refactor(remote): drop x-lancedb-min-version; use timestamp baseline for read-your-write

manylinux2_17 reached EOL in 2024 and pyarrow stopped publishing 2_17
wheels long ago. We already build manylinux2_28 wheels, so drop the 2_17
matrix entries.

Fixes #3452

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.29.1-beta.0"
+current_version = "0.30.0-beta.1"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

.github/workflows/build_windows_wheel/action.yml

@@ -29,7 +29,3 @@ runs:
         args: ${{ inputs.args }}
         docker-options: "-e PIP_EXTRA_INDEX_URL='https://pypi.fury.io/lance-format/ https://pypi.fury.io/lancedb/'"
         working-directory: python
-    - uses: actions/upload-artifact@v4
-      with:
-        name: windows-wheels
-        path: python\target\wheels

.github/workflows/pypi-publish.yml

@@ -8,6 +8,9 @@ on:
     # This should trigger a dry run (we skip the final publish step)
     paths:
       - .github/workflows/pypi-publish.yml
+      - .github/workflows/build_linux_wheel/action.yml
+      - .github/workflows/build_mac_wheel/action.yml
+      - .github/workflows/build_windows_wheel/action.yml
       - Cargo.toml # Change in dependency frequently breaks builds
       - Cargo.lock
 
@@ -21,32 +24,21 @@ jobs:
   linux:
     name: Python ${{ matrix.config.platform }} manylinux${{ matrix.config.manylinux }}
     timeout-minutes: 60
-    permissions:
-      id-token: write
-      contents: read
     strategy:
       matrix:
         config:
-          - platform: x86_64
-            manylinux: "2_17"
-            extra_args: ""
-            runner: ubuntu-22.04
           - platform: x86_64
             manylinux: "2_28"
             extra_args: "--features fp16kernels"
             runner: ubuntu-22.04
-          - platform: aarch64
+          # For successful fat LTO builds, we need a large runner to avoid OOM errors.
-            manylinux: "2_17"
-            extra_args: ""
-            # For successful fat LTO builds, we need a large runner to avoid OOM errors.
-            runner: ubuntu-2404-8x-arm64
           - platform: aarch64
             manylinux: "2_28"
             extra_args: "--features fp16kernels"
             runner: ubuntu-2404-8x-arm64
     runs-on: ${{ matrix.config.runner }}
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
         with:
           fetch-depth: 0
           lfs: true
@@ -60,15 +52,14 @@ jobs:
           args: "--release --strip ${{ matrix.config.extra_args }}"
           arm-build: ${{ matrix.config.platform == 'aarch64' }}
           manylinux: ${{ matrix.config.manylinux }}
-      - uses: ./.github/workflows/upload_wheel
+      - uses: actions/upload-artifact@v7
         if: startsWith(github.ref, 'refs/tags/python-v')
         with:
-          fury_token: ${{ secrets.FURY_TOKEN }}
+          name: wheels-linux-${{ matrix.config.platform }}-${{ matrix.config.manylinux }}
+          path: target/wheels/lancedb-*.whl
+          if-no-files-found: error
   mac:
     timeout-minutes: 90
-    permissions:
-      id-token: write
-      contents: read
     runs-on: ${{ matrix.config.runner }}
     strategy:
       matrix:
@@ -78,7 +69,7 @@ jobs:
     env:
       MACOSX_DEPLOYMENT_TARGET: 10.15
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
         with:
           fetch-depth: 0
           lfs: true
@@ -90,18 +81,21 @@ jobs:
         with:
           python-minor-version: 10
           args: "--release --strip --target ${{ matrix.config.target }} --features fp16kernels"
-      - uses: ./.github/workflows/upload_wheel
+      - uses: actions/upload-artifact@v7
         if: startsWith(github.ref, 'refs/tags/python-v')
         with:
-          fury_token: ${{ secrets.FURY_TOKEN }}
+          name: wheels-mac-${{ matrix.config.target }}
+          path: target/wheels/lancedb-*.whl
+          if-no-files-found: error
   windows:
-    timeout-minutes: 60
+    timeout-minutes: 90
-    permissions:
-      id-token: write
-      contents: read
     runs-on: windows-latest
+    env:
+      # link.exe is single-threaded and the long pole on Windows builds. Use
+      # rustc's bundled lld-link instead.
+      CARGO_TARGET_X86_64_PC_WINDOWS_MSVC_LINKER: rust-lld
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
         with:
           fetch-depth: 0
           lfs: true
@@ -113,18 +107,70 @@ jobs:
         with:
           python-minor-version: 10
           args: "--release --strip"
-          vcpkg_token: ${{ secrets.VCPKG_GITHUB_PACKAGES }}
+      - uses: actions/upload-artifact@v7
-      - uses: ./.github/workflows/upload_wheel
         if: startsWith(github.ref, 'refs/tags/python-v')
         with:
-          fury_token: ${{ secrets.FURY_TOKEN }}
+          name: wheels-windows
+          path: target/wheels/lancedb-*.whl
+          if-no-files-found: error
+  publish:
+    name: Publish wheels
+    if: startsWith(github.ref, 'refs/tags/python-v')
+    needs: [linux, mac, windows]
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write
+      contents: read
+    steps:
+      - uses: actions/checkout@v6
+      - name: Download wheel artifacts
+        uses: actions/download-artifact@v8
+        with:
+          pattern: wheels-*
+          path: target/wheels
+          merge-multiple: true
+      - name: List wheels
+        run: ls -la target/wheels
+      - name: Choose repo
+        id: choose_repo
+        run: |
+          if [[ ${{ github.ref }} == *beta* ]]; then
+            echo "repo=fury" >> $GITHUB_OUTPUT
+          else
+            echo "repo=pypi" >> $GITHUB_OUTPUT
+          fi
+      - name: Publish to Fury
+        if: steps.choose_repo.outputs.repo == 'fury'
+        env:
+          FURY_TOKEN: ${{ secrets.FURY_TOKEN }}
+        run: |
+          shopt -s nullglob
+          WHEELS=(target/wheels/lancedb-*.whl)
+          if [[ ${#WHEELS[@]} -eq 0 ]]; then
+            echo "No wheels found in target/wheels/" >&2
+            exit 1
+          fi
+          for WHEEL in "${WHEELS[@]}"; do
+            echo "Uploading $WHEEL to Fury"
+            curl -f -F package=@"$WHEEL" "https://$FURY_TOKEN@push.fury.io/lancedb/"
+          done
+      # NOTE: pypa/gh-action-pypi-publish must be invoked directly from a
+      # workflow file, not from inside a composite action. When called from a
+      # composite, `github.action_repository` is empty (actions/runner#2473)
+      # and the action falls back to `github.repository`, producing a bogus
+      # `docker://ghcr.io/<repo>:<ref>` image reference that GHA tries to pull.
+      - name: Publish to PyPI
+        if: steps.choose_repo.outputs.repo == 'pypi'
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          packages-dir: target/wheels/
   gh-release:
     if: startsWith(github.ref, 'refs/tags/python-v')
     runs-on: ubuntu-latest
     permissions:
       contents: write
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
         with:
           fetch-depth: 0
           lfs: true
@@ -187,13 +233,13 @@ jobs:
   report-failure:
     name: Report Workflow Failure
     runs-on: ubuntu-latest
-    needs: [linux, mac, windows]
+    needs: [linux, mac, windows, publish]
     permissions:
       contents: read
       issues: write
     if: always() && failure() && startsWith(github.ref, 'refs/tags/python-v')
     steps:
-      - uses: actions/checkout@v4
+      - uses: actions/checkout@v6
       - uses: ./.github/actions/create-failure-issue
         with:
           job-results: ${{ toJSON(needs) }}

.github/workflows/upload_wheel/action.yml

@@ -1,34 +0,0 @@
-name: upload-wheel
-
-description: "Upload wheels to Pypi"
-inputs:
-  fury_token:
-    required: true
-    description: "release token for the fury repo"
-
-runs:
-  using: "composite"
-  steps:
-  - name: Choose repo
-    shell: bash
-    id: choose_repo
-    run: |
-      if [[ ${{ github.ref }} == *beta* ]]; then
-        echo "repo=fury" >> $GITHUB_OUTPUT
-      else
-        echo "repo=pypi" >> $GITHUB_OUTPUT
-      fi
-  - name: Publish to Fury
-    if: steps.choose_repo.outputs.repo == 'fury'
-    shell: bash
-    env:
-      FURY_TOKEN: ${{ inputs.fury_token }}
-    run: |
-      WHEEL=$(ls target/wheels/lancedb-*.whl 2> /dev/null | head -n 1)
-      echo "Uploading $WHEEL to Fury"
-      curl -f -F package=@$WHEEL https://$FURY_TOKEN@push.fury.io/lancedb/
-  - name: Publish to PyPI
-    if: steps.choose_repo.outputs.repo == 'pypi'
-    uses: pypa/gh-action-pypi-publish@release/v1
-    with:
-      packages-dir: target/wheels/

AGENTS.md

@@ -37,10 +37,13 @@ Before committing changes, run formatting for every language you touched. At min
   and run targeted tests through `cd python && uv run ...`.
 * TypeScript changes: run the relevant `npm`/`pnpm` lint, format, build, and docs commands in `nodejs`.
 
-Before creating a PR, make sure the PR title follows Conventional Commits, such as
+Before creating a PR, the exact value passed to `gh pr create --title` must follow
-`fix: support nested field paths in native index creation` or
+Conventional Commits, such as `fix: support nested field paths in native index creation`
-`feat(python): add dataset multiprocessing support`. The semantic-release check uses the
+or `feat(python): add dataset multiprocessing support`. Do not use a plain natural
-PR title and body as the merge commit message, so a non-conventional PR title will fail CI.
+language summary like `Support nested field paths in native index creation` as the PR
+title. The semantic-release check uses the PR title and body as the merge commit message,
+so a non-conventional PR title will fail CI. After creating a PR, read the remote PR title
+back and fix it immediately if it is not conventional.
 
 ## Coding tips
 

Cargo.lock

@@ -3284,8 +3284,8 @@ checksum = "42703706b716c37f96a77aea830392ad231f44c9e9a67872fa5548707e11b11c"
 
 [[package]]
 name = "fsst"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-array",
  "rand 0.9.4",
@@ -4506,8 +4506,8 @@ checksum = "e037a2e1d8d5fdbd49b16a4ea09d5d6401c1f29eca5ff29d03d3824dba16256a"
 
 [[package]]
 name = "lance"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arc-swap",
  "arrow",
@@ -4525,6 +4525,7 @@ dependencies = [
  "async_cell",
  "aws-credential-types",
  "aws-sdk-dynamodb",
+ "bitpacking",
  "byteorder",
  "bytes",
  "chrono",
@@ -4551,9 +4552,11 @@ dependencies = [
  "lance-io",
  "lance-linalg",
  "lance-namespace",
+ "lance-select",
  "lance-table",
  "lance-tokenizer",
  "log",
+ "moka",
  "object_store",
  "permutation",
  "pin-project",
@@ -4577,8 +4580,8 @@ dependencies = [
 
 [[package]]
 name = "lance-arrow"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4598,8 +4601,8 @@ dependencies = [
 
 [[package]]
 name = "lance-bitpacking"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrayref",
  "paste",
@@ -4608,8 +4611,8 @@ dependencies = [
 
 [[package]]
 name = "lance-core"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4644,8 +4647,8 @@ dependencies = [
 
 [[package]]
 name = "lance-datafusion"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4675,8 +4678,8 @@ dependencies = [
 
 [[package]]
 name = "lance-datagen"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4694,8 +4697,8 @@ dependencies = [
 
 [[package]]
 name = "lance-encoding"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -4730,8 +4733,8 @@ dependencies = [
 
 [[package]]
 name = "lance-file"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-arith",
  "arrow-array",
@@ -4762,8 +4765,8 @@ dependencies = [
 
 [[package]]
 name = "lance-index"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arc-swap",
  "arrow",
@@ -4800,6 +4803,7 @@ dependencies = [
  "lance-file",
  "lance-io",
  "lance-linalg",
+ "lance-select",
  "lance-table",
  "lance-tokenizer",
  "libm",
@@ -4827,8 +4831,8 @@ dependencies = [
 
 [[package]]
 name = "lance-io"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow",
  "arrow-arith",
@@ -4870,8 +4874,8 @@ dependencies = [
 
 [[package]]
 name = "lance-linalg"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -4887,8 +4891,8 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow",
  "async-trait",
@@ -4900,8 +4904,8 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace-impls"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow",
  "arrow-ipc",
@@ -4936,9 +4940,9 @@ dependencies = [
 
 [[package]]
 name = "lance-namespace-reqwest-client"
-version = "0.7.6"
+version = "0.7.7"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f65e31bdaa13e01dab6e7cf566da31df243c34a542f0d915d3601ec0e01e61d2"
+checksum = "6369eee4682fb11edf538388b43c61ce288b8302fe89bb40944d7daa7faaae99"
 dependencies = [
  "reqwest 0.12.28",
  "serde",
@@ -4948,10 +4952,25 @@ dependencies = [
  "url",
 ]
 
+[[package]]
+name = "lance-select"
+version = "7.1.0-beta.4"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
+dependencies = [
+ "arrow-array",
+ "arrow-buffer",
+ "byteorder",
+ "bytes",
+ "deepsize",
+ "itertools 0.13.0",
+ "lance-core",
+ "roaring",
+]
+
 [[package]]
 name = "lance-table"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow",
  "arrow-array",
@@ -4970,6 +4989,7 @@ dependencies = [
  "lance-core",
  "lance-file",
  "lance-io",
+ "lance-select",
  "log",
  "object_store",
  "prost",
@@ -4990,8 +5010,8 @@ dependencies = [
 
 [[package]]
 name = "lance-testing"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "arrow-array",
  "arrow-schema",
@@ -5002,8 +5022,8 @@ dependencies = [
 
 [[package]]
 name = "lance-tokenizer"
-version = "7.0.0-beta.13"
+version = "7.1.0-beta.4"
-source = "git+https://github.com/lance-format/lance.git?tag=v7.0.0-beta.13#929166e3ff51ed61b1fa42de2c63feaf51967ea1"
+source = "git+https://github.com/lance-format/lance.git?tag=v7.1.0-beta.4#0c0b3e18c0a4c75bda1dd6ec9d6247ef75bd29d9"
 dependencies = [
  "jieba-rs",
  "lindera",
@@ -5014,7 +5034,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb"
-version = "0.29.1-beta.0"
+version = "0.30.0-beta.1"
 dependencies = [
  "ahash",
  "anyhow",
@@ -5084,6 +5104,7 @@ dependencies = [
  "serde",
  "serde_json",
  "serde_with",
+ "serial_test",
  "snafu 0.8.9",
  "tempfile",
  "test-log",
@@ -5096,7 +5117,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-nodejs"
-version = "0.29.1-beta.0"
+version = "0.30.0-beta.1"
 dependencies = [
  "arrow-array",
  "arrow-buffer",
@@ -5119,7 +5140,7 @@ dependencies = [
 
 [[package]]
 name = "lancedb-python"
-version = "0.32.1-beta.0"
+version = "0.33.0-beta.1"
 dependencies = [
  "arrow",
  "async-trait",
@@ -8128,6 +8149,15 @@ dependencies = [
  "winapi-util",
 ]
 
+[[package]]
+name = "scc"
+version = "2.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "46e6f046b7fef48e2660c57ed794263155d713de679057f2d0c169bfc6e756cc"
+dependencies = [
+ "sdd",
+]
+
 [[package]]
 name = "schannel"
 version = "0.1.29"
@@ -8194,6 +8224,12 @@ dependencies = [
  "untrusted 0.9.0",
 ]
 
+[[package]]
+name = "sdd"
+version = "3.0.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "490dcfcbfef26be6800d11870ff2df8774fa6e86d047e3e8c8a76b25655e41ca"
+
 [[package]]
 name = "sec1"
 version = "0.3.0"
@@ -8384,6 +8420,32 @@ dependencies = [
  "unsafe-libyaml",
 ]
 
+[[package]]
+name = "serial_test"
+version = "3.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "911bd979bf1070a3f3aa7b691a3b3e9968f339ceeec89e08c280a8a22207a32f"
+dependencies = [
+ "futures-executor",
+ "futures-util",
+ "log",
+ "once_cell",
+ "parking_lot",
+ "scc",
+ "serial_test_derive",
+]
+
+[[package]]
+name = "serial_test_derive"
+version = "3.4.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "0a7d91949b85b0d2fb687445e448b40d322b6b3e4af6b44a29b21d9a5f33e6d9"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.117",
+]
+
 [[package]]
 name = "sha1"
 version = "0.10.6"

Cargo.toml

@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"
 
 [workspace.dependencies]
-lance = { "version" = "=7.0.0-beta.13", default-features = false, "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=7.1.0-beta.4", default-features = false, "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=7.0.0-beta.13", default-features = false, "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=7.1.0-beta.4", default-features = false, "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=7.0.0-beta.13", default-features = false, "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=7.1.0-beta.4", default-features = false, "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=7.0.0-beta.13", "tag" = "v7.0.0-beta.13", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=7.1.0-beta.4", "tag" = "v7.1.0-beta.4", "git" = "https://github.com/lance-format/lance.git" }
 ahash = "0.8"
 # Note that this one does not include pyarrow
 arrow = { version = "58.0.0", optional = false }

ci/check_lance_release.py

@@ -112,25 +112,25 @@ def fetch_remote_tags() -> List[TagInfo]:
             "api",
             "-X",
             "GET",
-            f"repos/{LANCE_REPO}/git/refs/tags",
+            f"repos/{LANCE_REPO}/releases",
-            "--paginate",
             "--jq",
-            ".[].ref",
+            ".[].tag_name",
+            "-F",
+            "per_page=20",
         ]
     )
     tags: List[TagInfo] = []
     for line in output.splitlines():
-        ref = line.strip()
+        tag = line.strip()
-        if not ref.startswith("refs/tags/v"):
+        if not tag.startswith("v"):
             continue
-        tag = ref.split("refs/tags/")[-1]
         version = tag.lstrip("v")
         try:
             tags.append(TagInfo(tag=tag, version=version, semver=parse_semver(version)))
         except ValueError:
             continue
     if not tags:
-        raise RuntimeError("No Lance tags could be parsed from GitHub API output")
+        raise RuntimeError("No Lance releases could be parsed from GitHub API output")
     return tags
 
 

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>0.29.1-beta.0</version>
+    <version>0.30.0-beta.1</version>
 </dependency>
 ```
 

docs/src/js/interfaces/ConnectionOptions.md

@@ -70,16 +70,20 @@ client used by manifest-enabled native connections.
 optional readConsistencyInterval: number;
 ```
 
-(For LanceDB OSS only): The interval, in seconds, at which to check for
+The interval, in seconds, at which to check for updates to the table
-updates to the table from other processes. If None, then consistency is not
+from other processes. If None, then consistency is not checked. For
-checked. For performance reasons, this is the default. For strong
+performance reasons, this is the default. For strong consistency, set
-consistency, set this to zero seconds. Then every read will check for
+this to zero seconds. Then every read will check for updates from other
-updates from other processes. As a compromise, you can set this to a
+processes. As a compromise, you can set this to a non-zero value for
-non-zero value for eventual consistency. If more than that interval
+eventual consistency. If more than that interval has passed since the
-has passed since the last check, then the table will be checked for updates.
+last check, then the table will be checked for updates. Note: this
-Note: this consistency only applies to read operations. Write operations are
+consistency only applies to read operations. Write operations are
 always consistent.
 
+Stronger consistency is not free. The smaller the interval, the more
+often each read pays the cost of checking for updates against object
+storage, raising per-read latency and cost.
+
 ***
 
 ### region?

java/lancedb-core/pom.xml

@@ -8,7 +8,7 @@
     <parent>
       <groupId>com.lancedb</groupId>
       <artifactId>lancedb-parent</artifactId>
-      <version>0.29.1-beta.0</version>
+      <version>0.30.0-beta.1</version>
       <relativePath>../pom.xml</relativePath>
     </parent>
 

java/pom.xml

@@ -6,7 +6,7 @@
 
     <groupId>com.lancedb</groupId>
     <artifactId>lancedb-parent</artifactId>
-    <version>0.29.1-beta.0</version>
+    <version>0.30.0-beta.1</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>7.0.0-beta.13</lance-core.version>
+        <lance-core.version>7.1.0-beta.4</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 = "0.29.1-beta.0"
+version = "0.30.0-beta.1"
 publish = false
 license.workspace = true
 description.workspace = true

nodejs/__test__/connection.test.ts

@@ -171,18 +171,22 @@ describe("given a connection", () => {
 
     let manifestDir =
       tmpDir.name + "/test_manifest_paths_v2_empty.lance/_versions";
-    readdirSync(manifestDir).forEach((file) => {
+    readdirSync(manifestDir)
-      expect(file).toMatch(/^\d{20}\.manifest$/);
+      .filter((f) => f.endsWith(".manifest"))
-    });
+      .forEach((file) => {
+        expect(file).toMatch(/^\d{20}\.manifest$/);
+      });
 
     table = (await db.createTable("test_manifest_paths_v2", [{ id: 1 }], {
       enableV2ManifestPaths: true,
     })) as LocalTable;
     expect(await table.usesV2ManifestPaths()).toBe(true);
     manifestDir = tmpDir.name + "/test_manifest_paths_v2.lance/_versions";
-    readdirSync(manifestDir).forEach((file) => {
+    readdirSync(manifestDir)
-      expect(file).toMatch(/^\d{20}\.manifest$/);
+      .filter((f) => f.endsWith(".manifest"))
-    });
+      .forEach((file) => {
+        expect(file).toMatch(/^\d{20}\.manifest$/);
+      });
   });
 
   it("should be able to migrate tables to the V2 manifest paths", async () => {
@@ -199,16 +203,20 @@ describe("given a connection", () => {
 
     const manifestDir =
       tmpDir.name + "/test_manifest_path_migration.lance/_versions";
-    readdirSync(manifestDir).forEach((file) => {
+    readdirSync(manifestDir)
-      expect(file).toMatch(/^\d\.manifest$/);
+      .filter((f) => f.endsWith(".manifest"))
-    });
+      .forEach((file) => {
+        expect(file).toMatch(/^\d\.manifest$/);
+      });
 
     await table.migrateManifestPathsV2();
     expect(await table.usesV2ManifestPaths()).toBe(true);
 
-    readdirSync(manifestDir).forEach((file) => {
+    readdirSync(manifestDir)
-      expect(file).toMatch(/^\d{20}\.manifest$/);
+      .filter((f) => f.endsWith(".manifest"))
-    });
+      .forEach((file) => {
+        expect(file).toMatch(/^\d{20}\.manifest$/);
+      });
   });
 });
 

nodejs/npm/darwin-arm64/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-darwin-arm64",
-	"version": "0.29.1-beta.0",
+	"version": "0.30.0-beta.1",
 	"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": "0.29.1-beta.0",
+	"version": "0.30.0-beta.1",
 	"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": "0.29.1-beta.0",
+	"version": "0.30.0-beta.1",
 	"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": "0.29.1-beta.0",
+	"version": "0.30.0-beta.1",
 	"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": "0.29.1-beta.0",
+	"version": "0.30.0-beta.1",
 	"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": "0.29.1-beta.0",
+  "version": "0.30.0-beta.1",
   "os": [
     "win32"
   ],

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

@@ -1,6 +1,6 @@
 {
 	"name": "@lancedb/lancedb-win32-x64-msvc",
-	"version": "0.29.1-beta.0",
+	"version": "0.30.0-beta.1",
 	"os": ["win32"],
 	"cpu": ["x64"],
 	"main": "lancedb.win32-x64-msvc.node",

nodejs/package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@lancedb/lancedb",
-  "version": "0.29.1-beta.0",
+  "version": "0.30.0-beta.1",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "@lancedb/lancedb",
-      "version": "0.29.1-beta.0",
+      "version": "0.30.0-beta.1",
       "cpu": [
         "x64",
         "arm64"

nodejs/package.json

@@ -11,7 +11,7 @@
     "ann"
   ],
   "private": false,
-  "version": "0.29.1-beta.0",
+  "version": "0.30.0-beta.1",
   "main": "dist/index.js",
   "exports": {
     ".": "./dist/index.js",

nodejs/src/lib.rs

@@ -24,15 +24,19 @@ mod util;
 #[napi(object)]
 #[derive(Debug)]
 pub struct ConnectionOptions {
-    /// (For LanceDB OSS only): The interval, in seconds, at which to check for
+    /// The interval, in seconds, at which to check for updates to the table
-    /// updates to the table from other processes. If None, then consistency is not
+    /// from other processes. If None, then consistency is not checked. For
-    /// checked. For performance reasons, this is the default. For strong
+    /// performance reasons, this is the default. For strong consistency, set
-    /// consistency, set this to zero seconds. Then every read will check for
+    /// this to zero seconds. Then every read will check for updates from other
-    /// updates from other processes. As a compromise, you can set this to a
+    /// processes. As a compromise, you can set this to a non-zero value for
-    /// non-zero value for eventual consistency. If more than that interval
+    /// eventual consistency. If more than that interval has passed since the
-    /// has passed since the last check, then the table will be checked for updates.
+    /// last check, then the table will be checked for updates. Note: this
-    /// Note: this consistency only applies to read operations. Write operations are
+    /// consistency only applies to read operations. Write operations are
     /// always consistent.
+    ///
+    /// Stronger consistency is not free. The smaller the interval, the more
+    /// often each read pays the cost of checking for updates against object
+    /// storage, raising per-read latency and cost.
     pub read_consistency_interval: Option<f64>,
     /// (For LanceDB OSS only): configuration for object storage.
     ///

python/.bumpversion.toml

@@ -1,5 +1,5 @@
 [tool.bumpversion]
-current_version = "0.33.0-beta.0"
+current_version = "0.33.0-beta.1"
 parse = """(?x)
     (?P<major>0|[1-9]\\d*)\\.
     (?P<minor>0|[1-9]\\d*)\\.

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.33.0-beta.0"
+version = "0.33.0-beta.1"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"

python/python/lancedb/__init__.py

@@ -94,7 +94,6 @@ def connect(
     host_override: str, optional
         The override url for LanceDB Cloud.
     read_consistency_interval: timedelta, default None
-        (For LanceDB OSS only)
         The interval at which to check for updates to the table from other
         processes. If None, then consistency is not checked. For performance
         reasons, this is the default. For strong consistency, set this to
@@ -104,6 +103,10 @@ def connect(
         the last check, then the table will be checked for updates. Note: this
         consistency only applies to read operations. Write operations are
         always consistent.
+
+        Stronger consistency is not free. The smaller the interval, the more
+        often each read pays the cost of checking for updates against object
+        storage, raising per-read latency and cost.
     client_config: ClientConfig or dict, optional
         Configuration options for the LanceDB Cloud HTTP client. If a dict, then
         the keys are the attributes of the ClientConfig class. If None, then the
@@ -147,6 +150,13 @@ def connect(
     >>> db = lancedb.connect("s3://my-bucket/lancedb",
     ...                      storage_options={"aws_access_key_id": "***"})
 
+    For tests and temporary data, use an in-memory database:
+
+    >>> db = lancedb.connect("memory://")
+
+    In-memory databases are not persisted. Tables are dropped when the last
+    connection or table handle referencing them is closed.
+
     Connect to LanceDB cloud:
 
     >>> db = lancedb.connect("db://my_database", api_key="ldb_...",
@@ -210,6 +220,7 @@ def connect(
             request_thread_pool=request_thread_pool,
             client_config=client_config,
             storage_options=storage_options,
+            read_consistency_interval=read_consistency_interval,
             **kwargs,
         )
     _check_s3_bucket_with_dots(str(uri), storage_options)
@@ -336,7 +347,6 @@ async def connect_async(
     host_override: str, optional
         The override url for LanceDB Cloud.
     read_consistency_interval: timedelta, default None
-        (For LanceDB OSS only)
         The interval at which to check for updates to the table from other
         processes. If None, then consistency is not checked. For performance
         reasons, this is the default. For strong consistency, set this to
@@ -346,6 +356,10 @@ async def connect_async(
         the last check, then the table will be checked for updates. Note: this
         consistency only applies to read operations. Write operations are
         always consistent.
+
+        Stronger consistency is not free. The smaller the interval, the more
+        often each read pays the cost of checking for updates against object
+        storage, raising per-read latency and cost.
     client_config: ClientConfig or dict, optional
         Configuration options for the LanceDB Cloud HTTP client. If a dict, then
         the keys are the attributes of the ClientConfig class. If None, then the
@@ -378,6 +392,8 @@ async def connect_async(
     ...     db = await lancedb.connect_async("s3://my-bucket/lancedb",
     ...                                      storage_options={
     ...                                          "aws_access_key_id": "***"})
+    ...     # For tests and temporary data, use an in-memory database
+    ...     db = await lancedb.connect_async("memory://")
     ...     # Connect to LanceDB cloud
     ...     db = await lancedb.connect_async("db://my_database", api_key="ldb_...",
     ...                                      client_config={

python/python/lancedb/remote/db.py

@@ -50,6 +50,7 @@ class RemoteDBConnection(DBConnection):
         connection_timeout: Optional[float] = None,
         read_timeout: Optional[float] = None,
         storage_options: Optional[Dict[str, str]] = None,
+        read_consistency_interval: Optional[timedelta] = None,
     ):
         """Connect to a remote LanceDB database."""
         if isinstance(client_config, dict):
@@ -103,6 +104,7 @@ class RemoteDBConnection(DBConnection):
                 host_override=host_override,
                 client_config=client_config,
                 storage_options=storage_options,
+                read_consistency_interval=read_consistency_interval,
             )
         )
 

python/python/lancedb/rerankers/linear_combination.py

@@ -102,8 +102,15 @@ class LinearCombinationReranker(Reranker):
 
         combined_list = []
         for row_id, result in results.items():
+            # Convert vector distance to a relevance score in [0, 1] where
+            # higher is better.  Missing vector entries are penalised with
+            # `_invert_score(fill)` = 1 - fill (= 0.0 for the default fill=1).
             vector_score = self._invert_score(result.get("_distance", fill))
-            fts_score = result.get("_score", fill)
+            # FTS scores (BM25) are already in a "higher = more relevant" space.
+            # Missing FTS entries are penalised symmetrically: we use
+            # `1 - fill` so that the same `fill` value drives both missing-vector
+            # and missing-FTS penalties in the same direction.
+            fts_score = result.get("_score", 1 - fill)
             result["_relevance_score"] = self._combine_score(vector_score, fts_score)
             combined_list.append(result)
 
@@ -123,8 +130,12 @@ class LinearCombinationReranker(Reranker):
         return tbl
 
     def _combine_score(self, vector_score, fts_score):
-        # these scores represent distance
+        # Both vector_score (inverted distance) and fts_score are in a
-        return 1 - (self.weight * vector_score + (1 - self.weight) * fts_score)
+        # "higher = more relevant" space.  A straight weighted average gives
+        # higher _relevance_score to better matches, as expected.
+        # Previously this returned `1 - (...)` which inverted the final
+        # ranking so that the *least* relevant document ranked first.
+        return self.weight * vector_score + (1 - self.weight) * fts_score
 
     def _invert_score(self, dist: float):
         # Invert the score between relevance and distance

python/python/tests/test_db.py

@@ -466,7 +466,8 @@ async def test_create_table_v2_manifest_paths_async(tmp_path):
     assert await tbl.uses_v2_manifest_paths()
     manifests_dir = tmp_path / "test_v2_manifest_paths.lance" / "_versions"
     for manifest in os.listdir(manifests_dir):
-        assert re.match(r"\d{20}\.manifest", manifest)
+        if manifest.endswith(".manifest"):
+            assert re.match(r"\d{20}\.manifest", manifest)
 
     # Start a table in V1 mode then migrate
     tbl = await db_no_v2_paths.create_table(
@@ -476,13 +477,15 @@ async def test_create_table_v2_manifest_paths_async(tmp_path):
     assert not await tbl.uses_v2_manifest_paths()
     manifests_dir = tmp_path / "test_v2_migration.lance" / "_versions"
     for manifest in os.listdir(manifests_dir):
-        assert re.match(r"\d\.manifest", manifest)
+        if manifest.endswith(".manifest"):
+            assert re.match(r"\d\.manifest", manifest)
 
     await tbl.migrate_manifest_paths_v2()
     assert await tbl.uses_v2_manifest_paths()
 
     for manifest in os.listdir(manifests_dir):
-        assert re.match(r"\d{20}\.manifest", manifest)
+        if manifest.endswith(".manifest"):
+            assert re.match(r"\d{20}\.manifest", manifest)
 
 
 @pytest.mark.asyncio

python/python/tests/test_lsm_write_spec.py

@@ -40,16 +40,6 @@ def _make_table(tmp_path):
 def test_set_lsm_write_spec_validates(tmp_path):
     _db, table = _make_table(tmp_path)
 
-    # No PK set yet.
-    with pytest.raises(Exception, match="primary key"):
-        table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 4))
-
-    table.set_unenforced_primary_key("id")
-
-    # Column mismatch.
-    with pytest.raises(Exception, match="match"):
-        table.set_lsm_write_spec(LsmWriteSpec.bucket("v", 4))
-
     # Out-of-range num_buckets.
     with pytest.raises(Exception, match="num_buckets"):
         table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 0))
@@ -70,7 +60,6 @@ def test_unset_lsm_write_spec(tmp_path):
         table.unset_lsm_write_spec()
 
     # Install a spec, then remove it; afterwards a fresh spec can be set.
-    table.set_unenforced_primary_key("id")
     table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 4))
     table.unset_lsm_write_spec()
     # A second unset errors — there is no spec left to remove.

python/python/tests/test_remote_db.py

@@ -362,6 +362,22 @@ def test_table_create_indices():
                     schema=dict(
                         fields=[
                             dict(name="id", type={"type": "int64"}, nullable=False),
+                            dict(name="text", type={"type": "string"}, nullable=False),
+                            dict(
+                                name="vector",
+                                type={
+                                    "type": "fixed_size_list",
+                                    "fields": [
+                                        dict(
+                                            name="item",
+                                            type={"type": "float"},
+                                            nullable=True,
+                                        )
+                                    ],
+                                    "length": 2,
+                                },
+                                nullable=False,
+                            ),
                         ]
                     ),
                 )

python/python/tests/test_rerankers.py

@@ -603,3 +603,89 @@ def test_cross_encoder_reranker_return_all(tmp_path):
     assert "_relevance_score" in result.column_names
     assert "_score" in result.column_names
     assert "_distance" in result.column_names
+
+
+# ---------------------------------------------------------------------------
+# Regression tests for LinearCombinationReranker scoring bugs (issue #3154)
+# ---------------------------------------------------------------------------
+
+
+def test_linear_combination_best_match_ranks_first():
+    """
+    The document that is BOTH the closest vector match AND the only FTS match
+    must rank first.  Previously _combine_score subtracted from 1, inverting
+    the ranking so the worst document ranked highest.
+    """
+    reranker = LinearCombinationReranker(weight=0.7, return_score="all")
+
+    # rowid 0: perfect vector match, sole FTS match  → should rank 1st
+    # rowid 1: mediocre vector, no FTS match
+    # rowid 2: bad vector, no FTS match
+    vector_results = pa.Table.from_pydict(
+        {
+            "_rowid": [0, 1, 2],
+            "_distance": [0.0, 0.5, 0.9],
+        }
+    )
+    fts_results = pa.Table.from_pydict(
+        {
+            "_rowid": [0],
+            "_score": [1.0],
+        }
+    )
+
+    combined = reranker.merge_results(vector_results, fts_results, fill=1.0)
+    scores = dict(
+        zip(
+            combined["_rowid"].to_pylist(),
+            combined["_relevance_score"].to_pylist(),
+        )
+    )
+
+    # rowid 0 must have the highest relevance score
+    assert scores[0] > scores[1], (
+        f"Best match (rowid 0, score={scores[0]:.4f}) should beat "
+        f"mid match (rowid 1, score={scores[1]:.4f})"
+    )
+    assert scores[1] > scores[2], (
+        f"Mid match (rowid 1, score={scores[1]:.4f}) should beat "
+        f"bad match (rowid 2, score={scores[2]:.4f})"
+    )
+
+
+def test_linear_combination_missing_fts_is_penalised():
+    """
+    A document with no FTS match must score *lower* than a document that
+    has a mediocre FTS match, everything else being equal.  Previously
+    missing-FTS entries used fill=1.0 directly, which gave them a reward
+    (via the 1-(...) inversion) instead of a penalty.
+    """
+    reranker = LinearCombinationReranker(weight=0.5, return_score="all")
+
+    vector_results = pa.Table.from_pydict(
+        {
+            "_rowid": [0, 1],
+            "_distance": [0.2, 0.2],  # identical vector scores
+        }
+    )
+    fts_results = pa.Table.from_pydict(
+        {
+            "_rowid": [0],  # rowid 1 has no FTS match
+            "_score": [0.3],  # small FTS score
+        }
+    )
+
+    combined = reranker.merge_results(vector_results, fts_results, fill=1.0)
+    scores = dict(
+        zip(
+            combined["_rowid"].to_pylist(),
+            combined["_relevance_score"].to_pylist(),
+        )
+    )
+
+    # rowid 0 has a small FTS score; rowid 1 has none.
+    # Even a small FTS contribution should beat having none at all.
+    assert scores[0] > scores[1], (
+        f"Document with FTS score (rowid 0, {scores[0]:.4f}) should beat "
+        f"document with no FTS match (rowid 1, {scores[1]:.4f})"
+    )

rust/lancedb/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb"
-version = "0.29.1-beta.0"
+version = "0.30.0-beta.1"
 edition.workspace = true
 description = "LanceDB: A serverless, low-latency vector database for AI applications"
 license.workspace = true
@@ -104,6 +104,7 @@ datafusion.workspace = true
 http-body = "1"                                        # Matching reqwest
 rstest = "0.23.0"
 test-log = "0.2"
+serial_test = "3"
 
 
 [features]

rust/lancedb/src/connection.rs

@@ -812,8 +812,7 @@ impl ConnectBuilder {
         self
     }
 
-    /// The interval at which to check for updates from other processes. This
+    /// The interval at which to check for updates from other processes.
-    /// only affects LanceDB OSS.
     ///
     /// If left unset, consistency is not checked. For maximum read
     /// performance, this is the default. For strong consistency, set this to
@@ -825,8 +824,11 @@ impl ConnectBuilder {
     /// This only affects read operations. Write operations are always
     /// consistent.
     ///
-    /// LanceDB Cloud uses eventual consistency under the hood, and is not
+    /// # Cost
-    /// currently configurable.
+    ///
+    /// Stronger consistency is not free. The smaller the interval, the more
+    /// often each read pays the cost of checking for updates against object
+    /// storage, raising per-read latency and cost.
     pub fn read_consistency_interval(
         mut self,
         read_consistency_interval: std::time::Duration,
@@ -886,6 +888,7 @@ impl ConnectBuilder {
             options.host_override,
             self.request.client_config,
             storage_options.into(),
+            self.request.read_consistency_interval,
         )?);
         Ok(Connection {
             internal,

rust/lancedb/src/remote/client.rs

@@ -245,6 +245,9 @@ pub struct RestfulLanceDbClient<S: HttpSend = Sender> {
     pub(crate) sender: S,
     pub(crate) id_delimiter: String,
     pub(crate) header_provider: Option<Arc<dyn HeaderProvider>>,
+    /// Connection-level read consistency interval. Drives the
+    /// `x-lancedb-min-timestamp` freshness header sent on read requests.
+    pub(crate) read_consistency_interval: Option<Duration>,
 }
 
 impl<S: HttpSend> std::fmt::Debug for RestfulLanceDbClient<S> {
@@ -338,6 +341,7 @@ impl RestfulLanceDbClient<Sender> {
         host_override: Option<String>,
         default_headers: HeaderMap,
         client_config: ClientConfig,
+        read_consistency_interval: Option<Duration>,
     ) -> Result<Self> {
         // Get the timeouts
         let timeout =
@@ -435,6 +439,7 @@ impl RestfulLanceDbClient<Sender> {
                 .clone()
                 .unwrap_or("$".to_string()),
             header_provider: client_config.header_provider,
+            read_consistency_interval,
         })
     }
 }
@@ -840,6 +845,16 @@ pub mod test_utils {
     pub fn client_with_handler<T>(
         handler: impl Fn(reqwest::Request) -> http::response::Response<T> + Send + Sync + 'static,
     ) -> RestfulLanceDbClient<MockSender>
+    where
+        T: Into<reqwest::Body>,
+    {
+        client_with_handler_and_interval(handler, None)
+    }
+
+    pub fn client_with_handler_and_interval<T>(
+        handler: impl Fn(reqwest::Request) -> http::response::Response<T> + Send + Sync + 'static,
+        read_consistency_interval: Option<Duration>,
+    ) -> RestfulLanceDbClient<MockSender>
     where
         T: Into<reqwest::Body>,
     {
@@ -857,6 +872,7 @@ pub mod test_utils {
             },
             id_delimiter: "$".to_string(),
             header_provider: None,
+            read_consistency_interval,
         }
     }
 
@@ -881,6 +897,7 @@ pub mod test_utils {
             },
             id_delimiter: config.id_delimiter.unwrap_or_else(|| "$".to_string()),
             header_provider: config.header_provider,
+            read_consistency_interval: None,
         }
     }
 }
@@ -888,6 +905,7 @@ pub mod test_utils {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use serial_test::serial;
     use std::time::Duration;
 
     #[test]
@@ -1046,6 +1064,7 @@ mod tests {
             sender: Sender,
             id_delimiter: "+".to_string(),
             header_provider: Some(Arc::new(provider) as Arc<dyn HeaderProvider>),
+            read_consistency_interval: None,
         };
 
         // Apply dynamic headers
@@ -1081,6 +1100,7 @@ mod tests {
             sender: Sender,
             id_delimiter: "+".to_string(),
             header_provider: Some(Arc::new(provider) as Arc<dyn HeaderProvider>),
+            read_consistency_interval: None,
         };
 
         // Apply dynamic headers
@@ -1118,6 +1138,7 @@ mod tests {
             sender: Sender,
             id_delimiter: "+".to_string(),
             header_provider: Some(Arc::new(provider) as Arc<dyn HeaderProvider>),
+            read_consistency_interval: None,
         };
 
         // Header provider errors should fail the request
@@ -1143,6 +1164,7 @@ mod tests {
     }
 
     #[test]
+    #[serial(user_id_env)]
     fn test_resolve_user_id_none() {
         let config = ClientConfig::default();
         // Clear env vars that might be set from other tests
@@ -1155,6 +1177,7 @@ mod tests {
     }
 
     #[test]
+    #[serial(user_id_env)]
     fn test_resolve_user_id_from_env() {
         // SAFETY: This is only called in tests
         unsafe {
@@ -1169,6 +1192,7 @@ mod tests {
     }
 
     #[test]
+    #[serial(user_id_env)]
     fn test_resolve_user_id_from_env_key() {
         // SAFETY: This is only called in tests
         unsafe {
@@ -1189,6 +1213,7 @@ mod tests {
     }
 
     #[test]
+    #[serial(user_id_env)]
     fn test_resolve_user_id_direct_takes_precedence() {
         // SAFETY: This is only called in tests
         unsafe {
@@ -1206,6 +1231,7 @@ mod tests {
     }
 
     #[test]
+    #[serial(user_id_env)]
     fn test_resolve_user_id_empty_env_ignored() {
         // SAFETY: This is only called in tests
         unsafe {

rust/lancedb/src/remote/db.rs

@@ -206,6 +206,7 @@ impl RemoteDatabase {
         host_override: Option<String>,
         client_config: ClientConfig,
         options: RemoteOptions,
+        read_consistency_interval: Option<std::time::Duration>,
     ) -> Result<Self> {
         let parsed = super::client::parse_db_url(uri)?;
         let header_map = RestfulLanceDbClient::<Sender>::default_headers(
@@ -233,6 +234,7 @@ impl RemoteDatabase {
             host_override,
             header_map,
             client_config.clone(),
+            read_consistency_interval,
         )?;
 
         let table_cache = Cache::builder()

rust/lancedb/src/table.rs

@@ -253,6 +253,36 @@ pub enum Filter {
     Datafusion(Expr),
 }
 
+/// A predicate for filtering rows in delete operations.
+///
+/// Accepts either a SQL string or a DataFusion [`Expr`]. Use the [`From`]
+/// implementations to convert from `&str` or `&Expr` automatically.
+/// See [`Table::delete`] for usage examples.
+pub enum Predicate<'a> {
+    /// A SQL predicate string
+    String(&'a str),
+    /// A DataFusion logical expression
+    Expr(&'a Expr),
+}
+
+impl<'a> From<&'a str> for Predicate<'a> {
+    fn from(s: &'a str) -> Self {
+        Predicate::String(s)
+    }
+}
+
+impl<'a> From<&'a String> for Predicate<'a> {
+    fn from(s: &'a String) -> Self {
+        Predicate::String(s.as_str())
+    }
+}
+
+impl<'a> From<&'a Expr> for Predicate<'a> {
+    fn from(e: &'a Expr) -> Self {
+        Predicate::Expr(e)
+    }
+}
+
 #[async_trait]
 pub trait Tags: Send + Sync {
     /// List the tags of the table.
@@ -282,17 +312,15 @@ pub use self::merge::MergeResult;
 /// date) and [`LsmWriteSpec::with_writer_config_defaults`] (default
 /// `ShardWriter` configuration recorded in the MemWAL index).
 ///
-/// All variants require the table to have an unenforced primary key.
-///
 /// Install a spec with [`Table::set_lsm_write_spec`] and remove it with
 /// [`Table::unset_lsm_write_spec`]. The actual `merge_insert` dispatch
 /// onto the MemWAL writer is a follow-up.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub enum LsmWriteSpec {
-    /// Hash-bucket sharding by the unenforced primary key column.
+    /// Hash-bucket sharding by a scalar column.
     ///
-    /// `column` must equal the table's currently-set single-column
+    /// `column` must be a non-nested column with a supported scalar type.
-    /// unenforced primary key. `num_buckets` must be in `[1, 1024]`.
+    /// `num_buckets` must be in `[1, 1024]`.
     /// Iceberg-compatible Murmur3-x86-32 (seed 0) is used so each row's
     /// `bucket(column, num_buckets)` value is stable across processes.
     Bucket {
@@ -491,8 +519,8 @@ pub trait BaseTable: std::fmt::Display + std::fmt::Debug + Send + Sync {
 
     /// Add new records to the table.
     async fn add(&self, add: AddDataBuilder) -> Result<AddResult>;
-    /// Delete rows from the table.
+    /// Delete rows from the table matching the given [`Predicate`].
-    async fn delete(&self, predicate: &str) -> Result<DeleteResult>;
+    async fn delete(&self, predicate: Predicate<'_>) -> Result<DeleteResult>;
     /// Update rows in the table.
     async fn update(&self, update: UpdateBuilder) -> Result<UpdateResult>;
     /// Create an index on the provided column(s).
@@ -656,6 +684,30 @@ mod test_utils {
             }
         }
 
+        pub fn new_with_handler_and_interval<T>(
+            name: impl Into<String>,
+            handler: impl Fn(reqwest::Request) -> http::Response<T> + Clone + Send + Sync + 'static,
+            read_consistency_interval: Option<std::time::Duration>,
+        ) -> Self
+        where
+            T: Into<reqwest::Body>,
+        {
+            let inner = Arc::new(
+                crate::remote::table::RemoteTable::new_mock_with_consistency_interval(
+                    name.into(),
+                    handler.clone(),
+                    read_consistency_interval,
+                ),
+            );
+            let database = Arc::new(crate::remote::db::RemoteDatabase::new_mock(handler));
+            Self {
+                inner,
+                database: Some(database),
+                // Registry is unused.
+                embedding_registry: Arc::new(MemoryRegistry::new()),
+            }
+        }
+
         pub fn new_with_handler_version<T>(
             name: impl Into<String>,
             version: semver::Version,
@@ -860,7 +912,8 @@ impl Table {
     /// Delete the rows from table that match the predicate.
     ///
     /// # Arguments
-    /// - `predicate` - The SQL predicate string to filter the rows to be deleted.
+    /// - `predicate` - A SQL string (`&str`) or DataFusion expression (`&Expr`)
+    ///   that selects the rows to delete.
     ///
     /// # Example
     ///
@@ -869,6 +922,7 @@ impl Table {
     /// # use arrow_array::{FixedSizeListArray, types::Float32Type, RecordBatch,
     /// #   RecordBatchIterator, Int32Array};
     /// # use arrow_schema::{Schema, Field, DataType};
+    /// use datafusion_expr::{col, lit};
     /// # tokio::runtime::Runtime::new().unwrap().block_on(async {
     /// let tmpdir = tempfile::tempdir().unwrap();
     /// let db = lancedb::connect(tmpdir.path().to_str().unwrap())
@@ -898,11 +952,17 @@ impl Table {
     ///     .execute()
     ///     .await
     ///     .unwrap();
+    ///
+    /// // Using a SQL string:
     /// tbl.delete("id > 5").await.unwrap();
+    ///
+    /// // Using a DataFusion expression:
+    /// let expr = col("id").lt(lit(4));
+    /// tbl.delete(&expr).await.unwrap();
     /// # });
     /// ```
-    pub async fn delete(&self, predicate: &str) -> Result<DeleteResult> {
+    pub async fn delete(&self, predicate: impl Into<Predicate<'_>>) -> Result<DeleteResult> {
-        self.inner.delete(predicate).await
+        self.inner.delete(predicate.into()).await
     }
 
     /// Create an index on the provided column(s).
@@ -1298,21 +1358,15 @@ impl Table {
     ///
     /// [`LsmWriteSpec`] chooses one of three sharding strategies:
     ///
-    /// - [`LsmWriteSpec::bucket`] — hash-bucket writes by the single-column
+    /// - [`LsmWriteSpec::bucket`] — hash-bucket writes by a scalar column.
-    ///   unenforced primary key.
     /// - [`LsmWriteSpec::identity`] — shard by the raw value of a scalar column.
     /// - [`LsmWriteSpec::unsharded`] — route every write to a single shard.
     ///
-    /// All variants require the table to have an unenforced primary key
-    /// ([`Table::set_unenforced_primary_key`]); bucket sharding additionally
-    /// requires it to be the single column being bucketed.
-    ///
     /// # Example
     ///
     /// ```
     /// # use lancedb::table::{LsmWriteSpec, Table};
     /// # async fn example(table: &Table) -> Result<(), Box<dyn std::error::Error>> {
-    /// table.set_unenforced_primary_key(["id"]).await?;
     /// table
     ///     .set_lsm_write_spec(
     ///         LsmWriteSpec::bucket("id", 16).with_maintained_indexes(["id_idx"]),
@@ -2777,8 +2831,7 @@ impl BaseTable for NativeTable {
     }
 
     /// Delete rows from the table
-    async fn delete(&self, predicate: &str) -> Result<DeleteResult> {
+    async fn delete(&self, predicate: Predicate<'_>) -> Result<DeleteResult> {
-        // Delegate to the submodule implementation
         delete::execute_delete(self, predicate).await
     }
 
@@ -4600,21 +4653,6 @@ mod tests {
             .unwrap();
         let table = conn.create_table("t", reader).execute().await.unwrap();
 
-        // Reject when no PK is set.
-        let err = table
-            .set_lsm_write_spec(LsmWriteSpec::bucket("id", 4))
-            .await
-            .expect_err("should reject without PK");
-        assert!(matches!(err, Error::Lance { .. }), "got {:?}", err);
-
-        // Set PK, then a mismatched column on the spec must be rejected.
-        table.set_unenforced_primary_key(["id"]).await.unwrap();
-        let err = table
-            .set_lsm_write_spec(LsmWriteSpec::bucket("name", 4))
-            .await
-            .expect_err("should reject column != PK");
-        assert!(matches!(err, Error::Lance { .. }), "got {:?}", err);
-
         // Reject num_buckets out of range.
         for bad in [0u32, 1025] {
             let err = table
@@ -4680,9 +4718,6 @@ mod tests {
             .unwrap();
         let table = conn.create_table("t", reader).execute().await.unwrap();
 
-        // Lance's MemWAL still requires *some* unenforced primary key on
-        // the dataset; Unsharded just skips the per-row hashing step.
-        table.set_unenforced_primary_key(["id"]).await.unwrap();
         table
             .set_lsm_write_spec(LsmWriteSpec::unsharded())
             .await
@@ -4729,7 +4764,6 @@ mod tests {
             .unwrap();
         let table = conn.create_table("t", reader).execute().await.unwrap();
 
-        table.set_unenforced_primary_key(["id"]).await.unwrap();
         table
             .set_lsm_write_spec(
                 LsmWriteSpec::identity("region")
@@ -4785,7 +4819,6 @@ mod tests {
         table.unset_lsm_write_spec().await.unwrap_err();
 
         // Install a spec, then unset it.
-        table.set_unenforced_primary_key(["id"]).await.unwrap();
         table
             .set_lsm_write_spec(LsmWriteSpec::bucket("id", 4))
             .await

rust/lancedb/src/table/add_data.rs

@@ -982,4 +982,105 @@ mod tests {
         table2.add(struct_batch).execute().await.unwrap();
         assert_eq!(table2.count_rows(None).await.unwrap(), 2);
     }
+
+    /// Regression test: appending `arrow.json` (PyArrow `pa.json_()`) data into a table
+    /// whose schema was created with `pa.json_()` (internally stored as `lance.json`, backed
+    /// by `LargeBinary`) must succeed without a schema-mismatch error.
+    ///
+    /// Previously `build_field_exprs` would attempt a `Utf8 → LargeBinary` DataFusion cast,
+    /// which produced a field whose Arrow extension metadata still read `arrow.json` instead
+    /// of `lance.json`.  Lance-core then rejected the append with
+    /// `"json vs large_binary" schema mismatch`.
+    ///
+    /// PyArrow's `pa.json_()` may be backed by either `Utf8` or `LargeUtf8` depending on the
+    /// constructor used, so the test is parameterized over the input backing type.
+    #[rstest::rstest]
+    #[case::utf8(DataType::Utf8)]
+    #[case::large_utf8(DataType::LargeUtf8)]
+    #[tokio::test]
+    async fn test_add_arrow_json_into_lance_json_table(#[case] input_type: DataType) {
+        use arrow_array::{Array, cast::AsArray};
+        use lance_arrow::ARROW_EXT_NAME_KEY;
+        use lance_arrow::json::{ARROW_JSON_EXT_NAME, JSON_EXT_NAME};
+
+        // Build a table whose "data" column is lance.json (LargeBinary +
+        // ARROW:extension:name = "lance.json").
+        let lance_json_field = lance_arrow::json::json_field("data", true);
+        let table_schema = Arc::new(Schema::new(vec![lance_json_field]));
+
+        let db = connect("memory://").execute().await.unwrap();
+        let table = db
+            .create_empty_table("json_test", table_schema)
+            .execute()
+            .await
+            .unwrap();
+
+        // Sanity-check the stored schema.
+        let stored_field = table.schema().await.unwrap();
+        let data_field = stored_field.field_with_name("data").unwrap();
+        assert_eq!(data_field.data_type(), &DataType::LargeBinary);
+        assert_eq!(
+            data_field
+                .metadata()
+                .get(ARROW_EXT_NAME_KEY)
+                .map(|s| s.as_str()),
+            Some(JSON_EXT_NAME),
+        );
+
+        // Build an arrow.json input field (Utf8/LargeUtf8 + arrow.json extension).
+        // This is what PyArrow produces for pa.json_() arrays.
+        let arrow_json_metadata = std::collections::HashMap::from([(
+            ARROW_EXT_NAME_KEY.to_string(),
+            ARROW_JSON_EXT_NAME.to_string(),
+        )]);
+        let arrow_json_field =
+            Field::new("data", input_type.clone(), true).with_metadata(arrow_json_metadata);
+        let arrow_json_schema = Arc::new(Schema::new(vec![arrow_json_field]));
+
+        let rows: Vec<Option<&str>> = vec![None, Some(r#"{"a": 1}"#), Some(r#"{"b": 2}"#)];
+        let string_array: Arc<dyn arrow_array::Array> = match input_type {
+            DataType::Utf8 => Arc::new(arrow_array::StringArray::from(rows.clone())),
+            DataType::LargeUtf8 => Arc::new(arrow_array::LargeStringArray::from(rows.clone())),
+            other => panic!("unsupported arrow.json backing type for this test: {other:?}"),
+        };
+        let batch = RecordBatch::try_new(arrow_json_schema, vec![string_array]).unwrap();
+
+        // This must not fail with a schema-mismatch error.
+        table.add(batch).execute().await.unwrap();
+
+        assert_eq!(table.count_rows(None).await.unwrap(), rows.len());
+
+        // A lance.json column is read back as Utf8 carrying arrow.json extension metadata.
+        let results: Vec<RecordBatch> = table
+            .query()
+            .select(Select::columns(&["data"]))
+            .execute()
+            .await
+            .unwrap()
+            .try_collect()
+            .await
+            .unwrap();
+
+        assert_eq!(results.len(), 1);
+        let batch = &results[0];
+        assert_eq!(batch.num_rows(), rows.len());
+
+        let json_col = batch.column(0);
+        assert_eq!(json_col.data_type(), &DataType::Utf8);
+        let json_strs = json_col.as_string::<i32>();
+
+        for (i, expected) in rows.iter().enumerate() {
+            match expected {
+                None => assert!(json_strs.is_null(i), "row {i} expected null"),
+                Some(raw) => {
+                    assert!(!json_strs.is_null(i), "row {i} expected non-null");
+                    let actual: serde_json::Value = serde_json::from_str(json_strs.value(i))
+                        .expect("read-back JSON should be valid");
+                    let expected: serde_json::Value =
+                        serde_json::from_str(raw).expect("expected JSON should be valid");
+                    assert_eq!(actual, expected, "row {i} JSON mismatch");
+                }
+            }
+        }
+    }
 }

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

@@ -13,6 +13,7 @@ use datafusion_physical_expr::expressions::{CastExpr, Literal};
 use datafusion_physical_plan::expressions::Column;
 use datafusion_physical_plan::projection::ProjectionExec;
 use datafusion_physical_plan::{ExecutionPlan, PhysicalExpr};
+use lance_arrow::json::{is_arrow_json_field, is_json_field};
 
 use crate::{Error, Result};
 
@@ -64,6 +65,18 @@ fn build_field_exprs(
         let input_field = &input_fields[input_idx];
         let input_expr = get_input_expr(input_idx);
 
+        // Special case: input is arrow.json (PyArrow pa.json_() extension type backed by
+        // Utf8/LargeUtf8) and the table field is lance.json (backed by LargeBinary).
+        // Lance-core's write path already handles the arrow.json → lance.json conversion
+        // (including JSONB encoding), so we pass the expression through unchanged and let
+        // lance-core deal with it. Attempting to cast Utf8 → LargeBinary here would
+        // produce a field whose metadata still identifies it as arrow.json, which then
+        // causes a schema-mismatch error inside lance-core.
+        if is_arrow_json_field(input_field) && is_json_field(table_field) {
+            result.push((input_expr, Arc::clone(input_field) as FieldRef));
+            continue;
+        }
+
         let expr = match (input_field.data_type(), table_field.data_type()) {
             // Both are structs: recurse into sub-fields to handle subschemas and casts.
             (DataType::Struct(in_children), DataType::Struct(tbl_children))
@@ -618,4 +631,75 @@ mod tests {
             .unwrap();
         assert_eq!(a.values(), &[1, 3]);
     }
+
+    /// `arrow.json` input (PyArrow `pa.json_()`, Utf8/LargeUtf8 + extension metadata) against a
+    /// `lance.json` table field (LargeBinary + extension metadata) must be passed through
+    /// without a cast so that lance-core can perform its own arrow.json → JSONB conversion.
+    ///
+    /// Before the fix, `cast_to_table_schema` attempted a `Utf8 → LargeBinary` DataFusion
+    /// cast that preserved the wrong extension metadata, causing lance-core to reject the
+    /// batch with a "json vs large_binary" schema-mismatch error.
+    #[rstest::rstest]
+    #[case::utf8(DataType::Utf8)]
+    #[case::large_utf8(DataType::LargeUtf8)]
+    #[tokio::test]
+    async fn test_arrow_json_passthrough_to_lance_json(#[case] input_type: DataType) {
+        use lance_arrow::ARROW_EXT_NAME_KEY;
+        use lance_arrow::json::{ARROW_JSON_EXT_NAME, json_field};
+
+        // Build a table schema with a lance.json field (LargeBinary + lance.json metadata).
+        let lance_field = json_field("data", true);
+        let table_schema = Schema::new(vec![lance_field]);
+
+        // Build an input batch with an arrow.json field (Utf8/LargeUtf8 + arrow.json metadata).
+        let arrow_meta = std::collections::HashMap::from([(
+            ARROW_EXT_NAME_KEY.to_string(),
+            ARROW_JSON_EXT_NAME.to_string(),
+        )]);
+        let arrow_field = Field::new("data", input_type.clone(), true).with_metadata(arrow_meta);
+        let input_schema = Arc::new(Schema::new(vec![arrow_field]));
+
+        let values = vec![Some(r#"{"x": 1}"#), None, Some(r#"{"y": 2}"#)];
+        let input_array: Arc<dyn arrow_array::Array> = match input_type {
+            DataType::Utf8 => Arc::new(StringArray::from(values)),
+            DataType::LargeUtf8 => Arc::new(arrow_array::LargeStringArray::from(values)),
+            other => panic!("unsupported arrow.json backing type for this test: {other:?}"),
+        };
+        let input_batch = RecordBatch::try_new(input_schema, vec![input_array]).unwrap();
+
+        let plan = plan_from_batch(input_batch).await;
+        let projected = cast_to_table_schema(plan, &table_schema).unwrap();
+
+        // The projected schema's "data" field must carry arrow.json metadata
+        // (the input field), not be silently dropped or miscast.
+        let out_field = projected.schema().field_with_name("data").unwrap().clone();
+        assert_eq!(out_field.data_type(), &input_type);
+        assert_eq!(
+            out_field
+                .metadata()
+                .get(ARROW_EXT_NAME_KEY)
+                .map(|s| s.as_str()),
+            Some(ARROW_JSON_EXT_NAME),
+            "output field must still carry arrow.json metadata so lance-core can handle it"
+        );
+
+        // The data must flow through correctly (3 rows, no panic).
+        let result = collect(projected).await;
+        assert_eq!(result.num_rows(), 3);
+        let (v0, v2) = match input_type {
+            DataType::Utf8 => {
+                let col: &StringArray = result.column(0).as_any().downcast_ref().unwrap();
+                (col.value(0).to_string(), col.value(2).to_string())
+            }
+            DataType::LargeUtf8 => {
+                let col: &arrow_array::LargeStringArray =
+                    result.column(0).as_any().downcast_ref().unwrap();
+                (col.value(0).to_string(), col.value(2).to_string())
+            }
+            _ => unreachable!(),
+        };
+        assert_eq!(v0, r#"{"x": 1}"#);
+        assert!(result.column(0).is_null(1));
+        assert_eq!(v2, r#"{"y": 2}"#);
+    }
 }

rust/lancedb/src/table/delete.rs

@@ -1,9 +1,12 @@
+use std::sync::Arc;
+
 use futures::FutureExt;
+use lance::dataset::DeleteBuilder;
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright The LanceDB Authors
 use serde::{Deserialize, Serialize};
 
-use super::NativeTable;
+use super::{NativeTable, Predicate};
 use crate::Result;
 
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize, Default)]
@@ -21,17 +24,39 @@ pub struct DeleteResult {
 /// Internal implementation of the delete logic
 ///
 /// This logic was moved from NativeTable::delete to keep table.rs clean.
-pub(crate) async fn execute_delete(table: &NativeTable, predicate: &str) -> Result<DeleteResult> {
+pub(crate) async fn execute_delete(
+    table: &NativeTable,
+    predicate: Predicate<'_>,
+) -> Result<DeleteResult> {
     table.dataset.ensure_mutable()?;
-    let mut dataset = (*table.dataset.get().await?).clone();
+    match predicate {
-    let delete_result = dataset.delete(predicate).boxed().await?;
+        Predicate::String(s) => {
-    let num_deleted_rows = delete_result.num_deleted_rows;
+            let mut dataset = (*table.dataset.get().await?).clone();
-    let version = dataset.version().version;
+            let delete_result = dataset.delete(s).boxed().await?;
-    table.dataset.update(dataset);
+            let num_deleted_rows = delete_result.num_deleted_rows;
-    Ok(DeleteResult {
+            let version = dataset.version().version;
-        num_deleted_rows,
+            table.dataset.update(dataset);
-        version,
+            Ok(DeleteResult {
-    })
+                num_deleted_rows,
+                version,
+            })
+        }
+        Predicate::Expr(expr) => {
+            let dataset = table.dataset.get().await?;
+            let delete_result = DeleteBuilder::from_expr(Arc::clone(&dataset), expr.clone())
+                .execute()
+                .await?;
+            let num_deleted_rows = delete_result.num_deleted_rows;
+            let version = delete_result.new_dataset.version().version;
+            table.dataset.update(
+                Arc::try_unwrap(delete_result.new_dataset).unwrap_or_else(|arc| (*arc).clone()),
+            );
+            Ok(DeleteResult {
+                num_deleted_rows,
+                version,
+            })
+        }
+    }
 }
 
 #[cfg(test)]
@@ -176,4 +201,100 @@ mod tests {
             "Table version must increment after delete operation"
         );
     }
+
+    #[tokio::test]
+    async fn test_delete_expr() {
+        use datafusion_expr::{col, lit};
+
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // 1. Create a table with values 0 to 9
+        let schema = Arc::new(Schema::new(vec![Field::new("i", DataType::Int32, false)]));
+        let batch = RecordBatch::try_new(
+            schema.clone(),
+            vec![Arc::new(Int32Array::from_iter_values(0..10))],
+        )
+        .unwrap();
+
+        let table = conn
+            .create_table("test_delete_expr", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // 2. Verify initial state
+        assert_eq!(table.count_rows(None).await.unwrap(), 10);
+        let initial_version = table.version().await.unwrap();
+
+        // 3. Execute Delete with Expr (removes values > 5)
+        let expr = col("i").gt(lit(5));
+        table.delete(&expr).await.unwrap();
+
+        // 4. Verify results
+        assert_eq!(table.count_rows(None).await.unwrap(), 6); // 0, 1, 2, 3, 4, 5 remain
+        let current_version = table.version().await.unwrap();
+        assert!(
+            current_version > initial_version,
+            "Table version must increment after delete_expr operation"
+        );
+
+        // 5. Verify specific data consistency
+        let batches = table
+            .query()
+            .execute()
+            .await
+            .unwrap()
+            .try_collect::<Vec<_>>()
+            .await
+            .unwrap();
+        let batch = &batches[0];
+        let array = batch
+            .column(0)
+            .as_any()
+            .downcast_ref::<Int32Array>()
+            .unwrap();
+
+        // Ensure no value > 5 exists
+        for val in array.iter() {
+            assert!(val.unwrap() <= 5);
+        }
+    }
+
+    #[tokio::test]
+    async fn test_delete_expr_increments_version() {
+        use datafusion_expr::lit;
+
+        let conn = connect("memory://").execute().await.unwrap();
+
+        // Create a table with 5 rows
+        let batch = record_batch!(("id", Int32, [1, 2, 3, 4, 5])).unwrap();
+
+        let table = conn
+            .create_table("test_delete_expr_noop", batch)
+            .execute()
+            .await
+            .unwrap();
+
+        // Capture the initial state (Rows = 5, Version = 1)
+        let initial_rows = table.count_rows(None).await.unwrap();
+        let initial_version = table.version().await.unwrap();
+
+        assert_eq!(initial_rows, 5);
+        let expr = lit(false);
+        table.delete(&expr).await.unwrap();
+
+        // Rows should still be 5
+        let current_rows = table.count_rows(None).await.unwrap();
+        assert_eq!(
+            current_rows, initial_rows,
+            "Data should not change when predicate is false"
+        );
+
+        // version check
+        let current_version = table.version().await.unwrap();
+        assert!(
+            current_version > initial_version,
+            "Table version must increment after delete_expr operation"
+        );
+    }
 }

rust/lancedb/src/utils/mod.rs

@@ -6,7 +6,7 @@ pub(crate) mod background_cache;
 use std::sync::Arc;
 
 use arrow_array::RecordBatch;
-use arrow_schema::{DataType, Field, Fields, Schema, SchemaRef};
+use arrow_schema::{DataType, Field, Schema, SchemaRef};
 use datafusion_common::{DataFusionError, Result as DataFusionResult};
 use datafusion_execution::RecordBatchStream;
 use futures::{FutureExt, Stream};
@@ -199,38 +199,32 @@ fn collect_vector_columns(
     path.pop();
 }
 
-pub(crate) fn resolve_arrow_field_path(schema: &Schema, column: &str) -> Result<Field> {
+pub(crate) fn resolve_arrow_field_path(schema: &Schema, column: &str) -> Result<(String, Field)> {
-    let segments =
+    lance_core::datatypes::parse_field_path(column).map_err(|e| Error::InvalidInput {
-        lance_core::datatypes::parse_field_path(column).map_err(|e| Error::InvalidInput {
+        message: format!("Invalid field path `{}`: {}", column, e),
-            message: format!("Invalid field path `{}`: {}", column, e),
+    })?;
+
+    let lance_schema =
+        lance_core::datatypes::Schema::try_from(schema).map_err(|e| Error::Schema {
+            message: format!("Invalid schema: {}", e),
         })?;
-    let mut fields = schema.fields();
+    let field_path = lance_schema
-
+        .resolve_case_insensitive(column)
-    for (idx, segment) in segments.iter().enumerate() {
+        .ok_or_else(|| Error::Schema {
-        let field = find_field(fields, segment).ok_or_else(|| Error::Schema {
+            message: format!(
-            message: format!("Field path `{}` not found in schema", column),
+                "Field path `{}` not found in schema. Available field paths: {}",
+                column,
+                lance_schema.field_paths().join(", ")
+            ),
         })?;
-        if idx + 1 == segments.len() {
+    let field = field_path.last().expect("field path should be non-empty");
-            return Ok(field.clone());
+    let path_segments = field_path
-        }
-        fields = match field.data_type() {
-            DataType::Struct(fields) => fields,
-            _ => {
-                return Err(Error::Schema {
-                    message: format!("Field path `{}` not found in schema", column),
-                });
-            }
-        };
-    }
-
-    unreachable!("parse_field_path returns at least one segment")
-}
-
-fn find_field<'a>(fields: &'a Fields, name: &str) -> Option<&'a Field> {
-    fields
         .iter()
-        .find(|field| field.name() == name)
+        .map(|field| field.name.as_str())
-        .map(|field| field.as_ref())
+        .collect::<Vec<_>>();
+    let canonical_path = lance_core::datatypes::format_field_path(&path_segments);
+
+    Ok((canonical_path, Field::from(*field)))
 }
 
 pub fn supported_btree_data_type(dtype: &DataType) -> bool {