr = col("x").lt(lit(f64::INFINITY));
-        assert_eq!(
-            expr_to_sql_string(&expr).unwrap(),
-            "(x < arrow_cast('inf', 'Float64'))"
-        );
-    }
-
-    #[test]
-    fn test_cast_uses_arrow_type_name() {
-        let string = expr_cast(col("x"), DataType::Utf8);
-        assert_eq!(
-            expr_to_sql_string(&string).unwrap(),
-            "arrow_cast(x, 'Utf8')"
-        );
-
-        let int32 = expr_cast(col("x"), DataType::Int32);
-        assert_eq!(
-            expr_to_sql_string(&int32).unwrap(),
-            "arrow_cast(x, 'Int32')"
-        );
-
-        let expr = expr_cast(col("x"), DataType::Float16).lt(lit(2.0));
-        assert_eq!(
-            expr_to_sql_string(&expr).unwrap(),
-            "(arrow_cast(x, 'Float16') < 2.0)"
-        );
-
-        let decimal = expr_cast(lit("2.00"), DataType::Decimal256(40, 2));
-        assert_eq!(
-            expr_to_sql_string(&decimal).unwrap(),
-            "arrow_cast('2.00', 'Decimal256(40, 2)')"
-        );
-    }
-
-    #[test]
-    fn test_binary_placeholder_does_not_rewrite_user_string() {
-        use datafusion_common::ScalarValue;
-
-        let marker = "__lancedb_binary_placeholder_0__";
-        let expr = col("payload")
-            .eq(lit(ScalarValue::Binary(Some(vec![0x01]))))
-            .or(col("text").eq(lit(marker)));
-        assert_eq!(
-            expr_to_sql_string(&expr).unwrap(),
-            "((payload = X'01') OR (`text` = '__lancedb_binary_placeholder_0__'))"
-        );
-    }
-
-    #[test]
-    fn test_binary_binding_skips_quoted_identifiers() {
-        use datafusion_common::ScalarValue;
-
-        let expr = col("payload")
-            .eq(lit(ScalarValue::Binary(Some(vec![0x01]))))
-            .and(col("odd'name").eq(lit(1i64)))
-            .and(col("odd`'name").eq(lit(2i64)));
-        assert_eq!(
-            expr_to_sql_string(&expr).unwrap(),
-            "(((payload = X'01') AND (`odd'name` = 1)) AND (`odd``'name` = 2))"
-        );
-    }
-
-    #[test]
-    fn test_binary_placeholder_collision_search_is_linear() {
-        use datafusion_common::ScalarValue;
-
-        let collision_shaped = format!("__lancedb_binary_placeholder_0__{}", "_".repeat(64_000));
-        let expr = col("payload")
-            .eq(lit(ScalarValue::Binary(Some(vec![0x01]))))
-            .and(col("text").eq(lit(collision_shaped.clone())));
-        let sql = expr_to_sql_string(&expr).unwrap();
-        assert!(sql.contains("X'01'"));
-        assert!(sql.contains(&format!("'{collision_shaped}'")));
-    }
-
     #[test]
     fn test_multiple_binary_literals() {
         use datafusion_common::ScalarValue;

rust/lancedb/src/expr/sql.rs

@@ -1,24 +1,13 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use std::{
-    any::TypeId,
-    collections::{HashMap, HashSet},
-};
+use std::any::TypeId;
 
-use arrow_array::types::{
-    Decimal32Type, Decimal64Type, Decimal128Type, Decimal256Type, DecimalType,
-};
-use arrow_schema::DataType;
 use datafusion_common::ScalarValue;
 use datafusion_common::tree_node::{Transformed, TreeNode, TreeNodeRecursion};
 use datafusion_expr::Expr;
-use datafusion_functions::core::expr_fn::{
-    arrow_cast as datafusion_arrow_cast, arrow_try_cast as datafusion_arrow_try_cast,
-};
 use datafusion_sql::sqlparser::{
     dialect::{Dialect as SqlParserDialect, GenericDialect},
-    keywords::ALL_KEYWORDS,
     tokenizer::{Token, Tokenizer},
 };
 use datafusion_sql::unparser::{self, dialect::Dialect as UnparserDialect};
@@ -38,13 +27,11 @@ struct LanceSqlDialect;
 
 impl UnparserDialect for LanceSqlDialect {
     fn identifier_quote_style(&self, identifier: &str) -> Option<char> {
-        let identifier_upper = identifier.to_ascii_uppercase();
-        let needs_quote =
-            (identifier_upper != "ID" && ALL_KEYWORDS.contains(&identifier_upper.as_str()))
-                || identifier.chars().any(|c| c.is_ascii_uppercase())
-                || !identifier.chars().enumerate().all(|(i, c)| {
-                    c == '_' || c.is_ascii_alphabetic() || (i > 0 && c.is_ascii_digit())
-                });
+        let needs_quote = identifier.chars().any(|c| c.is_ascii_uppercase())
+            || !identifier
+                .chars()
+                .enumerate()
+                .all(|(i, c)| c == '_' || c.is_ascii_alphabetic() || (i > 0 && c.is_ascii_digit()));
         if needs_quote { Some('`') } else { None }
     }
 }
@@ -113,128 +100,24 @@ fn bytes_to_hex_sql(bytes: &[u8]) -> String {
     format!("X'{hex}'")
 }
 
-fn string_literals(expr: &Expr) -> HashSet<String> {
-    let mut literals = HashSet::new();
+/// Returns true if *expr* contains a `Binary` or `LargeBinary` scalar literal
+/// anywhere in its subtree.  DataFusion's SQL unparser cannot serialize those
+/// variants, so we route such expressions through a placeholder-substitution
+/// path that emits SQL `X'...'` byte-string literals.
+fn has_binary_literal(expr: &Expr) -> bool {
+    let mut found = false;
     let _ = expr.apply(&mut |e: &Expr| {
-        if let Expr::Literal(
-            ScalarValue::Utf8(Some(value))
-            | ScalarValue::LargeUtf8(Some(value))
-            | ScalarValue::Utf8View(Some(value)),
-            _,
-        ) = e
-        {
-            literals.insert(value.clone());
-        }
-        Ok(TreeNodeRecursion::Continue)
-    });
-    literals
-}
-
-fn typed_string_literal(value: String, data_type: DataType) -> Expr {
-    datafusion_arrow_cast(
-        Expr::Literal(ScalarValue::Utf8(Some(value)), None),
-        Expr::Literal(ScalarValue::Utf8(Some(data_type.to_string())), None),
-    )
-}
-
-fn next_binary_placeholder(user_strings: &HashSet<String>, next_id: &mut usize) -> String {
-    loop {
-        let placeholder = format!("{BINARY_PLACEHOLDER_PREFIX}{}__", *next_id);
-        *next_id += 1;
-        if !user_strings.contains(&placeholder) {
-            return placeholder;
-        }
-    }
-}
-
-fn bind_binary_literals(
-    sql: &str,
-    mut bindings: HashMap<String, Vec<u8>>,
-) -> crate::Result<String> {
-    let bytes = sql.as_bytes();
-    let mut output = Vec::with_capacity(bytes.len());
-    let mut index = 0;
-
-    // Walk SQL string tokens once. Placeholders are plain, unescaped string
-    // literals, so this remains linear even when user strings are large or
-    // deliberately resemble the placeholder prefix.
-    while index < bytes.len() {
-        if bytes[index] == b'`' {
-            let identifier_start = index;
-            index += 1;
-            let mut identifier_end = None;
-            while index < bytes.len() {
-                if bytes[index] == b'`' {
-                    if index + 1 < bytes.len() && bytes[index + 1] == b'`' {
-                        index += 2;
-                    } else {
-                        index += 1;
-                        identifier_end = Some(index);
-                        break;
-                    }
-                } else {
-                    index += 1;
-                }
-            }
-
-            let Some(identifier_end) = identifier_end else {
-                return Err(crate::Error::InvalidInput {
-                    message: "unterminated identifier while binding binary literal".to_string(),
-                });
-            };
-            output.extend_from_slice(&bytes[identifier_start..identifier_end]);
-            continue;
-        }
-
-        if bytes[index] != b'\'' {
-            output.push(bytes[index]);
-            index += 1;
-            continue;
-        }
-
-        let literal_start = index;
-        index += 1;
-        let content_start = index;
-        let mut escaped = false;
-        let mut content_end = None;
-        while index < bytes.len() {
-            if bytes[index] == b'\'' {
-                if index + 1 < bytes.len() && bytes[index + 1] == b'\'' {
-                    escaped = true;
-                    index += 2;
-                } else {
-                    content_end = Some(index);
-                    index += 1;
-                    break;
-                }
-            } else {
-                index += 1;
-            }
-        }
-
-        let Some(content_end) = content_end else {
-            return Err(crate::Error::InvalidInput {
-                message: "unterminated string while binding binary literal".to_string(),
-            });
-        };
-
-        let placeholder = &sql[content_start..content_end];
-        if !escaped && let Some(value) = bindings.remove(placeholder) {
-            output.extend_from_slice(bytes_to_hex_sql(&value).as_bytes());
+        if matches!(
+            e,
+            Expr::Literal(ScalarValue::Binary(_) | ScalarValue::LargeBinary(_), _)
+        ) {
+            found = true;
+            Ok(TreeNodeRecursion::Stop)
         } else {
-            output.extend_from_slice(&bytes[literal_start..index]);
+            Ok(TreeNodeRecursion::Continue)
         }
-    }
-
-    if !bindings.is_empty() {
-        return Err(crate::Error::InvalidInput {
-            message: "failed to bind binary literal while serializing expression".to_string(),
-        });
-    }
-
-    String::from_utf8(output).map_err(|e| crate::Error::InvalidInput {
-        message: format!("failed to bind binary literal: {e}"),
-    })
+    });
+    found
 }
 
 fn run_unparser(expr: &Expr) -> crate::Result<String> {
@@ -247,37 +130,25 @@ fn run_unparser(expr: &Expr) -> crate::Result<String> {
 }
 
 pub fn expr_to_sql_string(expr: &Expr) -> crate::Result<String> {
-    // DataFusion's unparser needs a few adaptations before its SQL can be
-    // reparsed by Lance without changing the typed expression's semantics:
-    //
-    // * decimal literals need an explicit cast to preserve precision and scale;
-    // * casts need exact Arrow type names rather than SQL type aliases;
-    // * an empty IN list is valid in DataFusion but invalid SQL;
-    // * binary literals are unsupported by the unparser and need placeholders.
-    // Eliminate empty membership expressions before visiting their children.
-    // Otherwise a discarded binary child could leave behind a stale binding.
+    // Fast path: no binary literals — DataFusion's unparser handles everything.
+    if !has_binary_literal(expr) {
+        return run_unparser(expr);
+    }
+
+    // Slow path: DataFusion's unparser cannot serialize `Binary`/`LargeBinary`
+    // scalars, so we rewrite each one to a unique string-literal placeholder,
+    // let the unparser do the rest of the work, then substitute the SQL
+    // `X'...'` byte-string literal back in.  This keeps the operator/function
+    // serialization logic centralized in DataFusion and works for every
+    // expression node type the unparser supports.
+    let mut bindings: Vec<Vec<u8>> = Vec::new();
     let rewritten = expr
         .clone()
-        .transform(|e: Expr| match e {
-            Expr::InList(in_list) if in_list.list.is_empty() => Ok(Transformed::yes(
-                Expr::Literal(ScalarValue::Boolean(Some(in_list.negated)), None),
-            )),
-            other => Ok(Transformed::no(other)),
-        })
-        .map_err(|e| crate::Error::InvalidInput {
-            message: format!("failed to rewrite expression: {e}"),
-        })?
-        .data;
-
-    let user_strings = string_literals(&rewritten);
-    let mut next_placeholder_id = 0;
-    let mut binary_bindings = HashMap::new();
-    let rewritten = rewritten
         .transform(|e: Expr| match e {
             Expr::Literal(ScalarValue::Binary(Some(bytes)), m)
             | Expr::Literal(ScalarValue::LargeBinary(Some(bytes)), m) => {
-                let placeholder = next_binary_placeholder(&user_strings, &mut next_placeholder_id);
-                binary_bindings.insert(placeholder.clone(), bytes);
+                let placeholder = format!("{}{}__", BINARY_PLACEHOLDER_PREFIX, bindings.len());
+                bindings.push(bytes);
                 Ok(Transformed::yes(Expr::Literal(
                     ScalarValue::Utf8(Some(placeholder)),
                     m,
@@ -287,57 +158,6 @@ pub fn expr_to_sql_string(expr: &Expr) -> crate::Result<String> {
             | Expr::Literal(ScalarValue::LargeBinary(None), m) => {
                 Ok(Transformed::yes(Expr::Literal(ScalarValue::Null, m)))
             }
-            Expr::Literal(ScalarValue::Decimal32(Some(value), precision, scale), _m) => {
-                let value = Decimal32Type::format_decimal(value, precision, scale);
-                Ok(Transformed::yes(typed_string_literal(
-                    value,
-                    DataType::Decimal32(precision, scale),
-                )))
-            }
-            Expr::Literal(ScalarValue::Decimal64(Some(value), precision, scale), _m) => {
-                let value = Decimal64Type::format_decimal(value, precision, scale);
-                Ok(Transformed::yes(typed_string_literal(
-                    value,
-                    DataType::Decimal64(precision, scale),
-                )))
-            }
-            Expr::Literal(ScalarValue::Decimal128(Some(value), precision, scale), _m) => {
-                let value = Decimal128Type::format_decimal(value, precision, scale);
-                Ok(Transformed::yes(typed_string_literal(
-                    value,
-                    DataType::Decimal128(precision, scale),
-                )))
-            }
-            Expr::Literal(ScalarValue::Decimal256(Some(value), precision, scale), _m) => {
-                let value = Decimal256Type::format_decimal(value, precision, scale);
-                Ok(Transformed::yes(typed_string_literal(
-                    value,
-                    DataType::Decimal256(precision, scale),
-                )))
-            }
-            Expr::Literal(ScalarValue::Float16(Some(value)), _m) if !value.is_finite() => Ok(
-                Transformed::yes(typed_string_literal(value.to_string(), DataType::Float16)),
-            ),
-            Expr::Literal(ScalarValue::Float32(Some(value)), _m) if !value.is_finite() => Ok(
-                Transformed::yes(typed_string_literal(value.to_string(), DataType::Float32)),
-            ),
-            Expr::Literal(ScalarValue::Float64(Some(value)), _m) if !value.is_finite() => Ok(
-                Transformed::yes(typed_string_literal(value.to_string(), DataType::Float64)),
-            ),
-            Expr::Cast(cast) => Ok(Transformed::yes(datafusion_arrow_cast(
-                *cast.expr,
-                Expr::Literal(
-                    ScalarValue::Utf8(Some(cast.field.data_type().to_string())),
-                    None,
-                ),
-            ))),
-            Expr::TryCast(cast) => Ok(Transformed::yes(datafusion_arrow_try_cast(
-                *cast.expr,
-                Expr::Literal(
-                    ScalarValue::Utf8(Some(cast.field.data_type().to_string())),
-                    None,
-                ),
-            ))),
             other => Ok(Transformed::no(other)),
         })
         .map_err(|e| crate::Error::InvalidInput {
@@ -345,12 +165,14 @@ pub fn expr_to_sql_string(expr: &Expr) -> crate::Result<String> {
         })?
         .data;
 
-    let sql = run_unparser(&rewritten)?;
-    if binary_bindings.is_empty() {
-        Ok(sql)
-    } else {
-        bind_binary_literals(&sql, binary_bindings)
+    let mut sql = run_unparser(&rewritten)?;
+    for (i, bytes) in bindings.iter().enumerate() {
+        // The unparser quotes string literals with single quotes, so the
+        // placeholder appears as `'__lancedb_binary_placeholder_<i>__'`.
+        let quoted = format!("'{}{}__'", BINARY_PLACEHOLDER_PREFIX, i);
+        sql = sql.replace(&quoted, &bytes_to_hex_sql(bytes));
     }
+    Ok(sql)
 }
 
 #[cfg(test)]

rust/lancedb/src/function.rs

@@ -15,9 +15,6 @@ use serde_json::Value;
 
 use crate::{Error, Result};
 
-/// Semantic Function type for a Blob v2 value.
-pub const FUNCTION_BLOB_V2_TYPE: &str = "blob_v2";
-
 fn invalid_json(error: impl std::fmt::Display) -> Error {
     Error::InvalidInput {
         message: format!("invalid remote Function JSON: {error}"),
@@ -210,33 +207,6 @@ pub enum PythonRuntimeSpec {
         environment: PythonEnvironmentSpec,
         env: BTreeMap<String, String>,
     },
-    /// The GPU-enabled Sophon-managed Python runtime.
-    ///
-    /// # Examples
-    ///
-    /// ```
-    /// use std::collections::BTreeMap;
-    /// use lancedb::function::{PythonEnvironmentSpec, PythonRuntimeSpec};
-    ///
-    /// let runtime = PythonRuntimeSpec::PythonV2 {
-    ///     python_version: "3.12".to_string(),
-    ///     environment: PythonEnvironmentSpec {
-    ///         kind: "pip".to_string(),
-    ///         packages: vec!["cupy-cuda12x".to_string()],
-    ///         channels: Vec::new(),
-    ///         path: None,
-    ///         modules: Vec::new(),
-    ///         image: None,
-    ///     },
-    ///     env: BTreeMap::new(),
-    /// };
-    /// assert!(runtime.requires_gpu());
-    /// ```
-    PythonV2 {
-        python_version: String,
-        environment: PythonEnvironmentSpec,
-        env: BTreeMap<String, String>,
-    },
     /// A runtime kind introduced by a newer server.
     ///
     /// Unknown payload fields are intentionally not retained because the
@@ -249,27 +219,22 @@ impl PythonRuntimeSpec {
     pub fn kind(&self) -> &str {
         match self {
             Self::Python { .. } => "python",
-            Self::PythonV2 { .. } => "python_v2",
             Self::Unrecognized { kind } => kind,
         }
     }
 
-    /// The Python version for a known Python runtime, or `None` for an unknown kind.
+    /// The Python version for the V1 runtime, or `None` for an unknown kind.
     pub fn python_version(&self) -> Option<&str> {
         match self {
-            Self::Python { python_version, .. } | Self::PythonV2 { python_version, .. } => {
-                Some(python_version)
-            }
+            Self::Python { python_version, .. } => Some(python_version),
             Self::Unrecognized { .. } => None,
         }
     }
 
-    /// The Python environment for a known Python runtime, or `None` for an unknown kind.
+    /// The Python environment for the V1 runtime, or `None` for an unknown kind.
     pub fn environment(&self) -> Option<&PythonEnvironmentSpec> {
         match self {
-            Self::Python { environment, .. } | Self::PythonV2 { environment, .. } => {
-                Some(environment)
-            }
+            Self::Python { environment, .. } => Some(environment),
             Self::Unrecognized { .. } => None,
         }
     }
@@ -277,73 +242,38 @@ impl PythonRuntimeSpec {
     /// Environment variables, or `None` for an unknown kind.
     pub fn env(&self) -> Option<&BTreeMap<String, String>> {
         match self {
-            Self::Python { env, .. } | Self::PythonV2 { env, .. } => Some(env),
+            Self::Python { env, .. } => Some(env),
             Self::Unrecognized { .. } => None,
         }
     }
-
-    /// Whether the runtime requires a GPU selected by the execution platform.
-    pub fn requires_gpu(&self) -> bool {
-        matches!(self, Self::PythonV2 { .. })
-    }
 }
 
 #[derive(Deserialize)]
-struct PythonRuntimeV1Wire {
-    python_version: String,
-    environment: PythonEnvironmentSpec,
+struct PythonRuntimeWire {
+    kind: String,
+    #[serde(default)]
+    python_version: Option<String>,
+    #[serde(default)]
+    environment: Option<PythonEnvironmentSpec>,
     #[serde(default)]
     env: BTreeMap<String, String>,
-    #[serde(default)]
-    gpu: Option<Value>,
-}
-
-#[derive(Deserialize)]
-struct PythonRuntimeV2Wire {
-    python_version: String,
-    environment: PythonEnvironmentSpec,
-    #[serde(default)]
-    env: BTreeMap<String, String>,
-    gpu: bool,
 }
 
 impl<'de> Deserialize<'de> for PythonRuntimeSpec {
     fn deserialize<D: Deserializer<'de>>(deserializer: D) -> std::result::Result<Self, D::Error> {
-        let value = Value::deserialize(deserializer)?;
-        let kind = value
-            .get("kind")
-            .ok_or_else(|| de::Error::missing_field("kind"))?
-            .as_str()
-            .ok_or_else(|| de::Error::custom("runtime.kind must be a string"))?
-            .to_string();
-        match kind.as_str() {
-            "python" => {
-                let wire: PythonRuntimeV1Wire =
-                    serde_json::from_value(value).map_err(de::Error::custom)?;
-                if wire.gpu.is_some() {
-                    return Err(de::Error::custom(
-                        "python runtime with gpu requires kind='python_v2'",
-                    ));
-                }
-                Ok(Self::Python {
-                    python_version: wire.python_version,
-                    environment: wire.environment,
-                    env: wire.env,
-                })
-            }
-            "python_v2" => {
-                let wire: PythonRuntimeV2Wire =
-                    serde_json::from_value(value).map_err(de::Error::custom)?;
-                if !wire.gpu {
-                    return Err(de::Error::custom("runtime.gpu must be true"));
-                }
-                Ok(Self::PythonV2 {
-                    python_version: wire.python_version,
-                    environment: wire.environment,
-                    env: wire.env,
-                })
-            }
-            _ => Ok(Self::Unrecognized { kind }),
+        let wire = PythonRuntimeWire::deserialize(deserializer)?;
+        if wire.kind == "python" {
+            Ok(Self::Python {
+                python_version: wire
+                    .python_version
+                    .ok_or_else(|| de::Error::missing_field("python_version"))?,
+                environment: wire
+                    .environment
+                    .ok_or_else(|| de::Error::missing_field("environment"))?,
+                env: wire.env,
+            })
+        } else {
+            Ok(Self::Unrecognized { kind: wire.kind })
         }
     }
 }
@@ -357,8 +287,6 @@ impl Serialize for PythonRuntimeSpec {
             environment: &'a PythonEnvironmentSpec,
             #[serde(skip_serializing_if = "BTreeMap::is_empty")]
             env: &'a BTreeMap<String, String>,
-            #[serde(skip_serializing_if = "Option::is_none")]
-            gpu: Option<bool>,
         }
 
         #[derive(Serialize)]
@@ -376,19 +304,6 @@ impl Serialize for PythonRuntimeSpec {
                 python_version,
                 environment,
                 env,
-                gpu: None,
-            }
-            .serialize(serializer),
-            Self::PythonV2 {
-                python_version,
-                environment,
-                env,
-            } => PythonRuntimeRef {
-                kind: "python_v2",
-                python_version,
-                environment,
-                env,
-                gpu: Some(true),
             }
             .serialize(serializer),
             Self::Unrecognized { kind } => UnrecognizedRuntimeRef { kind }.serialize(serializer),
@@ -398,8 +313,8 @@ impl Serialize for PythonRuntimeSpec {
 
 /// Immutable Function version returned by the Enterprise catalog.
 ///
-/// The GPU execution requirement is part of this identity. CPU and memory sizing,
-/// priority, concurrency, and retry policy belong to the execution platform.
+/// Scheduling resources, priority, concurrency, and retry policy belong to
+/// the submitting Job and are not part of this identity.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct FunctionVersion {
     name: String,
@@ -674,7 +589,7 @@ impl_json!(RefreshColumnResult);
 
 #[cfg(test)]
 mod conda_environment_tests {
-    use super::{PythonEnvironmentSpec, PythonRuntimeSpec};
+    use super::PythonEnvironmentSpec;
 
     #[test]
     fn conda_channels_round_trip_and_pip_stays_bare() {
@@ -693,45 +608,4 @@ mod conda_environment_tests {
             serde_json::from_str(r#"{"kind":"pip","packages":["numpy"]}"#).unwrap();
         assert!(!serde_json::to_string(&pip).unwrap().contains("channels"));
     }
-
-    #[test]
-    fn gpu_python_runtime_marker_round_trips_and_validates() {
-        let runtime: PythonRuntimeSpec = serde_json::from_str(
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":true}"#,
-        )
-        .unwrap();
-        assert_eq!(runtime.kind(), "python_v2");
-        assert!(runtime.requires_gpu());
-        assert_eq!(
-            super::canonical_json(&runtime).unwrap(),
-            r#"{"environment":{"kind":"pip"},"gpu":true,"kind":"python_v2","python_version":"3.12"}"#
-        );
-
-        for invalid in [
-            r#"{"kind":"python","python_version":"3.12","environment":{"kind":"pip"},"gpu":true}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"}}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":1}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":false}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":"true"}"#,
-            r#"{"kind":"python_v2","python_version":"3.12","environment":{"kind":"pip"},"gpu":"H100"}"#,
-        ] {
-            assert!(serde_json::from_str::<PythonRuntimeSpec>(invalid).is_err());
-        }
-    }
-
-    #[test]
-    fn unknown_runtime_discards_payload_before_known_field_validation() {
-        for encoded in [
-            r#"{"kind":"python_v3","gpu":{"model":"H100"}}"#,
-            r#"{"kind":"python_v3","resources":[]}"#,
-            r#"{"kind":"python_v3","python_version":3.15,"environment":{"kind":[]}}"#,
-        ] {
-            let runtime: PythonRuntimeSpec = serde_json::from_str(encoded).unwrap();
-            assert_eq!(runtime.kind(), "python_v3");
-            assert_eq!(
-                super::canonical_json(&runtime).unwrap(),
-                r#"{"kind":"python_v3"}"#
-            );
-        }
-    }
 }

rust/lancedb/src/io/object_store.rs

@@ -10,7 +10,7 @@ use lance::io::WrappingObjectStore;
 use object_store::{
     CopyOptions, Error, GetOptions, GetResult, ListResult, MultipartUpload, ObjectMeta,
     ObjectStore, ObjectStoreExt, PutMultipartOptions, PutOptions, PutPayload, PutResult, Result,
-    UploadPart, path::Path,
+    UploadPart, list::PaginatedListStore, path::Path,
 };
 
 use async_trait::async_trait;
@@ -187,6 +187,14 @@ impl WrappingObjectStore for MirroringObjectStoreWrapper {
             secondary: self.secondary.clone(),
         })
     }
+
+    fn wrap_paginated(
+        &self,
+        _store_prefix: &str,
+        original: Arc<dyn PaginatedListStore>,
+    ) -> Option<Arc<dyn PaginatedListStore>> {
+        Some(original)
+    }
 }
 
 // windows pathing can't be simply concatenated

rust/lancedb/src/io/object_store/io_tracking.rs

@@ -12,7 +12,7 @@ use lance::io::WrappingObjectStore;
 use object_store::{
     CopyOptions, GetOptions, GetResult, ListResult, MultipartUpload, ObjectMeta, ObjectStore,
     PutMultipartOptions, PutOptions, PutPayload, PutResult, RenameOptions, Result as OSResult,
-    UploadPart, path::Path,
+    UploadPart, list::PaginatedListStore, path::Path,
 };
 
 #[derive(Debug, Default)]
@@ -57,6 +57,14 @@ impl WrappingObjectStore for IoStatsHolder {
             stats: self.0.clone(),
         })
     }
+
+    fn wrap_paginated(
+        &self,
+        _store_prefix: &str,
+        original: Arc<dyn PaginatedListStore>,
+    ) -> Option<Arc<dyn PaginatedListStore>> {
+        Some(original)
+    }
 }
 
 impl IoTrackingStore {

rust/lancedb/src/remote/table.rs

@@ -3180,8 +3180,8 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
             self.schema().await?.as_ref(),
             "schema evolution",
         )?;
-        // The server plans the declaration against its table schema, including
-        // Blob v2 semantics inherited by a direct field projection.
+        // The server plans the declaration: expression validation, type
+        // inference and the persisted binding all happen there.
         let entries = columns
             .iter()
             .map(
@@ -5734,8 +5734,9 @@ mod tests {
             ))
             .execute()
             .await;
-        let Err(err) = result else {
-            panic!("legacy remote query unexpectedly succeeded")
+        let err = match result {
+            Ok(_) => panic!("legacy remote query unexpectedly succeeded"),
+            Err(err) => err,
         };
 
         assert!(
@@ -7387,8 +7388,8 @@ mod tests {
         assert_eq!(result.version, if old_server { 0 } else { 43 });
     }
 
-    /// A declaration is sent as `{name, computed}` for the server to plan; the
-    /// client never types the expression itself.
+    /// A declaration is sent as `{name, computed}` entries for the server to
+    /// plan; the client never types the expression itself.
     #[tokio::test]
     async fn test_add_computed_columns_sends_the_expression() {
         let table = Table::new_with_handler("my_table", |request| match request.url().path() {
@@ -7464,93 +7465,6 @@ mod tests {
         assert_eq!(result.version, 8);
     }
 
-    #[tokio::test]
-    async fn test_add_function_column_allows_an_existing_binding() {
-        let binding = crate::function::FunctionBinding::from_json(include_str!(
-            "../../tests/fixtures/first_class_functions/v1/remote_function_binding.json"
-        ))
-        .unwrap();
-        let binding_metadata = crate::table::computed_columns::function_bindings_metadata(
-            std::slice::from_ref(&binding),
-        )
-        .unwrap();
-        let mut fields = vec![
-            Field::new("title", DataType::Utf8, true),
-            Field::new("body", DataType::Utf8, true),
-        ];
-        fields.extend(binding.outputs().iter().map(|output| {
-            let data_type = match output.arrow_type.as_str() {
-                "utf8" => DataType::Utf8,
-                "int64" => DataType::Int64,
-                other => panic!("unexpected fixture output type {other}"),
-            };
-            Field::new(&output.output_name, data_type, true).with_metadata(
-                crate::table::computed_columns::function_computed_column_metadata(
-                    binding.binding_id(),
-                    output.output_ordinal,
-                    &["title".into(), "body".into()],
-                ),
-            )
-        }));
-        let schema = Schema::new_with_metadata(
-            fields,
-            HashMap::from([(
-                crate::table::computed_columns::FUNCTION_BINDINGS_META_KEY.to_string(),
-                binding_metadata,
-            )]),
-        );
-        let table =
-            Table::new_with_handler("my_table", move |request| match request.url().path() {
-                "/v1/table/my_table/describe/" => http::Response::builder()
-                    .status(200)
-                    .body(describe_response(&schema))
-                    .unwrap(),
-                "/v1/table/my_table/add_columns/" => {
-                    let actual: serde_json::Value =
-                        serde_json::from_slice(request.body().unwrap().as_bytes().unwrap())
-                            .unwrap();
-                    assert_eq!(
-                        actual["new_columns"],
-                        serde_json::json!([
-                            {"name":"secondary_text","all_null":true},
-                            {"name":"secondary_token_count","all_null":true}
-                        ])
-                    );
-                    http::Response::builder()
-                        .status(200)
-                        .body(r#"{"version":10}"#.to_string())
-                        .unwrap()
-                }
-                path => panic!("Unexpected path: {path}"),
-            });
-        let application = crate::function::FunctionApplication::from_json(
-            r#"{
-                "function":{"name":"text_features","version":"fv_01K3TEXT"},
-                "inputs":[
-                    {"parameter":"title","kind":"column","value":{"path":"title"}},
-                    {"parameter":"body","kind":"column","value":{"path":"body"}}
-                ],
-                "output":{"kind":"named_struct","fields":[
-                    {"name":"normalized_text","arrow_type":"utf8","nullable":false},
-                    {"name":"token_count","arrow_type":"int64","nullable":false}
-                ]},
-                "columns":{
-                    "normalized_text":"secondary_text",
-                    "token_count":"secondary_token_count"
-                }
-            }"#,
-        )
-        .unwrap();
-
-        let result = table
-            .add_columns()
-            .function(application)
-            .execute()
-            .await
-            .unwrap();
-        assert_eq!(result.version, 10);
-    }
-
     #[tokio::test]
     async fn test_add_fixed_size_list_function_column_declares_the_vector_type() {
         let table = Table::new_with_handler("my_table", |request| {

rust/lancedb/src/table.rs

@@ -750,8 +750,8 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
     /// Declare computed columns, each defined by a SQL expression.
     ///
     /// Where the declaration is planned depends on the backend: a local table
-    /// validates and types the expression itself, while a remote one sends the
-    /// expression for the server to plan.
+    /// validates and types the expression itself, a remote one sends the text
+    /// for the server to plan.
     async fn add_computed_columns(
         &self,
         _columns: &[(String, String)],
@@ -4183,6 +4183,14 @@ mod tests {
                 parent_list_calls: self.parent_list_calls.clone(),
             })
         }
+
+        fn wrap_paginated(
+            &self,
+            _store_prefix: &str,
+            _original: Arc<dyn object_store::list::PaginatedListStore>,
+        ) -> Option<Arc<dyn object_store::list::PaginatedListStore>> {
+            None
+        }
     }
 
     #[tokio::test]
@@ -4286,6 +4294,14 @@ mod tests {
             self.called.store(true, Ordering::Relaxed);
             original
         }
+
+        fn wrap_paginated(
+            &self,
+            _store_prefix: &str,
+            original: Arc<dyn object_store::list::PaginatedListStore>,
+        ) -> Option<Arc<dyn object_store::list::PaginatedListStore>> {
+            Some(original)
+        }
     }
 
     #[tokio::test]
@@ -5747,13 +5763,6 @@ mod tests {
         assert!(index_bytes > 0);
         assert_eq!(with_index, data_only + index_bytes);
 
-        // Release builds reject unstable overlay datasets unless explicitly opted in.
-        if !lance_table::feature_flags::can_read_dataset(
-            lance_table::feature_flags::FLAG_UNSTABLE_DATA_OVERLAY_FILES,
-        ) {
-            return;
-        }
-
         // Commit an overlay file supplying new `foo` values for the first three
         // rows of fragment 0. There is no high-level API that writes overlays
         // yet, so write the overlay's data file and commit the `DataOverlay`

rust/lancedb/src/table/datafusion/blob_coerce.rs

@@ -36,14 +36,6 @@ pub(super) fn coerce_blob_expr(
     };
 
     let input_shape = match input_field.data_type() {
-        DataType::Null => {
-            let expr: Arc<dyn PhysicalExpr> = Arc::new(CastExpr::new(
-                input_expr,
-                table_field.data_type().clone(),
-                None,
-            ));
-            return Ok((expr, table_field.clone()));
-        }
         DataType::Binary | DataType::LargeBinary | DataType::BinaryView => BlobInputShape::Bytes,
         DataType::Utf8 | DataType::LargeUtf8 | DataType::Utf8View => BlobInputShape::String,
         DataType::Struct(children) => {
@@ -163,7 +155,7 @@ mod tests {
     use crate::blob::blob;
     use arrow_array::{
         Array, ArrayRef, BinaryArray, BinaryViewArray, Int32Array, Int64Array, LargeBinaryArray,
-        NullArray, RecordBatch, StringArray, StringViewArray, StructArray, UInt8Array, UInt64Array,
+        RecordBatch, StringArray, StringViewArray, StructArray, UInt8Array, UInt64Array,
     };
     use arrow_schema::Schema;
     use datafusion::prelude::SessionContext;
@@ -287,18 +279,6 @@ mod tests {
         assert_eq!(data.value(0), b"view");
     }
 
-    #[tokio::test]
-    async fn null_column_coerces_to_all_null_blob_struct() {
-        let batch = batch_with_image(
-            Field::new("image", DataType::Null, true),
-            Arc::new(NullArray::new(2)),
-        );
-        let coerced = coerce(batch, &blob_table_schema()).await;
-        let image = image_struct(&coerced);
-        assert!(image.is_null(0));
-        assert!(image.is_null(1));
-    }
-
     #[tokio::test]
     async fn binary_nulls_stay_null_after_coercion() {
         let batch = batch_with_image(

rust/lancedb/src/table/merge.rs

@@ -1431,4 +1431,195 @@ mod lsm_tests {
             "LSM vector search must rank the memtable row first"
         );
     }
+
+    #[tokio::test]
+    async fn lsm_cosine_distance_scale_and_mixed_tier_ordering() {
+        use arrow::array::{FixedSizeListBuilder, Float32Builder};
+        use arrow::datatypes::Float32Type;
+
+        use crate::index::Index;
+        use crate::index::vector::IvfPqIndexBuilder;
+
+        const DIM: usize = 8;
+        const N: usize = 256;
+
+        fn normalized_vector(state: &mut u64) -> Vec<f32> {
+            let mut vector = (0..DIM)
+                .map(|_| {
+                    *state = state
+                        .wrapping_mul(6_364_136_223_846_793_005)
+                        .wrapping_add(1);
+                    ((*state >> 32) as u32 as f32 / u32::MAX as f32) * 2.0 - 1.0
+                })
+                .collect::<Vec<_>>();
+            let norm = vector.iter().map(|value| value * value).sum::<f32>().sqrt();
+            vector.iter_mut().for_each(|value| *value /= norm);
+            vector
+        }
+
+        let schema = Arc::new(Schema::new(vec![
+            Field::new("id", DataType::Int64, false),
+            Field::new(
+                "vec",
+                DataType::FixedSizeList(
+                    Arc::new(Field::new("item", DataType::Float32, true)),
+                    DIM as i32,
+                ),
+                false,
+            ),
+        ]));
+        let make_batch = |rows: Vec<(i64, Vec<f32>)>| {
+            let ids = rows.iter().map(|(id, _)| *id).collect::<Vec<_>>();
+            let mut vectors = FixedSizeListBuilder::new(Float32Builder::new(), DIM as i32);
+            for (_, vector) in &rows {
+                vectors.values().append_slice(vector);
+                vectors.append(true);
+            }
+            RecordBatch::try_new(
+                schema.clone(),
+                vec![Arc::new(Int64Array::from(ids)), Arc::new(vectors.finish())],
+            )
+            .unwrap()
+        };
+        let first_result = |batches: &[RecordBatch]| {
+            let batch = &batches[0];
+            let id = batch["id"].as_primitive::<Int64Type>().value(0);
+            let distance = batch["_distance"].as_primitive::<Float32Type>().value(0);
+            (id, distance)
+        };
+
+        let mut state = 42;
+        let base_rows = (0..N)
+            .map(|id| (id as i64, normalized_vector(&mut state)))
+            .collect::<Vec<_>>();
+        let query = normalized_vector(&mut state);
+
+        let dir = tempdir().unwrap();
+        let conn = connect(dir.path().to_str().unwrap())
+            .execute()
+            .await
+            .unwrap();
+        let base = make_batch(base_rows);
+        let reader: Box<dyn RecordBatchReader + Send> =
+            Box::new(RecordBatchIterator::new(vec![Ok(base)], schema.clone()));
+        let table = conn
+            .create_table("cosine_lsm", reader)
+            .execute()
+            .await
+            .unwrap();
+        table.set_unenforced_primary_key(["id"]).await.unwrap();
+        table
+            .create_index(
+                &["vec"],
+                Index::IvfPq(
+                    IvfPqIndexBuilder::default()
+                        .distance_type(crate::DistanceType::Cosine)
+                        .num_partitions(1)
+                        .num_sub_vectors(1),
+                ),
+            )
+            .name("vec_cosine".to_string())
+            .execute()
+            .await
+            .unwrap();
+        table
+            .set_lsm_write_spec(
+                LsmWriteSpec::unsharded().with_maintained_indexes(vec!["vec_cosine".to_string()]),
+            )
+            .await
+            .unwrap();
+
+        let base_only = table
+            .query()
+            .nearest_to(query.as_slice())
+            .unwrap()
+            .limit(1)
+            .use_lsm(false)
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let (base_id, public_distance) = first_result(&base_only);
+
+        let lsm = table
+            .query()
+            .nearest_to(query.as_slice())
+            .unwrap()
+            .limit(1)
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let (lsm_id, lsm_distance) = first_result(&lsm);
+        assert_eq!(lsm_id, base_id);
+        assert!(
+            (lsm_distance - public_distance).abs() < 1e-5,
+            "LSM cosine distance {lsm_distance} did not use the public scale {public_distance}"
+        );
+
+        // Add an exact memtable result whose distance lies between the public ANN
+        // score and its doubled internal score. Correctly normalized plans still
+        // rank the ANN row first; mixed units would incorrectly rank this row first.
+        assert!(public_distance > 0.0 && public_distance < 4.0 / 3.0);
+        let memtable_distance = public_distance * 1.5;
+        let cosine_similarity = 1.0 - memtable_distance;
+        let mut orthogonal = normalized_vector(&mut state);
+        let projection = orthogonal
+            .iter()
+            .zip(&query)
+            .map(|(left, right)| left * right)
+            .sum::<f32>();
+        for (value, query_value) in orthogonal.iter_mut().zip(&query) {
+            *value -= projection * query_value;
+        }
+        let norm = orthogonal
+            .iter()
+            .map(|value| value * value)
+            .sum::<f32>()
+            .sqrt();
+        orthogonal.iter_mut().for_each(|value| *value /= norm);
+        let sine = (1.0 - cosine_similarity * cosine_similarity).sqrt();
+        let memtable_vector = query
+            .iter()
+            .zip(&orthogonal)
+            .map(|(query_value, orthogonal_value)| {
+                cosine_similarity * query_value + sine * orthogonal_value
+            })
+            .collect::<Vec<_>>();
+
+        let mut merge = table.merge_insert(&[]);
+        merge
+            .when_matched_update_all(None)
+            .when_not_matched_insert_all();
+        let memtable = make_batch(vec![(N as i64, memtable_vector)]);
+        merge
+            .execute(Box::new(RecordBatchIterator::new(
+                vec![Ok(memtable)],
+                schema,
+            )))
+            .await
+            .unwrap();
+
+        let mixed = table
+            .query()
+            .nearest_to(query.as_slice())
+            .unwrap()
+            .limit(1)
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let (mixed_id, mixed_distance) = first_result(&mixed);
+        assert_eq!(
+            mixed_id, base_id,
+            "mixed LSM tiers must compare ANN and exact distances in public units"
+        );
+        assert!((mixed_distance - public_distance).abs() < 1e-5);
+    }
 }

rust/lancedb/src/table/query.rs

@@ -1,7 +1,10 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-use std::sync::Arc;
+use std::{
+    collections::{HashSet, VecDeque},
+    sync::Arc,
+};
 
 mod lsm;
 
@@ -17,15 +20,23 @@ use arrow::array::{AsArray, FixedSizeListBuilder, Float32Builder};
 use arrow::datatypes::{Float32Type, UInt8Type};
 use arrow_array::Array;
 use arrow_schema::{DataType, Schema};
-use datafusion_common::{Column, DataFusionError, SchemaError};
-use datafusion_physical_plan::ExecutionPlan;
+use datafusion_common::{Column, DataFusionError, ScalarValue, SchemaError};
+use datafusion_expr::Operator;
+use datafusion_physical_expr::expressions::{BinaryExpr, Column as PhysicalColumn, Literal};
+use datafusion_physical_plan::PhysicalExpr;
 use datafusion_physical_plan::projection::ProjectionExec;
 use datafusion_physical_plan::repartition::RepartitionExec;
 use datafusion_physical_plan::union::UnionExec;
+use datafusion_physical_plan::{ExecutionPlan, with_new_children_if_necessary};
 use lance::dataset::mem_wal::DatasetMemWalExt;
 use lance::dataset::scanner::DatasetRecordBatchStream;
 use lance::dataset::scanner::Scanner;
+use lance::index::DatasetIndexInternalExt;
+use lance::io::exec::ANNIvfSubIndexExec;
 use lance_datafusion::exec::{analyze_plan as lance_analyze_plan, execute_plan};
+use lance_index::metrics::NoOpMetricsCollector;
+use lance_index::vector::{DIST_COL, quantizer::QuantizationType};
+use lance_linalg::distance::DistanceType as LanceDistanceType;
 use lance_namespace::LanceNamespace;
 use lance_namespace::models::{
     QueryTableRequest as NsQueryTableRequest, QueryTableRequestColumns,
@@ -375,10 +386,30 @@ pub async fn create_plan(
         scanner.order_by(Some(order_by.clone()))?;
     }
 
-    scanner
+    let mut plan = scanner
         .create_plan()
         .await
-        .map_err(|error| enrich_lance_field_not_found(error, schema))
+        .map_err(|error| enrich_lance_field_not_found(error, schema))?;
+    let normalized_l2_indices = normalized_l2_ann_indices(plan.as_ref()).await?;
+    if !normalized_l2_indices.is_empty() {
+        // Rebuild only the affected ANN nodes with internal normalized squared-L2
+        // bounds. Exact branches keep the public cosine bounds from `plan`.
+        let internal_plan = if query.lower_bound.is_some() || query.upper_bound.is_some() {
+            scanner.distance_range(
+                query.lower_bound.map(|bound| bound / COSINE_ANN_SCALE),
+                query.upper_bound.map(|bound| bound / COSINE_ANN_SCALE),
+            );
+            scanner
+                .create_plan()
+                .await
+                .map_err(|error| enrich_lance_field_not_found(error, schema))?
+        } else {
+            plan.clone()
+        };
+        plan = normalize_ann_branches(plan, internal_plan, &normalized_l2_indices)?;
+    }
+
+    Ok(plan)
 }
 
 /// Replace DataFusion's top-level field candidates with qualified leaf paths.
@@ -470,6 +501,184 @@ fn leaf_field_paths(schema: &Schema) -> Vec<String> {
 
 //Helper functions below
 
+const COSINE_ANN_SCALE: f32 = 0.5;
+
+/// Find ANN index segments whose scores use normalized squared L2 for cosine search.
+///
+/// Cosine PQ/SQ/RQ indices normalize their vectors and use squared L2 internally.  This
+/// preserves ranking, but squared L2 over unit vectors is twice the cosine distance.  Flat
+/// cosine indices calculate cosine directly, so they are not included.
+async fn normalized_l2_ann_indices(plan: &dyn ExecutionPlan) -> Result<HashSet<String>> {
+    let mut ann_plans = Vec::new();
+    find_ann_plans(plan, &mut ann_plans);
+
+    let mut checked = HashSet::new();
+    let mut normalized_l2 = HashSet::new();
+    for ann in ann_plans {
+        if ann.query().metric_type != Some(LanceDistanceType::Cosine) {
+            continue;
+        }
+        for index in ann.indices() {
+            let uuid = index.uuid.to_string();
+            if !checked.insert(uuid.clone()) {
+                continue;
+            }
+            let vector_index = ann
+                .dataset()
+                .open_vector_index(&ann.query().column, &index.uuid, &NoOpMetricsCollector)
+                .await?;
+            let (_, quantization_type) = vector_index.sub_index_type();
+            if matches!(
+                quantization_type,
+                QuantizationType::Product | QuantizationType::Scalar | QuantizationType::Rabit
+            ) {
+                normalized_l2.insert(uuid);
+            }
+        }
+    }
+    Ok(normalized_l2)
+}
+
+/// Normalize affected ANN outputs before their parent plan nodes consume them.
+///
+/// This is used by planners that do not support distance ranges, such as the MemWAL
+/// LSM planner. The standard scanner path rebuilds a second plan when it also needs
+/// to translate range bounds, then calls [`normalize_ann_branches`] directly.
+pub(super) async fn normalize_cosine_ann_branches(
+    plan: Arc<dyn ExecutionPlan>,
+) -> Result<Arc<dyn ExecutionPlan>> {
+    let normalized_l2_indices = normalized_l2_ann_indices(plan.as_ref()).await?;
+    if normalized_l2_indices.is_empty() {
+        return Ok(plan);
+    }
+    normalize_ann_branches(plan.clone(), plan, &normalized_l2_indices)
+}
+
+fn find_ann_plans<'a>(plan: &'a dyn ExecutionPlan, ann_plans: &mut Vec<&'a ANNIvfSubIndexExec>) {
+    if let Some(ann) = plan.downcast_ref::<ANNIvfSubIndexExec>() {
+        ann_plans.push(ann);
+    }
+    for child in plan.children() {
+        find_ann_plans(child.as_ref(), ann_plans);
+    }
+}
+
+fn collect_ann_plans(
+    plan: &Arc<dyn ExecutionPlan>,
+    ann_plans: &mut VecDeque<Arc<dyn ExecutionPlan>>,
+) {
+    if plan.downcast_ref::<ANNIvfSubIndexExec>().is_some() {
+        ann_plans.push_back(plan.clone());
+        return;
+    }
+    for child in plan.children() {
+        collect_ann_plans(child, ann_plans);
+    }
+}
+
+/// Replace normalized-L2 ANN nodes with equivalent nodes that use internal bounds, then
+/// convert their output to the public cosine scale before any generic plan node consumes it.
+fn normalize_ann_branches(
+    public_plan: Arc<dyn ExecutionPlan>,
+    internal_plan: Arc<dyn ExecutionPlan>,
+    normalized_l2_indices: &HashSet<String>,
+) -> Result<Arc<dyn ExecutionPlan>> {
+    let mut internal_ann_plans = VecDeque::new();
+    collect_ann_plans(&internal_plan, &mut internal_ann_plans);
+    let normalized =
+        replace_ann_branches(public_plan, &mut internal_ann_plans, normalized_l2_indices)?;
+    if !internal_ann_plans.is_empty() {
+        return Err(Error::Runtime {
+            message: "internal and public vector plans contained different ANN branches"
+                .to_string(),
+        });
+    }
+    Ok(normalized)
+}
+
+fn replace_ann_branches(
+    public_plan: Arc<dyn ExecutionPlan>,
+    internal_ann_plans: &mut VecDeque<Arc<dyn ExecutionPlan>>,
+    normalized_l2_indices: &HashSet<String>,
+) -> Result<Arc<dyn ExecutionPlan>> {
+    if let Some(public_ann) = public_plan.downcast_ref::<ANNIvfSubIndexExec>() {
+        let internal_plan = internal_ann_plans
+            .pop_front()
+            .ok_or_else(|| Error::Runtime {
+                message: "internal vector plan was missing an ANN branch".to_string(),
+            })?;
+        let internal_ann = internal_plan
+            .downcast_ref::<ANNIvfSubIndexExec>()
+            .expect("collected only ANN plans");
+        let same_indices = public_ann
+            .indices()
+            .iter()
+            .map(|index| &index.uuid)
+            .eq(internal_ann.indices().iter().map(|index| &index.uuid));
+        if public_ann.query().column != internal_ann.query().column
+            || public_ann.query().metric_type != internal_ann.query().metric_type
+            || !same_indices
+        {
+            return Err(Error::Runtime {
+                message: "internal and public vector plans had mismatched ANN branches".to_string(),
+            });
+        }
+
+        let normalized_count = public_ann
+            .indices()
+            .iter()
+            .filter(|index| normalized_l2_indices.contains(&index.uuid.to_string()))
+            .count();
+        if normalized_count == 0 {
+            return Ok(public_plan);
+        }
+        if normalized_count != public_ann.indices().len() {
+            return Err(Error::Runtime {
+                message: "one ANN branch mixed public and normalized-L2 distance scales"
+                    .to_string(),
+            });
+        }
+        return scale_distance_column(internal_plan, COSINE_ANN_SCALE);
+    }
+
+    let children = public_plan
+        .children()
+        .into_iter()
+        .cloned()
+        .map(|child| replace_ann_branches(child, internal_ann_plans, normalized_l2_indices))
+        .collect::<Result<Vec<_>>>()?;
+    Ok(with_new_children_if_necessary(public_plan, children)?)
+}
+
+fn scale_distance_column(
+    plan: Arc<dyn ExecutionPlan>,
+    scale: f32,
+) -> Result<Arc<dyn ExecutionPlan>> {
+    let schema = plan.schema();
+    if schema.column_with_name(DIST_COL).is_none() {
+        return Ok(plan);
+    }
+
+    let expressions: Vec<(Arc<dyn PhysicalExpr>, String)> = schema
+        .fields()
+        .iter()
+        .enumerate()
+        .map(|(index, field)| {
+            let column: Arc<dyn PhysicalExpr> = Arc::new(PhysicalColumn::new(field.name(), index));
+            let expression = if field.name() == DIST_COL {
+                let scale: Arc<dyn PhysicalExpr> =
+                    Arc::new(Literal::new(ScalarValue::Float32(Some(scale))));
+                Arc::new(BinaryExpr::new(column, Operator::Multiply, scale))
+                    as Arc<dyn PhysicalExpr>
+            } else {
+                column
+            };
+            (expression, field.name().clone())
+        })
+        .collect();
+    Ok(Arc::new(ProjectionExec::try_new(expressions, plan)?))
+}
+
 // Take many execution plans and map them into a single plan that adds
 // a query_index column and unions them.
 pub(crate) fn create_multi_vector_plan(
@@ -1455,6 +1664,206 @@ mod tests {
         );
     }
 
+    #[tokio::test]
+    async fn test_cosine_pq_distance_uses_public_cosine_scale() {
+        use arrow_array::{Int32Array, RecordBatch, types::Float32Type};
+        use arrow_schema::{DataType, Field, Schema};
+
+        use crate::connect;
+        use crate::index::{Index, vector::IvfPqIndexBuilder};
+
+        fn normalized_vector(state: &mut u64, dimension: usize) -> Vec<f32> {
+            let mut vector = (0..dimension)
+                .map(|_| {
+                    *state = state
+                        .wrapping_mul(6_364_136_223_846_793_005)
+                        .wrapping_add(1);
+                    ((*state >> 32) as u32 as f32 / u32::MAX as f32) * 2.0 - 1.0
+                })
+                .collect::<Vec<_>>();
+            let norm = vector.iter().map(|value| value * value).sum::<f32>().sqrt();
+            vector.iter_mut().for_each(|value| *value /= norm);
+            vector
+        }
+
+        fn distances(batches: &[RecordBatch]) -> Vec<f32> {
+            batches
+                .iter()
+                .flat_map(|batch| {
+                    batch[DIST_COL]
+                        .as_primitive::<Float32Type>()
+                        .values()
+                        .to_vec()
+                })
+                .collect()
+        }
+
+        let conn = connect("memory://").execute().await.unwrap();
+        let dimension = 8;
+        let num_rows = 256;
+        let mut state = 42;
+        let values = (0..num_rows)
+            .flat_map(|_| normalized_vector(&mut state, dimension))
+            .collect::<Vec<_>>();
+        let query_vector = normalized_vector(&mut state, dimension);
+        let vectors = Arc::new(fixed_size_list_array(values, dimension as i32));
+        let schema = Arc::new(Schema::new(vec![
+            Field::new("id", DataType::Int32, false),
+            Field::new("vector", vectors.data_type().clone(), false),
+        ]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..num_rows)), vectors],
+        )
+        .unwrap();
+        let table = conn
+            .create_table("test_cosine_pq_distance", batch)
+            .execute()
+            .await
+            .unwrap();
+        table
+            .create_index(
+                &["vector"],
+                Index::IvfPq(
+                    IvfPqIndexBuilder::default()
+                        .distance_type(crate::DistanceType::Cosine)
+                        .num_partitions(1)
+                        .num_sub_vectors(1),
+                ),
+            )
+            .execute()
+            .await
+            .unwrap();
+
+        let approximate = table
+            .vector_search(query_vector.as_slice())
+            .unwrap()
+            .limit(5)
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let refined = table
+            .vector_search(query_vector.as_slice())
+            .unwrap()
+            .limit(5)
+            .refine_factor(1)
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let approximate_distances = distances(&approximate);
+        let refined_distances = distances(&refined);
+        assert_eq!(approximate_distances.len(), refined_distances.len());
+        for (approximate, refined) in approximate_distances.iter().zip(&refined_distances) {
+            assert!(
+                (approximate - refined).abs() < 1e-5,
+                "approximate cosine distance {approximate} did not use the public scale; refined distance was {refined}"
+            );
+        }
+
+        // Distance range bounds are public cosine distances too. Lance applies them to
+        // internal ANN scores, so the planner must translate the bounds before execution.
+        let nearest = approximate_distances[0];
+        let ranged = table
+            .vector_search(query_vector.as_slice())
+            .unwrap()
+            .limit(1)
+            .distance_range(Some(nearest - 1e-5), Some(nearest + 1e-5))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let ranged_distances = distances(&ranged);
+        assert_eq!(ranged_distances.len(), 1);
+        assert!((ranged_distances[0] - nearest).abs() < 1e-5);
+
+        let refined_ranged = table
+            .vector_search(query_vector.as_slice())
+            .unwrap()
+            .limit(1)
+            .refine_factor(1)
+            .distance_range(None, Some(nearest + 1e-5))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        assert_eq!(
+            distances(&refined_ranged).len(),
+            1,
+            "refinement must not apply public cosine bounds to internal ANN scores"
+        );
+
+        let aliased = table
+            .vector_search(query_vector.as_slice())
+            .unwrap()
+            .limit(1)
+            .select(Select::dynamic(&[("aliased_distance", "_distance")]))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let batch = &aliased[0];
+        let aliased_distance = batch["aliased_distance"]
+            .as_primitive::<Float32Type>()
+            .value(0);
+        let public_distance = batch[DIST_COL].as_primitive::<Float32Type>().value(0);
+        assert!(
+            (aliased_distance - public_distance).abs() < 1e-5,
+            "distance aliases and auto-projected distances must use the same public scale"
+        );
+
+        // Appended rows take an exact fallback branch. Its public range filter must stay
+        // independent of the translated ANN bounds before both branches are merged.
+        let mut orthogonal = normalized_vector(&mut state, dimension);
+        let projection = orthogonal
+            .iter()
+            .zip(&query_vector)
+            .map(|(left, right)| left * right)
+            .sum::<f32>();
+        for (value, query_value) in orthogonal.iter_mut().zip(&query_vector) {
+            *value -= projection * query_value;
+        }
+        let norm = orthogonal
+            .iter()
+            .map(|value| value * value)
+            .sum::<f32>()
+            .sqrt();
+        orthogonal.iter_mut().for_each(|value| *value /= norm);
+        let appended_vectors = Arc::new(fixed_size_list_array(orthogonal, dimension as i32));
+        let appended = RecordBatch::try_new(
+            schema,
+            vec![Arc::new(Int32Array::from(vec![num_rows])), appended_vectors],
+        )
+        .unwrap();
+        table.add(appended).execute().await.unwrap();
+
+        let mixed = table
+            .vector_search(query_vector.as_slice())
+            .unwrap()
+            .limit(5)
+            .distance_range(None, Some(nearest + 1e-5))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let mixed_distances = distances(&mixed);
+        assert_eq!(mixed_distances.len(), 1);
+        assert!((mixed_distances[0] - nearest).abs() < 1e-5);
+    }
+
     #[tokio::test]
     async fn test_create_plan_applies_approx_mode_to_ann_query() {
         use arrow_array::RecordBatch;

rust/lancedb/src/table/query/lsm.rs

@@ -130,6 +130,10 @@ pub(super) async fn create_lsm_plan(
         .await?
     };
 
+    // Normalize cosine ANN arms before LSM merge and sort nodes compare their
+    // distances with exact SSTable and memtable arms.
+    let plan = super::normalize_cosine_ann_branches(plan).await?;
+
     // Lance appends the primary-key columns internally for dedup and keeps them in
     // the output; drop the ones the user did not request so the projection matches.
     restore_projection(plan, &query, &pk_columns)
@@ -300,7 +304,7 @@ async fn build_read_context(
     for shard_id in shard_ids {
         let manifest_store =
             ShardManifestStore::new(store.clone(), &base_path, shard_id, scan_batch_size);
-        if let Some(manifest) = manifest_store.read_latest().await? {
+        if let Some(manifest) = manifest_store.latest().await? {
             snapshots.push(snapshot_from_manifest(shard_id, &manifest, &exclude));
         }
     }

rust/lancedb/src/table/refresh.rs

@@ -7,16 +7,6 @@
 //! therefore idempotent and does not observe input mutation -- once a row is
 //! filled, changing what the expression reads leaves the stored result alone.
 //!
-//! A column's computed inputs are filled first -- the dependency graph is
-//! walked once, each reachable column filled once in dependency order, each
-//! fill its own commit. Every fill in the pass, the requested column's
-//! included, covers only the fragments of the snapshot the pass started
-//! from: a commit may rebase over a concurrent append, and the fragment that
-//! admits carries placeholder nulls no earlier fill covered, so it waits for
-//! a later refresh rather than being read as values. Two concurrent fills of
-//! one input collide on its field in lance's conflict check, so a dependent
-//! fill can only commit over inputs that were durable when it read them.
-//!
 //! Two passes per fragment. The first scans only the unfilled live rows and
 //! evaluates the expression over them, which yields the exact fill count and
 //! decides whether the fragment is staged at all -- a fragment where nothing
@@ -29,14 +19,10 @@
 //! inputs masked to null first, so a poison value in a row nobody is filling
 //! cannot fail the refresh.
 
-use std::collections::HashSet;
 use std::sync::Arc;
 
-use arrow_array::{
-    Array, ArrayRef, BooleanArray, LargeBinaryArray, RecordBatch, RecordBatchOptions, StructArray,
-    new_null_array,
-};
-use arrow_schema::{DataType, Field as ArrowField, Schema as ArrowSchema};
+use arrow_array::{ArrayRef, BooleanArray, RecordBatch, RecordBatchOptions};
+use arrow_schema::Schema as ArrowSchema;
 use datafusion_expr::ColumnarValue;
 use futures::{Stream, StreamExt, TryStreamExt};
 use lance::Dataset;
@@ -44,7 +30,7 @@ use lance::dataset::WriteDestination;
 use lance::dataset::fragment::FileFragment;
 use lance::dataset::transaction::Operation;
 use lance_core::ROW_ID;
-use lance_core::datatypes::{BlobHandling, Schema as LanceSchema};
+use lance_core::datatypes::Schema as LanceSchema;
 use serde::{Deserialize, Serialize};
 
 use super::computed_columns::{BoundExpression, ComputedColumnKind, computed_column_from_field};
@@ -55,8 +41,7 @@ use crate::{Error, Result};
 /// The result of refreshing a computed column.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
 pub struct RefreshColumnResult {
-    /// Rows that had a value computed, in the requested column only; inputs
-    /// filled on its behalf are not counted.
+    /// Rows that had a value computed.
     #[serde(default)]
     pub rows_filled: u64,
     /// The commit version associated with the operation.
@@ -67,7 +52,6 @@ pub struct RefreshColumnResult {
 struct RefreshExecution {
     result: RefreshColumnResult,
     source_version: u64,
-    published_version: Option<u64>,
 }
 
 /// Internal implementation of the refresh logic.
@@ -90,12 +74,7 @@ async fn execute_refresh_column_with_source(
 
     let expression = declared_expression(&dataset, column)?;
     let schema = Arc::new(ArrowSchema::from(dataset.schema()));
-    let bound = Arc::new(super::computed_columns::bind(
-        schema.clone(),
-        column,
-        &expression,
-    )?);
-    ensure_inputs_filled(&dataset, &schema, column, &bound).await?;
+    let bound = Arc::new(super::computed_columns::bind(schema, column, &expression)?);
     let field = dataset
         .schema()
         .field(column)
@@ -108,7 +87,6 @@ async fn execute_refresh_column_with_source(
         fields: vec![field.clone()],
         metadata: Default::default(),
     };
-    let output_is_blob = field.is_blob_v2();
 
     let mut rows_filled = 0u64;
     let mut replacements = Vec::new();
@@ -118,30 +96,29 @@ async fn execute_refresh_column_with_source(
             continue;
         }
         rows_filled += gained;
-        let values =
-            fill_stream(&dataset, &fragment, bound.clone(), column, output_is_blob).await?;
+        let values = fill_stream(&dataset, &fragment, bound.clone(), column).await?;
         replacements.push(fragment.write_columns(values, &column_schema).await?);
     }
 
-    let source_version = dataset.version().version;
     if replacements.is_empty() {
+        let source_version = dataset.version().version;
         return Ok(RefreshExecution {
             result: RefreshColumnResult {
                 rows_filled: 0,
                 version: source_version,
             },
             source_version,
-            published_version: None,
         });
     }
 
+    let read_version = dataset.version().version;
     // The dataset's own session, so registrations and caches survive the
     // commit being installed on the handle.
     let session = dataset.session();
     let new_dataset = Dataset::commit(
         WriteDestination::Dataset(dataset.clone()),
         Operation::DataReplacement { replacements },
-        Some(source_version),
+        Some(read_version),
         None,
         None,
         session,
@@ -156,52 +133,10 @@ async fn execute_refresh_column_with_source(
             rows_filled,
             version,
         },
-        source_version,
-        published_version: Some(version),
+        source_version: read_version,
     })
 }
 
-/// Refuse while a computed input still has rows a refresh of it would fill:
-/// read now, its placeholder null would be evaluated as a value and kept.
-async fn ensure_inputs_filled(
-    dataset: &Dataset,
-    schema: &Arc<ArrowSchema>,
-    column: &str,
-    bound: &BoundExpression,
-) -> Result<()> {
-    for input in &bound.roots {
-        let Some(declaration) = schema
-            .field_with_name(input)
-            .ok()
-            .and_then(computed_column_from_field)
-        else {
-            continue;
-        };
-        let ComputedColumnKind::Sql { expression } = &declaration.kind else {
-            return Err(Error::NotSupported {
-                message: format!(
-                    "computed column '{column}' reads '{input}', whose fill state this \
-                     refresh cannot check; refresh '{input}' first"
-                ),
-            });
-        };
-        let input_bound = super::computed_columns::bind(schema.clone(), input, expression)?;
-        let mut unfilled = 0u64;
-        for fragment in dataset.get_fragments() {
-            unfilled += count_fragment_gains(dataset, &fragment, &input_bound, input).await?;
-        }
-        if unfilled > 0 {
-            return Err(Error::InvalidInput {
-                message: format!(
-                    "computed column '{column}' reads '{input}', which has {unfilled} unfilled \
-                     rows; refresh '{input}' first"
-                ),
-            });
-        }
-    }
-    Ok(())
-}
-
 /// Run the refresh as a [`Job`] in this process.
 pub(crate) async fn execute_refresh_column_async(
     table: &NativeTable,
@@ -225,7 +160,8 @@ pub(crate) async fn execute_refresh_column_async(
             rows_failed: 0,
             rows_remaining: 0,
             source_version: execution.source_version,
-            published_version: execution.published_version,
+            published_version: (execution.result.rows_filled > 0)
+                .then_some(execution.result.version),
         })
     })))
 }
@@ -300,15 +236,12 @@ fn evaluation_batch(
     mask_out: Option<&BooleanArray>,
 ) -> lance_core::Result<RecordBatch> {
     let mut columns = Vec::with_capacity(bound.roots.len());
-    let mut fields = Vec::with_capacity(bound.roots.len());
     for name in &bound.roots {
-        let index = batch.schema_ref().index_of(name).map_err(|_| {
+        let column = batch.column_by_name(name).ok_or_else(|| {
             lance_core::Error::invalid_input(format!(
                 "refreshing a computed column read no {name} column"
             ))
         })?;
-        let column = batch.column(index);
-        fields.push(batch.schema_ref().field(index).clone());
         // Rows outside the mask must not reach the expression: a value in a
         // deleted or already-filled row can be one it would choke on.
         columns.push(match mask_out {
@@ -317,7 +250,7 @@ fn evaluation_batch(
         });
     }
     Ok(RecordBatch::try_new_with_options(
-        Arc::new(ArrowSchema::new(fields)),
+        bound.read_schema.clone(),
         columns,
         &RecordBatchOptions::new().with_row_count(Some(batch.num_rows())),
     )?)
@@ -338,99 +271,6 @@ fn evaluate(bound: &BoundExpression, batch: &RecordBatch) -> lance_core::Result<
     }
 }
 
-fn materialized_blob_ids(schema: &LanceSchema, paths: &[String]) -> Result<HashSet<u32>> {
-    paths
-        .iter()
-        .map(|path| {
-            let field = schema
-                .resolve(path)
-                .and_then(|fields| fields.last().copied())
-                .ok_or_else(|| Error::InvalidInput {
-                    message: format!("computed Blob input '{path}' no longer exists"),
-                })?;
-            if !field.is_blob_v2() {
-                return Err(Error::InvalidInput {
-                    message: format!("computed Blob input '{path}' is no longer Blob v2"),
-                });
-            }
-            u32::try_from(field.id).map_err(|_| Error::InvalidInput {
-                message: format!(
-                    "computed Blob input '{path}' has invalid field id {}",
-                    field.id
-                ),
-            })
-        })
-        .collect()
-}
-
-fn configure_blob_inputs(
-    scanner: &mut lance::dataset::scanner::Scanner,
-    schema: &LanceSchema,
-    bound: &BoundExpression,
-    extra_blob_id: Option<u32>,
-) -> Result<()> {
-    let mut ids = materialized_blob_ids(schema, &bound.blob_paths)?;
-    ids.extend(extra_blob_id);
-    scanner.blob_handling(BlobHandling::SomeBlobsBinary(ids));
-    Ok(())
-}
-
-fn blob_array_from_binary(
-    array: &ArrayRef,
-    target_field: &ArrowField,
-) -> lance_core::Result<ArrayRef> {
-    let values = array
-        .as_any()
-        .downcast_ref::<LargeBinaryArray>()
-        .ok_or_else(|| {
-            lance_core::Error::invalid_input(format!(
-                "a Blob v2 computed output produced {}, expected LargeBinary",
-                array.data_type()
-            ))
-        })?;
-    let mut builder = lance::blob::BlobArrayBuilder::new(values.len());
-    for index in 0..values.len() {
-        if values.is_null(index) {
-            builder.push_null()?;
-        } else {
-            builder.push_bytes(values.value(index))?;
-        }
-    }
-    let minimal = builder.finish()?;
-    let minimal = minimal
-        .as_any()
-        .downcast_ref::<StructArray>()
-        .ok_or_else(|| lance_core::Error::internal("Blob builder returned a non-struct array"))?;
-    let DataType::Struct(target_fields) = target_field.data_type() else {
-        return Err(lance_core::Error::invalid_input(format!(
-            "Blob v2 output field '{}' has non-struct type {}",
-            target_field.name(),
-            target_field.data_type()
-        )));
-    };
-    let columns = target_fields
-        .iter()
-        .map(|field| match field.name().as_str() {
-            "data" | "uri" => minimal
-                .column_by_name(field.name())
-                .cloned()
-                .ok_or_else(|| {
-                    lance_core::Error::internal(format!("Blob builder omitted '{}'", field.name()))
-                }),
-            "position" | "size" => Ok(new_null_array(field.data_type(), minimal.len())),
-            name => Err(lance_core::Error::invalid_input(format!(
-                "Blob v2 output field '{}' has unsupported logical child '{name}'",
-                target_field.name()
-            ))),
-        })
-        .collect::<lance_core::Result<Vec<_>>>()?;
-    Ok(Arc::new(StructArray::try_new(
-        target_fields.clone(),
-        columns,
-        minimal.nulls().cloned(),
-    )?))
-}
-
 /// How many rows of one fragment would gain a value.
 ///
 /// Scans only the unfilled live rows -- deleted rows never reach the
@@ -449,7 +289,6 @@ async fn count_fragment_gains(
         .with_row_id()
         .filter(&format!("{} IS NULL", quote_identifier(column)))?
         .project(&bound.roots)?;
-    configure_blob_inputs(&mut scanner, dataset.schema(), bound, None)?;
 
     let mut gained = 0u64;
     let mut batches = scanner.try_into_stream().await?;
@@ -471,7 +310,6 @@ async fn fill_stream(
     fragment: &FileFragment,
     bound: Arc<BoundExpression>,
     column: &str,
-    output_is_blob: bool,
 ) -> Result<impl Stream<Item = lance_core::Result<RecordBatch>> + Send + use<>> {
     let mut projection: Vec<String> = bound.roots.clone();
     projection.push(column.to_string());
@@ -481,20 +319,6 @@ async fn fill_stream(
         .with_row_id()
         .include_deleted_rows()
         .project(&projection)?;
-    let output_blob_id = output_is_blob
-        .then(|| {
-            dataset
-                .schema()
-                .field(column)
-                .and_then(|field| u32::try_from(field.id).ok())
-        })
-        .flatten();
-    configure_blob_inputs(
-        &mut scanner,
-        dataset.schema(),
-        bound.as_ref(),
-        output_blob_id,
-    )?;
 
     let projected = Arc::new(ArrowSchema::new(vec![
         ArrowSchema::from(dataset.schema())
@@ -530,11 +354,6 @@ async fn fill_stream(
 
         let computed = evaluate(&bound, &evaluation_batch(&batch, &bound, Some(&keep))?)?;
         let merged = arrow_select::zip::zip(&fill, &computed, existing)?;
-        let merged = if output_is_blob {
-            blob_array_from_binary(&merged, projected.field(0))?
-        } else {
-            merged
-        };
         Ok(RecordBatch::try_new(projected.clone(), vec![merged])?)
     }))
 }
@@ -543,12 +362,8 @@ async fn fill_stream(
 mod tests {
     use std::sync::Arc;
 
-    use arrow_array::{
-        Array, ArrayRef, Int32Array, LargeBinaryArray, RecordBatch, StructArray, record_batch,
-    };
-    use arrow_schema::Field as ArrowField;
+    use arrow_array::{Int32Array, record_batch};
     use futures::TryStreamExt;
-    use lance_core::ROW_ID;
 
     use crate::connect;
     use crate::query::{ExecutableQuery, QueryBase, Select};
@@ -569,8 +384,7 @@ mod tests {
             .version)
     }
 
-    async fn read(table: &Table, column: &str) -> Vec<Option<i64>> {
-        use arrow_array::{Array, Int64Array};
+    async fn read(table: &Table, column: &str) -> Vec<Option<i32>> {
         let batches = table
             .query()
             .select(Select::columns(&[column]))
@@ -580,19 +394,15 @@ mod tests {
             .try_collect::<Vec<_>>()
             .await
             .unwrap();
-        let mut values: Vec<Option<i64>> = batches
+        let mut values: Vec<Option<i32>> = batches
             .iter()
             .flat_map(|batch| {
-                let array = &batch[column];
-                match array.as_any().downcast_ref::<Int32Array>() {
-                    Some(ints) => ints.iter().map(|v| v.map(i64::from)).collect::<Vec<_>>(),
-                    None => array
-                        .as_any()
-                        .downcast_ref::<Int64Array>()
-                        .unwrap()
-                        .iter()
-                        .collect::<Vec<_>>(),
-                }
+                batch[column]
+                    .as_any()
+                    .downcast_ref::<Int32Array>()
+                    .unwrap()
+                    .iter()
+                    .collect::<Vec<_>>()
             })
             .collect();
         values.sort();
@@ -604,117 +414,6 @@ mod tests {
         table.add(batch).execute().await.unwrap();
     }
 
-    #[test]
-    fn test_blob_output_matches_complete_logical_field() {
-        let values: ArrayRef = Arc::new(LargeBinaryArray::from(vec![
-            Some(b"hello".as_slice()),
-            None,
-        ]));
-        let field = ArrowField::new(
-            "image",
-            lance_core::datatypes::BLOB_V2_LOGICAL_TYPE.clone(),
-            true,
-        );
-
-        let output = super::blob_array_from_binary(&values, &field).unwrap();
-        assert_eq!(output.data_type(), field.data_type());
-        let output = output.as_any().downcast_ref::<StructArray>().unwrap();
-        assert_eq!(output.column_by_name("position").unwrap().null_count(), 2);
-        assert_eq!(output.column_by_name("size").unwrap().null_count(), 2);
-    }
-
-    /// The gate's reproducer: `b = coalesce(a, 0)` refreshed before `a`
-    /// must not bake zeros from `a`'s placeholder null. It is refused, and
-    /// names the input, until `a` is filled -- after every append too.
-    #[tokio::test]
-    async fn test_dependent_refresh_refuses_an_unfilled_input() {
-        let table = table_with("dependent_refresh_order", vec![1, 2, 3]).await;
-        table
-            .add_columns()
-            .computed("a", "x + 1")
-            .computed("b", "coalesce(a, 0)")
-            .execute()
-            .await
-            .unwrap();
-
-        let err = table.refresh_column("b").await.unwrap_err();
-        assert!(
-            matches!(&err, Error::InvalidInput { message } if message.contains("refresh 'a' first")),
-            "{err}"
-        );
-        assert_eq!(read(&table, "b").await, vec![None, None, None]);
-
-        assert_eq!(table.refresh_column("a").await.unwrap().rows_filled, 3);
-        assert_eq!(table.refresh_column("b").await.unwrap().rows_filled, 3);
-        assert_eq!(read(&table, "b").await, vec![Some(2), Some(3), Some(4)]);
-
-        append(&table, vec![10]).await;
-        assert!(table.refresh_column("b").await.is_err());
-        table.refresh_column("a").await.unwrap();
-        assert_eq!(table.refresh_column("b").await.unwrap().rows_filled, 1);
-        assert_eq!(
-            table.count_rows(Some("b = 0".to_string())).await.unwrap(),
-            0
-        );
-    }
-
-    /// Names that need quoting, and a nested input, survive the trip through
-    /// declaration metadata and the dependency check: the recorded inputs
-    /// are matched by name, never re-parsed as SQL.
-    #[tokio::test]
-    async fn test_dependent_refresh_handles_awkward_column_names() {
-        use arrow_array::{Int32Array, StructArray};
-        use arrow_schema::{DataType, Field, Fields};
-
-        let conn = connect("memory://").execute().await.unwrap();
-        let age_fields = Fields::from(vec![Field::new("age", DataType::Int32, true)]);
-        let meta = StructArray::new(
-            age_fields.clone(),
-            vec![Arc::new(Int32Array::from(vec![10, 20])) as _],
-            None,
-        );
-        let schema = Arc::new(arrow_schema::Schema::new(vec![
-            Field::new("camelCase", DataType::Int32, true),
-            Field::new("with-hyphen", DataType::Int32, true),
-            Field::new("meta", DataType::Struct(age_fields), true),
-        ]));
-        let batch = arrow_array::RecordBatch::try_new(
-            schema,
-            vec![
-                Arc::new(Int32Array::from(vec![1, 2])) as _,
-                Arc::new(Int32Array::from(vec![100, 200])) as _,
-                Arc::new(meta) as _,
-            ],
-        )
-        .unwrap();
-        let table = conn
-            .create_table("awkward_names", batch)
-            .execute()
-            .await
-            .unwrap();
-
-        table
-            .add_columns()
-            .computed("y", "`camelCase` * 2")
-            .computed("z", "coalesce(y, 0) + `with-hyphen` + meta.age")
-            .execute()
-            .await
-            .unwrap();
-        let z = crate::table::computed_columns::computed_columns(
-            table.schema().await.unwrap().as_ref(),
-        )
-        .into_iter()
-        .find(|c| c.name == "z")
-        .unwrap();
-        assert_eq!(z.inputs, vec!["meta.age", "with-hyphen", "y"]);
-
-        let err = table.refresh_column("z").await.unwrap_err();
-        assert!(err.to_string().contains("refresh 'y' first"), "{err}");
-        assert_eq!(table.refresh_column("y").await.unwrap().rows_filled, 2);
-        assert_eq!(table.refresh_column("z").await.unwrap().rows_filled, 2);
-        assert_eq!(read(&table, "z").await, vec![Some(112), Some(224)]);
-    }
-
     #[tokio::test]
     async fn test_refresh_fills_a_declared_column() {
         let table = table_with("refresh_fills", vec![1, 2, 3]).await;
@@ -952,8 +651,7 @@ mod tests {
 
         let read_back = read(&table, "doubled").await;
         assert_eq!(read_back.len(), 20_000);
-        let mut expected: Vec<Option<i64>> =
-            values.iter().map(|v| Some(i64::from(v * 2))).collect();
+        let mut expected: Vec<Option<i32>> = values.iter().map(|v| Some(v * 2)).collect();
         expected.sort();
         assert_eq!(read_back, expected);
     }
@@ -1310,366 +1008,4 @@ mod tests {
         let err = table.refresh_column("embedding").await.unwrap_err();
         assert!(matches!(err, Error::NotSupported { message } if message.contains("udf")));
     }
-
-    fn blob_batch(ids: Vec<i32>, payloads: Vec<Option<&[u8]>>) -> RecordBatch {
-        use arrow_array::Int32Array;
-        use arrow_schema::{Field, Schema};
-
-        let mut builder = lance::blob::BlobArrayBuilder::new(payloads.len());
-        for payload in payloads {
-            match payload {
-                Some(payload) => builder.push_bytes(payload).unwrap(),
-                None => builder.push_null().unwrap(),
-            }
-        }
-        RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", arrow_schema::DataType::Int32, false),
-                crate::blob("image", true),
-            ])),
-            vec![Arc::new(Int32Array::from(ids)), builder.finish().unwrap()],
-        )
-        .unwrap()
-    }
-
-    async fn create_blob_table(path: &std::path::Path, batch: RecordBatch) -> Table {
-        let conn = connect(path.to_str().unwrap()).execute().await.unwrap();
-        conn.create_table("blobs", batch).execute().await.unwrap()
-    }
-
-    #[tokio::test]
-    async fn test_refresh_inherits_and_publishes_blob_output() {
-        use arrow_array::UInt64Array;
-        use lance_arrow::{
-            BLOB_DEDICATED_SIZE_THRESHOLD_META_KEY, BLOB_INLINE_SIZE_THRESHOLD_META_KEY,
-        };
-        use lance_core::datatypes::BlobKind;
-
-        use crate::table::schema_evolution::FieldMetadataUpdate;
-
-        let tmp = tempfile::tempdir().unwrap();
-        let table = create_blob_table(
-            tmp.path(),
-            blob_batch(
-                vec![1, 2, 3, 4],
-                vec![Some(b"hello"), Some(b"ab"), Some(b""), None],
-            ),
-        )
-        .await;
-        table
-            .add_columns()
-            .computed("image_copy", "image")
-            .execute()
-            .await
-            .unwrap();
-        table
-            .update_field_metadata(&[FieldMetadataUpdate::new("image_copy")
-                .set(BLOB_INLINE_SIZE_THRESHOLD_META_KEY, "1")
-                .set(BLOB_DEDICATED_SIZE_THRESHOLD_META_KEY, "4")])
-            .await
-            .unwrap();
-
-        let first_refresh = table.refresh_column("image_copy").await.unwrap();
-        assert_eq!(first_refresh.rows_filled, 3);
-        assert_eq!(
-            table.blob_columns().await.unwrap(),
-            vec!["image".to_string(), "image_copy".to_string()]
-        );
-
-        let batches = table
-            .query()
-            .with_row_id()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let batch = arrow_select::concat::concat_batches(&batches[0].schema(), &batches).unwrap();
-        assert!(
-            batch
-                .column_by_name("image_copy")
-                .unwrap()
-                .as_any()
-                .is::<arrow_array::StructArray>()
-        );
-        let row_ids = batch
-            .column_by_name(ROW_ID)
-            .unwrap()
-            .as_any()
-            .downcast_ref::<UInt64Array>()
-            .unwrap()
-            .values()
-            .to_vec();
-        let original = table.fetch_blobs("image", &row_ids).await.unwrap();
-        let copied = table.fetch_blobs("image_copy", &row_ids).await.unwrap();
-        assert_eq!(original, copied);
-        let ids = batch
-            .column_by_name("id")
-            .unwrap()
-            .as_any()
-            .downcast_ref::<Int32Array>()
-            .unwrap();
-        let files = table
-            .fetch_blob_files("image_copy", &row_ids)
-            .await
-            .unwrap();
-        let mut layouts = ids
-            .values()
-            .iter()
-            .copied()
-            .zip(files)
-            .map(|(id, file)| (id, file.and_then(|file| file.kind())))
-            .collect::<Vec<_>>();
-        layouts.sort_by_key(|(id, _)| *id);
-        assert_eq!(
-            layouts,
-            vec![
-                (1, Some(BlobKind::Dedicated)),
-                (2, Some(BlobKind::Packed)),
-                (3, Some(BlobKind::Inline)),
-                (4, None),
-            ]
-        );
-
-        table
-            .add(blob_batch(vec![5], vec![Some(b"appended")]))
-            .execute()
-            .await
-            .unwrap();
-        table
-            .optimize(crate::table::OptimizeAction::Compact {
-                options: crate::table::CompactionOptions::default(),
-                remap_options: None,
-            })
-            .await
-            .unwrap();
-        assert_eq!(
-            table
-                .refresh_column("image_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            1
-        );
-        assert_eq!(
-            table
-                .refresh_column("image_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            0
-        );
-
-        table.checkout(first_refresh.version).await.unwrap();
-        assert_eq!(table.count_rows(None).await.unwrap(), 4);
-        assert_eq!(
-            table.blob_columns().await.unwrap(),
-            vec!["image".to_string(), "image_copy".to_string()]
-        );
-        table.checkout_latest().await.unwrap();
-    }
-
-    #[tokio::test]
-    async fn test_refresh_inherits_nested_struct_blob_input() {
-        use arrow_array::{Int32Array, StructArray, UInt64Array};
-        use arrow_schema::{DataType, Field, Fields, Schema};
-
-        let tmp = tempfile::tempdir().unwrap();
-        let mut blob_builder = lance::blob::BlobArrayBuilder::new(2);
-        blob_builder.push_bytes(b"nested").unwrap();
-        blob_builder.push_null().unwrap();
-        let blob_field = crate::blob("image", true);
-        let metadata_fields = Fields::from(vec![blob_field.clone()]);
-        let metadata = StructArray::new(
-            metadata_fields.clone(),
-            vec![blob_builder.finish().unwrap()],
-            None,
-        );
-        let batch = RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", DataType::Int32, false),
-                Field::new("metadata", DataType::Struct(metadata_fields), true),
-            ])),
-            vec![Arc::new(Int32Array::from(vec![1, 2])), Arc::new(metadata)],
-        )
-        .unwrap();
-        let table = create_blob_table(tmp.path(), batch).await;
-        table
-            .add_columns()
-            .computed("payload_copy", "metadata.image")
-            .execute()
-            .await
-            .unwrap();
-
-        assert_eq!(
-            table
-                .refresh_column("payload_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            1
-        );
-        assert_eq!(
-            table.blob_columns().await.unwrap(),
-            vec!["metadata.image".to_string(), "payload_copy".to_string()]
-        );
-        let batches = table
-            .query()
-            .with_row_id()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let row_ids = batches[0]
-            .column_by_name(ROW_ID)
-            .unwrap()
-            .as_any()
-            .downcast_ref::<UInt64Array>()
-            .unwrap()
-            .values();
-        let payloads = table.fetch_blobs("payload_copy", row_ids).await.unwrap();
-        assert_eq!(payloads.value(0), b"nested");
-        assert!(payloads.is_null(1));
-    }
-
-    #[tokio::test]
-    async fn test_refresh_preserves_list_shape_when_materializing_blob_input() {
-        use arrow_array::{Int32Array, ListArray};
-        use arrow_buffer::{OffsetBuffer, ScalarBuffer};
-        use arrow_schema::{DataType, Field, Schema};
-
-        let tmp = tempfile::tempdir().unwrap();
-        let mut blob_builder = lance::blob::BlobArrayBuilder::new(3);
-        blob_builder.push_bytes(b"a").unwrap();
-        blob_builder.push_bytes(b"bb").unwrap();
-        blob_builder.push_null().unwrap();
-        let item = Arc::new(crate::blob("item", true));
-        let images = ListArray::new(
-            item.clone(),
-            OffsetBuffer::new(ScalarBuffer::from(vec![0, 2, 3])),
-            blob_builder.finish().unwrap(),
-            None,
-        );
-        let batch = RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", DataType::Int32, false),
-                Field::new("images", DataType::List(item), true),
-            ])),
-            vec![Arc::new(Int32Array::from(vec![1, 2])), Arc::new(images)],
-        )
-        .unwrap();
-        let table = create_blob_table(tmp.path(), batch).await;
-        table
-            .add_columns()
-            .computed("image_payloads", "images")
-            .execute()
-            .await
-            .unwrap();
-
-        assert_eq!(
-            table
-                .refresh_column("image_payloads")
-                .await
-                .unwrap()
-                .rows_filled,
-            2
-        );
-        let batches = table
-            .query()
-            .select(Select::columns(&["image_payloads"]))
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let output = batches[0]
-            .column_by_name("image_payloads")
-            .unwrap()
-            .as_any()
-            .downcast_ref::<ListArray>()
-            .unwrap();
-        assert_eq!(output.value_offsets(), &[0, 2, 3]);
-        assert!(output.values().as_any().is::<LargeBinaryArray>());
-    }
-
-    #[tokio::test]
-    async fn test_refresh_inherits_external_blob_input() {
-        use arrow_array::{Int32Array, StringArray, UInt64Array};
-        use arrow_schema::{DataType, Field, Schema};
-
-        let tmp = tempfile::tempdir().unwrap();
-        let payload = b"external-payload";
-        let path = tmp.path().join("payload.bin");
-        std::fs::write(&path, payload).unwrap();
-        let uri = url::Url::from_file_path(path).unwrap().to_string();
-        let conn = connect(tmp.path().join("db").to_str().unwrap())
-            .execute()
-            .await
-            .unwrap();
-        let table = conn
-            .create_empty_table(
-                "external",
-                Arc::new(Schema::new(vec![
-                    Field::new("id", DataType::Int32, false),
-                    crate::blob("image", true),
-                ])),
-            )
-            .execute()
-            .await
-            .unwrap();
-        let batch = RecordBatch::try_new(
-            Arc::new(Schema::new(vec![
-                Field::new("id", DataType::Int32, false),
-                Field::new("image", DataType::Utf8, true),
-            ])),
-            vec![
-                Arc::new(Int32Array::from(vec![1])),
-                Arc::new(StringArray::from(vec![Some(uri)])),
-            ],
-        )
-        .unwrap();
-        table
-            .add(batch)
-            .allow_external_blob_outside_bases(true)
-            .execute()
-            .await
-            .unwrap();
-        table
-            .add_columns()
-            .computed("payload_copy", "image")
-            .execute()
-            .await
-            .unwrap();
-
-        assert_eq!(
-            table
-                .refresh_column("payload_copy")
-                .await
-                .unwrap()
-                .rows_filled,
-            1
-        );
-        let batches = table
-            .query()
-            .with_row_id()
-            .execute()
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        let row_ids = batches[0]
-            .column_by_name(ROW_ID)
-            .unwrap()
-            .as_any()
-            .downcast_ref::<UInt64Array>()
-            .unwrap()
-            .values();
-        let payloads = table.fetch_blobs("payload_copy", row_ids).await.unwrap();
-        assert_eq!(payloads.value(0), payload);
-    }
 }

rust/lancedb/tests/fixtures/first_class_functions/v1/arrow_types.json

@@ -78,12 +78,6 @@
         "type": "utf8"
       }
     },
-    {
-      "arrow_type": "large_utf8",
-      "json": {
-        "type": "large_utf8"
-      }
-    },
     {
       "arrow_type": "binary",
       "json": {
@@ -177,21 +171,6 @@
         ]
       }
     },
-    {
-      "arrow_type": "list<large_utf8>",
-      "json": {
-        "type": "list",
-        "fields": [
-          {
-            "name": "item",
-            "nullable": false,
-            "type": {
-              "type": "large_utf8"
-            }
-          }
-        ]
-      }
-    },
     {
       "arrow_type": "large_list<utf8>",
       "json": {
@@ -207,21 +186,6 @@
         ]
       }
     },
-    {
-      "arrow_type": "large_list<large_utf8>",
-      "json": {
-        "type": "large_list",
-        "fields": [
-          {
-            "name": "item",
-            "nullable": false,
-            "type": {
-              "type": "large_utf8"
-            }
-          }
-        ]
-      }
-    },
     {
       "arrow_type": "fixed_size_list<float32, 384>",
       "json": {
@@ -366,4 +330,4 @@
     "timestamp[us]",
     "struct<a: int32>"
   ]
-}
+}