fix(rust): reuse SQ preflight compaction plan

fix(rust): safely classify legacy vector indexes
test(rust): use valid legacy compaction fixture
2026-08-31 10:38:31 +00:00 · 2026-08-06 11:30:18 +00:00 · 2026-08-06 11:13:04 +00:00 · 2026-08-06 10:31:30 +00:00 · 2026-08-06 10:12:16 +00:00 · 2026-08-06 09:14:00 +00:00
33 changed files with 602 additions and 1092 deletions
@@ -1,222 +0,0 @@
-name: Check doc links
-
-# Checking external links is inherently noisy: third-party sites rate-limit
-# automated clients, reject non-browser user agents, and go down temporarily.
-# Blocking pull requests on that trades a lot of false failures for very little
-# signal, so this runs on a schedule and reports findings in a single tracking
-# issue instead of failing anyone's build.
-on:
-  schedule:
-    - cron: "0 7 * * *"
-  workflow_dispatch:
-
-# The report lives in one repository-global issue, so runs must not overlap: a
-# lookup racing a create produces duplicate issues, and a healthy run closing
-# the issue while a failing run only rewrites its body would leave a broken
-# report closed. The group is deliberately ref-independent so that a manual
-# dispatch serializes against the scheduled run.
-concurrency:
-  group: docs-link-check
-  cancel-in-progress: false
-
-permissions: {}
-
-env:
-  REPORT_TITLE: "Docs link checker report"
-
-jobs:
-  scan:
-    name: Scan links
-    runs-on: ubuntu-24.04
-    # lychee-action is pinned by SHA, but its wrapper downloads the lychee
-    # release tarball at run time without verifying a digest, and hands the
-    # resulting binary a GitHub token. Release assets remain replaceable, so
-    # that binary is confined to a job whose token can only read public
-    # content; everything that writes runs in the report job below.
-    permissions:
-      contents: read
-    outputs:
-      exit_code: ${{ steps.lychee.outputs.exit_code }}
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v6
-        with:
-          # workflow_dispatch can run from any ref, but the report is
-          # repository-global. Always measure the default branch so a manual
-          # run from a topic branch cannot close a report that main warrants,
-          # or overwrite it with branch-only findings.
-          ref: ${{ github.event.repository.default_branch }}
-          persist-credentials: false
-
-      - name: Check links
-        id: lychee
-        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 `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
-          # broken on every run.
-          args: >-
-            --scheme https
-            --scheme http
-            --no-progress
-            --max-retries 3
-            --timeout 20
-            'docs/src/**/*.md'
-          format: json
-          output: ./lychee/out.json
-          jobSummary: false
-          # The report, not a red build, is the signal for broken links. The
-          # validation step below still fails the run if the check itself
-          # breaks.
-          fail: false
-
-      - name: Validate report
-        # lychee does not reserve exit code 2 for broken links: its CLI
-        # parser also exits 2 on an invalid option, before any link was
-        # checked or any report written. Only a parseable report whose
-        # counts agree with the exit code counts as a link verdict; anything
-        # else fails here, and the report job below is skipped entirely, so
-        # the tracking issue is never touched. Exit 2 covers timeouts as
-        # well as errors, and a timed-out host is exactly the transient
-        # unavailability this report exists to surface, so both count as
-        # findings. Requiring total > 0 also catches a glob that silently
-        # stopped matching any file.
-        if: steps.lychee.outputs.exit_code == 0 || steps.lychee.outputs.exit_code == 2
-        env:
-          EXIT_CODE: ${{ steps.lychee.outputs.exit_code }}
-        run: |
-          jq -e --argjson code "$EXIT_CODE" '
-            (.total > 0) and
-            (if $code == 0
-             then .errors == 0 and .timeouts == 0
-               and (.error_map | length == 0) and (.timeout_map | length == 0)
-             else (.errors + .timeouts) > 0
-               and ((.error_map | length) + (.timeout_map | length)) > 0
-             end)
-          ' ./lychee/out.json
-
-      - name: Upload report
-        if: steps.lychee.outputs.exit_code == 2
-        uses: actions/upload-artifact@v7
-        with:
-          name: link-report
-          path: ./lychee/out.json
-          retention-days: 7
-
-  report:
-    name: Update report issue
-    needs: scan
-    runs-on: ubuntu-24.04
-    # Deliberately no checkout: this job needs the report artifact and the
-    # issues API, not the repository contents.
-    permissions:
-      issues: write
-    env:
-      EXIT_CODE: ${{ needs.scan.outputs.exit_code }}
-      GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-    steps:
-      - name: Classify checker result
-        # lychee exits 0 when every link resolves and 2 when links fail,
-        # both already cross-checked against the report by the scan job's
-        # validation step. Anything else (1 runtime, 3 bad config) means the
-        # check never produced a link verdict, which must surface as a failed
-        # run rather than be published as "broken documentation links".
-        run: |
-          case "$EXIT_CODE" in
-            0|2)
-              echo "lychee exit code $EXIT_CODE"
-              ;;
-            *)
-              echo "::error::lychee exited with '$EXIT_CODE': the link check did not complete. Leaving the report issue untouched."
-              exit 1
-              ;;
-          esac
-
-      - name: Find existing report issue
-        id: report
-        # Matched on title alone, and through search rather than a listing:
-        # the issue action applies labels in a separate call after creating the
-        # issue, so a label filter misses a half-created report, and this
-        # repository has far more open issues than one listing page holds.
-        # Closed issues are included because a healthy run closes the report:
-        # an open-only lookup would forget that identity and the next failing
-        # run would open a duplicate. The oldest match stays the canonical
-        # report and is reopened below when links break again.
-        run: |
-          match=$(gh issue list --repo "$GITHUB_REPOSITORY" --state all \
-            --search "in:title \"$REPORT_TITLE\" author:app/github-actions" \
-            --limit 50 --json number,title,state \
-            --jq "[.[] | select(.title == \"$REPORT_TITLE\")] | sort_by(.number) | first // empty")
-          echo "number=$(jq -r '.number // empty' <<<"$match")" >> "$GITHUB_OUTPUT"
-          echo "state=$(jq -r '.state // empty' <<<"$match")" >> "$GITHUB_OUTPUT"
-
-      - name: Download report
-        if: env.EXIT_CODE == 2
-        uses: actions/download-artifact@v8
-        with:
-          name: link-report
-          path: ./lychee
-
-      - name: Compose report
-        if: env.EXIT_CODE == 2
-        run: |
-          run_url="$GITHUB_SERVER_URL/$GITHUB_REPOSITORY/actions/runs/$GITHUB_RUN_ID"
-          {
-            echo "Broken documentation links found by [\`$GITHUB_WORKFLOW\`]($run_url)."
-            echo
-            echo "This issue is rewritten by every scheduled run and closed automatically once all links resolve."
-            echo
-            echo "Entries can be false positives: some sites rate-limit or block automated clients while working fine in a browser. Confirm before editing the docs, and add persistent offenders to \`--exclude\` in \`.github/workflows/docs-link-check.yml\`."
-            echo
-            # Timeouts are reported alongside errors: entries land in
-            # timeout_map with a status text instead of an HTTP code.
-            jq -r '
-              "\(.errors) of \(.total) links failed, \(.timeouts) timed out.",
-              "",
-              ([(.error_map | to_entries[]), (.timeout_map | to_entries[])]
-               | group_by(.key)[] |
-                "### Errors in \(.[0].key)",
-                "",
-                (map(.value[])[] | "* [\(.status.code // .status.text // "ERR")] <\(.url)> — \(.status.details // .status.text // "unknown error")"),
-                "")
-            ' ./lychee/out.json
-          } > ./lychee/issue.md
-
-      - name: Reopen report issue
-        # A healthy run closes the report, and the issue action below only
-        # rewrites the body of whatever number it is given. Without an
-        # explicit reopen, the 2 -> 0 -> 2 sequence would keep rewriting a
-        # closed issue while links are broken. A CLOSED state implies the
-        # lookup found a canonical issue, so no separate emptiness check.
-        if: env.EXIT_CODE == 2 && steps.report.outputs.state == 'CLOSED'
-        env:
-          ISSUE_NUMBER: ${{ steps.report.outputs.number }}
-        run: |
-          run_url="$GITHUB_SERVER_URL/$GITHUB_REPOSITORY/actions/runs/$GITHUB_RUN_ID"
-          gh issue reopen "$ISSUE_NUMBER" --repo "$GITHUB_REPOSITORY" \
-            --comment "Broken documentation links found again in [the latest run]($run_url)."
-
-      - name: Report broken links
-        if: env.EXIT_CODE == 2
-        uses: peter-evans/create-issue-from-file@fca9117c27cdc29c6c4db3b86c48e4115a786710 # v6.0.0
-        with:
-          # Empty on the first failing run, which creates the issue; afterwards
-          # the same issue is updated in place.
-          issue-number: ${{ steps.report.outputs.number }}
-          title: ${{ env.REPORT_TITLE }}
-          content-filepath: ./lychee/issue.md
-          labels: documentation
-
-      - name: Close report issue once links are healthy
-        # An OPEN state implies the lookup found a canonical issue; a report
-        # that is already closed needs nothing.
-        if: env.EXIT_CODE == 0 && steps.report.outputs.state == 'OPEN'
-        env:
-          ISSUE_NUMBER: ${{ steps.report.outputs.number }}
-        run: |
-          run_url="$GITHUB_SERVER_URL/$GITHUB_REPOSITORY/actions/runs/$GITHUB_RUN_ID"
-          gh issue close "$ISSUE_NUMBER" --repo "$GITHUB_REPOSITORY" \
-            --comment "All documentation links resolved in [the latest run]($run_url)."
@@ -296,18 +296,16 @@ jobs:
          cargo update -p aws-types --precise 1.3.9
          cargo update -p aws-sigv4 --precise 1.3.5
          cargo update -p aws-credential-types --precise 1.2.8
-          # aws-smithy-checksums must stay at or above 0.63.13: OpenDAL's S3
-          # service needs crc-fast ~1.9, and older releases pin it to ~1.3.
-          cargo update -p aws-smithy-checksums --precise 0.63.13
+          cargo update -p aws-smithy-checksums --precise 0.63.9
          cargo update -p aws-smithy-runtime --precise 1.9.3
-          cargo update -p aws-smithy-http --precise 0.62.6
-          cargo update -p aws-smithy-eventstream --precise 0.60.14
+          cargo update -p aws-smithy-http --precise 0.62.4
+          cargo update -p aws-smithy-eventstream --precise 0.60.12
          cargo update -p aws-smithy-http-client --precise 1.1.3
          cargo update -p aws-smithy-observability --precise 0.1.4
          cargo update -p aws-smithy-query --precise 0.60.8
-          cargo update -p aws-smithy-runtime-api --precise 1.9.3
-          cargo update -p aws-smithy-async --precise 1.2.7
-          cargo update -p aws-smithy-types --precise 1.3.6
+          cargo update -p aws-smithy-runtime-api --precise 1.9.1
+          cargo update -p aws-smithy-async --precise 1.2.6
+          cargo update -p aws-smithy-types --precise 1.3.5
          cargo update -p aws-smithy-xml --precise 0.60.11
          cargo update -p home --precise 0.5.9
      - name: cargo +${{ matrix.msrv }} check
@@ -13,20 +13,20 @@ categories = ["database-implementations"]
 rust-version = "1.91.0"

 [workspace.dependencies]
-lance = { "version" = "=11.0.0-beta.2", default-features = false, "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-core = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-datagen = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-file = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-io = { "version" = "=11.0.0-beta.2", default-features = false, "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-index = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-linalg = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-namespace-impls = { "version" = "=11.0.0-beta.2", default-features = false, "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-table = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-testing = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-datafusion = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-encoding = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
-lance-arrow = { "version" = "=11.0.0-beta.2", "tag" = "v11.0.0-beta.2", "git" = "https://github.com/lance-format/lance.git" }
+lance = { "version" = "=10.1.0-beta.1", default-features = false, "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-core = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-datagen = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-file = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-io = { "version" = "=10.1.0-beta.1", default-features = false, "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-index = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-linalg = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-namespace-impls = { "version" = "=10.1.0-beta.1", default-features = false, "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-table = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-testing = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-datafusion = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-encoding = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "git" = "https://github.com/lance-format/lance.git" }
+lance-arrow = { "version" = "=10.1.0-beta.1", "tag" = "v10.1.0-beta.1", "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 }
@@ -52,7 +52,7 @@ env_logger = "0.11"
 half = { "version" = "2.7.1", default-features = false, features = [
    "num-traits",
 ] }
-futures = "0.3"
+futures = "0"
 log = "0.4"
 metrics = "0.24"
 metrics-util = "0.19"
@@ -31,7 +31,7 @@ is also an [asynchronous API client](#connections-asynchronous).
 ## Namespaces (Synchronous)

 A namespace-backed connection resolves tables through a
-[Lance namespace](https://lance-format.github.io/lance-namespace/) service instead of
+[Lance namespace](https://lancedb.github.io/lance-namespace/) service instead of
 listing a storage directory.

 ::: lancedb.connect_namespace
@@ -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-beta.2</lance-core.version>
+        <lance-core.version>10.1.0-beta.1</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>
@@ -11,11 +11,8 @@ import {
  Float16,
  Float32,
  Float64,
-  Int32,
  Schema,
  Utf8,
-  fromDataToBuffer,
-  tableFromIPC,
 } from "../lancedb/arrow";
 import { EmbeddingFunction, LanceSchema } from "../lancedb/embedding";
 import { getRegistry, register } from "../lancedb/embedding/registry";
@@ -187,63 +184,6 @@ describe("embedding functions", () => {
    const vector0 = JSON.parse(JSON.stringify(arr[0].vector));
    expect(vector0).toEqual([1, 2, 3]);
  });
-
-  it("should append generated vectors to a non-nullable schema", async () => {
-    @register("non_nullable_schema_test")
-    class MockEmbeddingFunction extends EmbeddingFunction<string> {
-      ndims() {
-        return 3;
-      }
-      embeddingDataType(): Float {
-        return new Float64();
-      }
-      async computeSourceEmbeddings(data: string[]) {
-        return data.map(() => [1, 2, 3]);
-      }
-    }
-
-    const schema = new Schema([
-      new Field("id", new Int32()),
-      new Field("text", new Utf8()),
-      new Field("type", new Utf8()),
-      new Field(
-        "vector",
-        new FixedSizeList(3, new Field("item", new Float64())),
-      ),
-    ]);
-    const func = new MockEmbeddingFunction();
-    const db = await connect(tmpDir.name);
-    const table = await db.createEmptyTable("test_non_nullable", schema, {
-      embeddingFunction: {
-        function: func,
-        sourceColumn: "text",
-      },
-    });
-
-    const data = [
-      { id: 1, text: "Carrot", type: "vegetable" },
-      { id: 2, text: "Apple", type: "fruit" },
-    ];
-    const buffer = await fromDataToBuffer(
-      data,
-      undefined,
-      await table.schema(),
-    );
-    const generatedTable = tableFromIPC(buffer);
-    const vectorField = generatedTable.schema.fields.find(
-      (field) => field.name === "vector",
-    );
-    expect(vectorField?.nullable).toBe(false);
-
-    await table.add(data);
-
-    const rows = await table.query().toArray();
-    expect(rows).toHaveLength(2);
-    for (const row of rows) {
-      expect([...row.vector]).toEqual([1, 2, 3]);
-    }
-  });
-
  it("should error when appending to a table with an unregistered embedding function", async () => {
    @register("mock")
    class MockEmbeddingFunction extends EmbeddingFunction<string> {
@@ -1,14 +0,0 @@
-// SPDX-License-Identifier: Apache-2.0
-// SPDX-FileCopyrightText: Copyright The LanceDB Authors
-
-import packageJson = require("../package.json");
-
-describe("package metadata", () => {
-  it("requires Node.js type declarations compatible with the runtime", () => {
-    expect(packageJson.engines.node).toBe(">= 18");
-    expect(packageJson.peerDependencies["@types/node"]).toBe(">=18");
-    expect(packageJson.peerDependenciesMeta["@types/node"]).toEqual({
-      optional: true,
-    });
-  });
-});
@@ -110,81 +110,6 @@ describe("Query outputSchema", () => {
  });
 });

-describe("Search pagination", () => {
-  let tmpDir: tmp.DirResult;
-  let table: Table;
-
-  beforeEach(async () => {
-    tmpDir = tmp.dirSync({ unsafeCleanup: true });
-    const db = await connect(tmpDir.name);
-    const schema = new Schema([
-      new Field("id", new Int64(), false),
-      new Field("text", new Utf8(), false),
-      new Field(
-        "vector",
-        new FixedSizeList(2, new Field("item", new Float32())),
-        false,
-      ),
-    ]);
-    const data = makeArrowTable(
-      [
-        { id: 1n, text: "common", vector: [0, 0] },
-        { id: 2n, text: "common common", vector: [1, 1] },
-        { id: 3n, text: "common common common", vector: [2, 2] },
-        { id: 4n, text: "common common common common", vector: [3, 3] },
-      ],
-      { schema },
-    );
-    table = await db.createTable("test", data);
-  });
-
-  afterEach(() => {
-    tmpDir.removeCallback();
-  });
-
-  it("applies offset after the vector search limit", async () => {
-    const allResults = await table
-      .vectorSearch([0, 0])
-      .select(["id"])
-      .limit(4)
-      .toArray();
-    const secondPage = await table
-      .vectorSearch([0, 0])
-      .select(["id"])
-      .limit(2)
-      .offset(2)
-      .toArray();
-
-    expect(allResults).toHaveLength(4);
-    expect(secondPage).toHaveLength(2);
-    expect(secondPage.map((row) => row.id)).toEqual(
-      allResults.slice(2, 4).map((row) => row.id),
-    );
-  });
-
-  it("applies offset after the full-text search limit", async () => {
-    await table.createIndex("text", { config: Index.fts() });
-
-    const allResults = await table
-      .search("common", "fts")
-      .select(["id"])
-      .limit(4)
-      .toArray();
-    const secondPage = await table
-      .search("common", "fts")
-      .select(["id"])
-      .limit(2)
-      .offset(2)
-      .toArray();
-
-    expect(allResults).toHaveLength(4);
-    expect(secondPage).toHaveLength(2);
-    expect(secondPage.map((row) => row.id)).toEqual(
-      allResults.slice(2, 4).map((row) => row.id),
-    );
-  });
-});
-
 describe("Query orderBy", () => {
  let tmpDir: tmp.DirResult;
  let table: Table;
@@ -55,13 +55,7 @@
        "openai": "4.29.2"
      },
      "peerDependencies": {
-        "@types/node": ">=18",
        "apache-arrow": ">=15.0.0 <=18.1.0"
-      },
-      "peerDependenciesMeta": {
-        "@types/node": {
-          "optional": true
-        }
      }
    },
    "node_modules/@aws-crypto/crc32": {
@@ -101,12 +101,6 @@
    "openai": "4.29.2"
  },
  "peerDependencies": {
-    "@types/node": ">=18",
    "apache-arrow": ">=15.0.0 <=18.1.0"
-  },
-  "peerDependenciesMeta": {
-    "@types/node": {
-      "optional": true
-    }
  }
 }
@@ -60,7 +60,7 @@ tests = [
    "pytest-asyncio>=0.21",
    "duckdb>=0.9.0",
    "pytz>=2023.3",
-    "polars>=0.19, <=1.32.3",
+    "polars>=0.19, <=1.3.0",
    "pyarrow<25",
    "pyarrow-stubs>=16.0",
    "pylance==9.0.0rc1",
@@ -140,7 +140,6 @@ include = [
    "python/lancedb/remote/errors.py",
    "python/lancedb/embeddings/__init__.py",
    "python/lancedb/_lancedb.pyi",
-    "python/type_tests/connect.py",
 ]
 exclude = ["python/tests/"]
 pythonVersion = "3.13"
@@ -183,11 +183,7 @@ class EmbeddingFunction(BaseModel, ABC):
    def VectorField(self, **kwargs):
        """
        Creates a pydantic Field that can automatically annotate
-        the target vector column for this embedding function.
-
-        The field can be annotated as ``list[float]`` for compatibility with
-        static type checkers. LanceDB will infer the fixed vector dimension from
-        this embedding function.
+        the target vector column for this embedding function
        """
        return Field(json_schema_extra={"vector_column_for": self}, **kwargs)

@@ -1 +0,0 @@
-
@@ -99,17 +99,6 @@ def Vector(
    ...     pa.field("url", pa.utf8(), False),
    ...     pa.field("embeddings", pa.list_(pa.float32(), 768))
    ... ])
-
-    Notes
-    -----
-    ``Vector`` creates a type dynamically, so calls such as ``Vector(768)`` are
-    not valid static type annotations. For an embedding field, use the standard
-    ``list[float]`` annotation when running mypy; ``VectorField`` supplies the
-    fixed dimension to LanceDB::
-
-        class MyModel(LanceModel):
-            text: str = embeddings.SourceField()
-            vector: list[float] = embeddings.VectorField()
    """

    # TODO: make a public parameterized type.
@@ -164,16 +153,6 @@ def Vector(
    return FixedSizeList


-def _raise_bare_vector_error(*_args):
-    raise TypeError("Vector must be parameterized with a dimension, e.g. Vector(128).")
-
-
-# Pydantic v1 and v2 otherwise treat the bare Vector factory as a field validator
-# and inspect its signature, which produces misleading errors about internal types.
-setattr(Vector, "__get_validators__", _raise_bare_vector_error)
-setattr(Vector, "__get_pydantic_core_schema__", _raise_bare_vector_error)
-
-
 def MultiVector(
    dim: int, value_type: pa.DataType = pa.float32(), nullable: bool = True
 ) -> Type:
@@ -390,10 +369,6 @@ def _unwrap_optional_annotation(annotation: Any) -> Any | None:

 def _pydantic_to_arrow_type(field: FieldInfo) -> pa.DataType:
    """Convert a Pydantic FieldInfo to Arrow DataType"""
-    embedding_vector_type = _embedding_vector_to_arrow_type(field)
-    if embedding_vector_type is not None:
-        return embedding_vector_type
-
    unwrapped = _unwrap_optional_annotation(field.annotation)
    if unwrapped is not None:
        return _pydantic_type_to_arrow_type(unwrapped, field)
@@ -407,32 +382,8 @@ def _pydantic_to_arrow_type(field: FieldInfo) -> pa.DataType:
    return _pydantic_type_to_arrow_type(field.annotation, field)


-def _embedding_vector_to_arrow_type(field: FieldInfo) -> pa.DataType | None:
-    """Infer a fixed-size vector type from ``VectorField`` metadata."""
-    if not _is_embedding_vector_annotation(field):
-        return None
-
-    function = get_extras(field, "vector_column_for")
-    return pa.list_(pa.float32(), function.ndims())
-
-
-def _is_embedding_vector_annotation(field: FieldInfo) -> bool:
-    if get_extras(field, "vector_column_for") is None:
-        return False
-
-    annotation = _unwrap_optional_annotation(field.annotation)
-    if annotation is None:
-        annotation = field.annotation
-
-    origin = getattr(annotation, "__origin__", None)
-    args = getattr(annotation, "__args__", ())
-    return origin is list and args == (float,)
-
-
 def is_nullable(field: FieldInfo) -> bool:
    """Check if a Pydantic FieldInfo is nullable."""
-    if _is_embedding_vector_annotation(field):
-        return True
    if _unwrap_optional_annotation(field.annotation) is not None:
        return True
    if isinstance(field.annotation, (_GenericAlias, GenericAlias)):
@@ -108,11 +108,6 @@ def _should_push_down_query_table(
    return namespace_client is not None and "QueryTable" in pushdown_operations


-def _polars_predicate_pushdown_barrier(frame: Any) -> Any:
-    """Return a Polars frame unchanged while blocking predicate pushdown."""
-    return frame
-
-
 _MODEL_BACKED_TOKENIZER_PREFIXES = ("jieba", "lindera")
 _MODEL_BACKED_TOKENIZER_ERRORS = (
    "unknown base tokenizer",
@@ -869,18 +864,12 @@ class Table(ABC):
        """
        raise NotImplementedError

-    def to_polars(self, **kwargs) -> "pl.LazyFrame":
-        """Return the table as a Polars LazyFrame.
-
-        Note
-        ----
-        The Polars streaming engine is not supported because it does not currently
-        implement Python PyArrow dataset scans. Use the default engine when collecting
-        this LazyFrame.
+    def to_polars(self, **kwargs) -> "pl.DataFrame":
+        """Return the table as a polars.DataFrame.

        Returns
        -------
-        polars.LazyFrame
+        polars.DataFrame
        """
        raise NotImplementedError

@@ -2580,9 +2569,6 @@ class LanceTable(Table):
        2. Currently we've disabled push-down of the filters from polars
           because polars pushdown into pyarrow uses pyarrow compute
           expressions rather than SQl strings (which LanceDB supports)
-        3. The Polars streaming engine is not supported because it does not
-           currently implement Python PyArrow dataset scans. Use the default
-           engine when collecting this LazyFrame.

        Returns
        -------
@@ -2591,12 +2577,8 @@ class LanceTable(Table):
        from lancedb.integrations.pyarrow import PyarrowDatasetAdapter

        dataset = PyarrowDatasetAdapter(self)
-        # Polars 1.32's non-PyArrow callback path passes batch_size twice.  Keep
-        # the compatible PyArrow path, but block predicates because this adapter
-        # cannot translate PyArrow expressions into LanceDB filters.
-        return pl.scan_pyarrow_dataset(dataset, batch_size=batch_size).map_batches(
-            _polars_predicate_pushdown_barrier,
-            predicate_pushdown=False,
+        return pl.scan_pyarrow_dataset(
+            dataset, allow_pyarrow_filter=False, batch_size=batch_size
        )

    # New unified API overload
@@ -6,7 +6,6 @@ import inspect
 import re
 import sys
 from datetime import timedelta
-from importlib import resources
 import os
 from types import SimpleNamespace

@@ -19,10 +18,6 @@ from lance_namespace.errors import NamespaceNotEmptyError, TableNotFoundError
 from lancedb.pydantic import LanceModel, Vector


-def test_package_includes_pep_561_marker():
-    assert resources.files(lancedb).joinpath("py.typed").is_file()
-
-
 def test_basic(tmp_path):
    db = lancedb.connect(tmp_path)

@@ -9,7 +9,6 @@ from typing import List, Optional, Tuple
 import pyarrow as pa
 import pydantic
 import pytest
-from lancedb.conftest import MockTextEmbeddingFunction
 from lancedb.pydantic import (
    PYDANTIC_VERSION,
    LanceModel,
@@ -416,36 +415,6 @@ def test_nullable_vector():
    assert schema == pa.schema([pa.field("vec", pa.list_(pa.float32(), 16), True)])


-def test_bare_vector_raises_clear_error():
-    namespace = {
-        "__name__": "test_model_without_pyarrow",
-        "LanceModel": LanceModel,
-        "Vector": Vector,
-    }
-
-    with pytest.raises(TypeError, match=r"Vector must be parameterized.*Vector\(128\)"):
-        exec("class TestModel(LanceModel):\n    vector: Vector", namespace)
-
-
-def test_embedding_vector_list_annotation():
-    embedding = MockTextEmbeddingFunction.create()
-
-    class StaticTypingModel(LanceModel):
-        text: str = embedding.SourceField()
-        vector: list[float] = embedding.VectorField()
-
-    schema = pydantic_to_schema(StaticTypingModel)
-    assert schema == pa.schema(
-        [
-            pa.field("text", pa.utf8(), False),
-            pa.field("vector", pa.list_(pa.float32(), embedding.ndims()), True),
-        ]
-    )
-
-    model = StaticTypingModel(text="hello", vector=[0.0] * embedding.ndims())
-    assert model.vector == [0.0] * embedding.ndims()
-
-
 def test_fixed_size_list_field():
    class TestModel(pydantic.BaseModel):
        vec: Vector(16)
@@ -570,15 +570,6 @@ def test_query_builder(table):
    assert all(np.array(rs[0]["vector"]) == [1, 2])


-def test_query_multiple_vectors(table):
-    results = table.search([np.array([1, 2]), np.array([4, 5])]).limit(1).to_list()
-
-    assert len(results) == 2
-    results_by_query = {result["query_index"]: result for result in results}
-    assert results_by_query[0]["id"] == 1
-    assert results_by_query[1]["id"] == 2
-
-
 def test_with_row_id(table: lancedb.table.Table):
    rs = table.search().with_row_id(True).to_arrow()
    assert "_rowid" in rs.column_names
@@ -929,7 +929,6 @@ def test_polars(mem_db: DBConnection):

    # enter table to polars dataframe
    result = table.to_polars()
-    assert isinstance(result, pl.LazyFrame)
    assert np.allclose(result.collect()["vector"].to_list(), data["vector"])

    # make sure filtering isn't broken
@@ -1846,27 +1845,6 @@ def test_add_with_empty_fixed_size_list_drops_bad_rows(mem_db: DBConnection):
    assert np.allclose(data["embedding"].to_pylist()[0], np.array([0.1] * 16))


-def test_add_nullable_fixed_size_list_with_none(mem_db: DBConnection):
-    """Regression test for issue #2340."""
-    table = mem_db.create_table(
-        "test_nullable_fixed_size_list",
-        schema=pa.schema(
-            [
-                pa.field("id", pa.string()),
-                pa.field("feature", pa.list_(pa.float32(), 256)),
-                pa.field("tags", pa.list_(pa.string())),
-            ]
-        ),
-    )
-
-    table.add([{"id": "1", "feature": None, "tags": ["tag1", "tag2"]}])
-
-    result = table.to_arrow()
-    assert result.to_pylist() == [
-        {"id": "1", "feature": None, "tags": ["tag1", "tag2"]}
-    ]
-
-
 def test_add_nullable_struct_with_none(mem_db: DBConnection):
    """Regression test for issue #2654: a nullable struct column whose
    first batch contains only None values must not crash in
@@ -2218,45 +2196,6 @@ def test_merge(tmp_db: DBConnection, tmp_path):
    table.merge(other_dataset, left_on="id")


-@pytest.mark.parametrize("storage_version", ["legacy", "stable"])
-def test_search_after_merge(tmp_path, storage_version):
-    pytest.importorskip("lance")
-    pd = pytest.importorskip("pandas")
-
-    db = lancedb.connect(
-        tmp_path,
-        storage_options={"new_table_data_storage_version": storage_version},
-    )
-    rng = np.random.default_rng(42)
-    row_count = 512
-    vectors = rng.standard_normal((row_count, 8)).astype(np.float32)
-    table = db.create_table(
-        "search_after_merge",
-        data=pd.DataFrame(
-            {
-                "id": [str(i) for i in range(row_count)],
-                "vector": list(vectors),
-            }
-        ),
-    )
-    table.create_index("vector", config=IvfPq(num_partitions=1, num_sub_vectors=2))
-
-    links = pd.DataFrame(
-        {
-            "id": [str(i) for i in range(row_count // 2)],
-            "link": [f"https://example.com/{i}" for i in range(row_count // 2)],
-        }
-    )
-    table.merge(links, left_on="id")
-
-    query = table.search(vectors[-1]).refine_factor(50).limit(10)
-    assert "ANN" in query.explain_plan(verbose=True)
-
-    result = query.to_arrow()
-    links_by_id = dict(zip(result["id"].to_pylist(), result["link"].to_pylist()))
-    assert links_by_id[str(row_count - 1)] is None
-
-
 def test_delete(mem_db: DBConnection):
    table = mem_db.create_table(
        "my_table",
@@ -2799,40 +2738,15 @@ def test_create_with_embedding_function(mem_db: DBConnection):
    assert actual == expected


-def test_create_f16_table_from_arrow_data(mem_db: DBConnection):
-    dimension = 32
-    num_rows = 512
-    values = pa.array(
-        np.random.default_rng(42)
-        .standard_normal(num_rows * dimension)
-        .astype(np.float16)
-    )
-    df = pa.table(
-        {
-            "text": [f"s-{i}" for i in range(num_rows)],
-            "vector": pa.FixedSizeListArray.from_arrays(values, dimension),
-        }
-    )
-    table = mem_db.create_table("f16_tbl", data=df)
-    assert table.schema.field("vector").type == pa.list_(pa.float16(), dimension)
-    table.create_index(num_partitions=2, num_sub_vectors=2)
-
-    query = df["vector"][2].as_py()
-    expected = table.search(query).limit(2).to_arrow()
-
-    assert "s-2" in expected["text"].to_pylist()
-
-
 def test_create_f16_table(mem_db: DBConnection):
    class MyTable(LanceModel):
        text: str
        vector: Vector(32, value_type=pa.float16())

-    rng = np.random.default_rng(42)
    df = pa.table(
        {
            "text": [f"s-{i}" for i in range(512)],
-            "vector": [rng.standard_normal(32).astype(np.float16) for _ in range(512)],
+            "vector": [np.random.randn(32).astype(np.float16) for _ in range(512)],
        }
    )
    table = mem_db.create_table(
@@ -1,15 +0,0 @@
-# SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright The LanceDB Authors
-
-from typing import assert_type
-
-import lancedb
-from lancedb import AsyncConnection, DBConnection
-
-
-def check_connect_type() -> None:
-    assert_type(lancedb.connect("memory://"), DBConnection)
-
-
-async def check_connect_async_type() -> None:
-    assert_type(await lancedb.connect_async("memory://"), AsyncConnection)
@@ -1998,7 +1998,7 @@ requires-dist = [
    { name = "pillow", marker = "extra == 'clip'", specifier = ">=12.1.1" },
    { name = "pillow", marker = "extra == 'embeddings'", specifier = ">=12.1.1" },
    { name = "pillow", marker = "extra == 'siglip'", specifier = ">=12.1.1" },
-    { name = "polars", marker = "extra == 'tests'", specifier = ">=0.19,<=1.32.3" },
+    { name = "polars", marker = "extra == 'tests'", specifier = ">=0.19,<=1.3.0" },
    { name = "pre-commit", marker = "extra == 'dev'", specifier = ">=3.5.0" },
    { name = "pyarrow", specifier = ">=16" },
    { name = "pyarrow", marker = "extra == 'tests'", specifier = "<25" },
@@ -49,8 +49,8 @@ lance-namespace = { workspace = true }
 lance-namespace-impls = { workspace = true }
 metrics = { workspace = true, optional = true }
 metrics-util = { workspace = true, optional = true }
-# Pin the GooseFS SDK to the version required by Lance's OpenDAL dependency.
-goosefs-sdk = { version = "=0.1.9", optional = true }
+# Pin the transitive GooseFS SDK until the 0.1.6 compile break is fixed upstream.
+goosefs-sdk = { version = "=0.1.5", optional = true }
 moka = { workspace = true }
 pin-project = { workspace = true }
 tokio = { version = "1.23", features = ["rt-multi-thread", "sync"] }
@@ -17,7 +17,7 @@ use arrow_array::builder::LargeBinaryBuilder;
 use arrow_schema::{DataType, Field, Schema};
 use lance::dataset::{BlobRangeRequest as LanceBlobRangeRequest, Dataset, WriteParams};
 use lance_arrow::FieldExt;
-use lance_file::version::LanceFileVersion;
+use lance_encoding::version::LanceFileVersion;
 use lance_io::object_store::ObjectStore;
 use object_store::path::Path;

@@ -34,7 +34,7 @@ use crate::remote::{
    db::{OPT_REMOTE_API_KEY, OPT_REMOTE_HOST_OVERRIDE, OPT_REMOTE_REGION},
 };
 use lance::io::ObjectStoreParams;
-pub use lance_file::version::LanceFileVersion;
+pub use lance_encoding::version::LanceFileVersion;
 #[cfg(feature = "remote")]
 use lance_io::object_store::StorageOptions;
 use lance_io::object_store::{StorageOptionsAccessor, StorageOptionsProvider};
@@ -12,7 +12,7 @@ use lance::dataset::refs::Ref;
 use lance::dataset::{ReadParams, WriteMode, builder::DatasetBuilder};
 use lance::io::{ObjectStore, ObjectStoreParams, WrappingObjectStore};
 use lance_datafusion::utils::StreamingWriteSource;
-use lance_file::version::LanceFileVersion;
+use lance_encoding::version::LanceFileVersion;
 use lance_io::object_store::{StorageOptionsAccessor, StorageOptionsProvider};
 use lance_table::io::commit::commit_handler_from_url;
 use object_store::local::LocalFileSystem;
@@ -1294,11 +1294,9 @@ mod tests {
    use crate::connection::ConnectRequest;
    use crate::data::scannable::Scannable;
    use crate::database::{CreateTableMode, CreateTableRequest};
-    use crate::query::QueryRequest;
-    use crate::table::{AnyQuery, WriteOptions};
+    use crate::table::WriteOptions;
    use arrow_array::{Int32Array, RecordBatch, StringArray};
    use arrow_schema::{DataType, Field, Schema};
-    use futures::TryStreamExt;
    use std::path::PathBuf;
    use tempfile::tempdir;

@@ -1440,94 +1438,6 @@ mod tests {
        assert!(after_open.hits >= before_open.hits + 3);
    }

-    /// Regression test for https://github.com/lancedb/lancedb/issues/3197.
-    #[cfg(unix)]
-    #[tokio::test]
-    async fn test_open_table_follows_hugging_face_symlinks() {
-        let (tempdir, db) = setup_database().await;
-        let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int32, false)]));
-        db.create_table(CreateTableRequest {
-            name: "test".to_string(),
-            namespace_path: vec![],
-            data: Box::new(
-                RecordBatch::try_new(schema, vec![Arc::new(Int32Array::from(vec![1, 2, 3]))])
-                    .unwrap(),
-            ) as Box<dyn Scannable>,
-            mode: CreateTableMode::Create,
-            write_options: Default::default(),
-            location: None,
-            namespace_client: None,
-        })
-        .await
-        .unwrap();
-
-        let table_dir = tempdir.path().join("test.lance");
-        let versions_dir = table_dir.join("_versions");
-        let manifest_path = std::fs::read_dir(&versions_dir)
-            .unwrap()
-            .map(|entry| entry.unwrap().path())
-            .find(|path| path.extension().is_some_and(|ext| ext == "manifest"))
-            .unwrap();
-        let data_path = std::fs::read_dir(table_dir.join("data"))
-            .unwrap()
-            .map(|entry| entry.unwrap().path())
-            .find(|path| path.extension().is_some_and(|ext| ext == "lance"))
-            .unwrap();
-
-        // Hugging Face snapshots keep dataset objects in a separate blob directory and
-        // expose them through relative symlinks.
-        let blobs_dir = tempdir.path().join("blobs");
-        std::fs::create_dir(&blobs_dir).unwrap();
-        let manifest_blob = "9b603c63d0e692e05d58be25605f2f2064cc781e5ff94fe983a405059547b816";
-        let data_blob = "be64f20e5723bd0a27cfdbdb41cf7d6fad94cd572a71973b717fb8340f4310c5";
-        std::fs::rename(&manifest_path, blobs_dir.join(manifest_blob)).unwrap();
-        std::fs::rename(&data_path, blobs_dir.join(data_blob)).unwrap();
-        std::os::unix::fs::symlink(Path::new("../../blobs").join(manifest_blob), &manifest_path)
-            .unwrap();
-        std::os::unix::fs::symlink(Path::new("../../blobs").join(data_blob), &data_path).unwrap();
-        let symlink_len = std::fs::symlink_metadata(&manifest_path).unwrap().len();
-        let target_len = std::fs::metadata(&manifest_path).unwrap().len();
-        assert_ne!(symlink_len, target_len);
-
-        drop(db);
-        let db = ListingDatabase::connect_with_options(&ConnectRequest {
-            uri: tempdir.path().to_str().unwrap().to_string(),
-            #[cfg(feature = "remote")]
-            client_config: Default::default(),
-            options: Default::default(),
-            namespace_client_properties: Default::default(),
-            manifest_enabled: false,
-            read_consistency_interval: None,
-            session: None,
-        })
-        .await
-        .unwrap();
-
-        let table = db
-            .open_table(OpenTableRequest {
-                name: "test".to_string(),
-                namespace_path: vec![],
-                index_cache_size: None,
-                lance_read_params: None,
-                location: None,
-                namespace_client: None,
-                managed_versioning: None,
-            })
-            .await
-            .unwrap();
-        let batches = table
-            .query(
-                &AnyQuery::Query(QueryRequest::default()),
-                Default::default(),
-            )
-            .await
-            .unwrap()
-            .try_collect::<Vec<_>>()
-            .await
-            .unwrap();
-        assert_eq!(batches.iter().map(RecordBatch::num_rows).sum::<usize>(), 3);
-    }
-
    #[tokio::test]
    async fn test_clone_table_basic() {
        let (_tempdir, db) = setup_database().await;
@@ -2432,7 +2342,7 @@ mod tests {

    #[tokio::test]
    async fn test_table_uri() {
-        let (_tempdir, mut db) = setup_database().await;
+        let (_tempdir, db) = setup_database().await;

        let mut pb = PathBuf::new();
        pb.push(db.uri.clone());
@@ -2441,18 +2351,6 @@ mod tests {
        let expected = pb.to_str().unwrap();
        let uri = db.table_uri("test").ok().unwrap();
        assert_eq!(uri, expected);
-
-        // URI paths always use forward slashes, even on Windows. Using
-        // `Path::join` here used to produce `az://container/prefix\\test.lance`,
-        // which Azure treated as a different object from the table returned by
-        // `table_names` (https://github.com/lancedb/lancedb/issues/1072).
-        for base_uri in ["az://container/prefix", "az://container/prefix/"] {
-            db.uri = base_uri.to_string();
-            assert_eq!(
-                db.table_uri("test").unwrap(),
-                "az://container/prefix/test.lance"
-            );
-        }
    }

    /// Regression: connecting via a URL-style URI (which goes through
@@ -201,7 +201,7 @@ impl LanceNamespaceDatabase {
        &self,
        request: &DbCreateTableRequest,
    ) -> Result<(
-        Option<lance_file::version::LanceFileVersion>,
+        Option<lance_encoding::version::LanceFileVersion>,
        Option<bool>,
        Option<bool>,
    )> {
@@ -214,7 +214,7 @@ impl LanceNamespaceDatabase {

        let storage_version_override = storage_options
            .and_then(|opts| opts.get(OPT_NEW_TABLE_STORAGE_VERSION))
-            .map(|s| s.parse::<lance_file::version::LanceFileVersion>())
+            .map(|s| s.parse::<lance_encoding::version::LanceFileVersion>())
            .transpose()?;

        let v2_manifest_override = storage_options
@@ -169,12 +169,6 @@ impl From<DataFusionError> for Error {

 impl From<lance::Error> for Error {
    fn from(source: lance::Error) -> Self {
-        if has_unsupported_local_filesystem_source(&source) {
-            return Self::NotSupported {
-                message: "the filesystem does not support an operation required for safe Lance commits (such as atomic rename). Object-storage mounts such as Mountpoint for Amazon S3 are not supported; use the native object-store URI (for example, s3://bucket/path) instead".to_string(),
-            };
-        }
-
        // Try to unwrap external errors that were wrapped by lance
        match source {
            lance::Error::Wrapped { error, .. } => Self::from_box_error(error),
@@ -187,27 +181,6 @@ impl From<lance::Error> for Error {
    }
 }

-fn has_unsupported_local_filesystem_source(error: &(dyn std::error::Error + 'static)) -> bool {
-    let mut current = Some(error);
-    let mut is_local_filesystem = false;
-    let mut is_unsupported = false;
-    while let Some(error) = current {
-        is_local_filesystem |= error
-            .downcast_ref::<object_store::Error>()
-            .is_some_and(|error| {
-                matches!(error, object_store::Error::Generic { store, .. } if *store == "LocalFileSystem")
-            });
-        is_unsupported |= error
-            .downcast_ref::<std::io::Error>()
-            .is_some_and(|error| error.kind() == std::io::ErrorKind::Unsupported);
-        if is_local_filesystem && is_unsupported {
-            return true;
-        }
-        current = error.source();
-    }
-    false
-}
-
 impl Error {
    fn from_box_error(mut source: Box<dyn std::error::Error + Send + Sync>) -> Self {
        source = match source.downcast::<Self>() {
@@ -297,46 +270,3 @@ impl From<candle_core::Error> for Error {
        }
    }
 }
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-
-    #[test]
-    fn unsupported_filesystem_operations_have_actionable_error() {
-        let object_store_error = object_store::Error::Generic {
-            store: "LocalFileSystem",
-            source: Box::new(std::io::Error::from(std::io::ErrorKind::Unsupported)),
-        };
-        let lance_error = lance::Error::io_source(Box::new(object_store_error));
-
-        let error = Error::from(lance_error);
-
-        assert!(matches!(
-            error,
-            Error::NotSupported { message }
-                if message.contains("Mountpoint for Amazon S3")
-                    && message.contains("s3://bucket/path")
-        ));
-    }
-
-    #[test]
-    fn other_io_errors_remain_lance_errors() {
-        let object_store_error = object_store::Error::Generic {
-            store: "LocalFileSystem",
-            source: Box::new(std::io::Error::from(std::io::ErrorKind::PermissionDenied)),
-        };
-        let lance_error = lance::Error::io_source(Box::new(object_store_error));
-
-        assert!(matches!(Error::from(lance_error), Error::Lance { .. }));
-    }
-
-    #[test]
-    fn unsupported_non_filesystem_errors_remain_lance_errors() {
-        let lance_error = lance::Error::io_source(Box::new(std::io::Error::from(
-            std::io::ErrorKind::Unsupported,
-        )));
-
-        assert!(matches!(Error::from(lance_error), Error::Lance { .. }));
-    }
-}
@@ -2942,7 +2942,7 @@ impl<S: HttpSend> BaseTable for RemoteTable<S> {
 }

 #[derive(Serialize, Clone, Debug)]
-pub struct MergeInsertRequest {
+pub(crate) struct MergeInsertRequest {
    on: String,
    when_matched_update_all: bool,
    when_matched_update_all_filt: Option<String>,
@@ -5907,18 +5907,16 @@ mod tests {
            .await
            .unwrap();

-        // Positions are relative to the first retained token, so dropping the
-        // leading "hello" stop word does not shift the remaining tokens.
        assert_eq!(
            tokens,
            vec![
                FtsToken {
                    text: "こんにちは".to_string(),
-                    position: 0,
+                    position: 1,
                },
                FtsToken {
                    text: "世界".to_string(),
-                    position: 1,
+                    position: 2,
                },
            ]
        );
@@ -90,7 +90,7 @@ struct RemoteBlobState {

 /// Seekable Cloud blob handle over HTTP Range.
 #[derive(Debug)]
-pub struct RemoteBlobFile {
+pub(crate) struct RemoteBlobFile {
    requester: Arc<dyn BlobRangeRequester>,
    state: Mutex<RemoteBlobState>,
    closed: AtomicBool,
@@ -33,7 +33,7 @@ use crate::table::{AddResult, MergeResult};
 /// same Arrow-IPC streaming body and error side-channel; only the target
 /// endpoint, query parameters, and parsed result type differ.
 #[derive(Debug, Clone)]
-pub enum WriteOp {
+pub(crate) enum WriteOp {
    /// `add`: stream to `/v1/table/{id}/insert/`, optionally overwriting.
    Insert { overwrite: bool },
    /// `merge_insert`: stream to `/v1/table/{id}/merge_insert/` with the merge
@@ -49,7 +49,7 @@ pub enum WriteOp {
 /// The parsed server response for a completed write, discriminated by the
 /// operation that produced it.
 #[derive(Debug, Clone)]
-pub enum WriteResult {
+pub(crate) enum WriteResult {
    Add(AddResult),
    Merge(MergeResult),
 }
@@ -6,19 +6,162 @@
 //! This module contains the implementation of optimization operations that help
 //! maintain good performance for LanceDB tables.

-use std::sync::Arc;
+use std::{collections::HashSet, sync::Arc};

+use arrow_schema::DataType;
 use lance::dataset::cleanup::RemovalStats;
-use lance::dataset::optimize::{CompactionMetrics, IndexRemapperOptions, compact_files};
-use lance::index::DatasetIndexExt;
+use lance::dataset::optimize::{
+    CompactionMetrics, CompactionPlan, CompactionPlanner, IndexRemapperOptions,
+    compact_files_with_planner, plan_compaction,
+};
+use lance::index::{DatasetIndexExt, DatasetIndexInternalExt};
+use lance_index::IndexType;
+use lance_index::metrics::NoOpMetricsCollector;
 use lance_index::optimize::OptimizeOptions;
-use log::info;
+use log::{debug, info};

+use super::NativeTable;
+use crate::error::{Error, Result};
 pub use chrono::Duration;
 pub use lance::dataset::optimize::CompactionOptions;

-use super::NativeTable;
-use crate::error::Result;
+const MAX_ARROW_FIXED_SIZE_LIST_CHILD_INDEX: u64 = u32::MAX as u64;
+
+struct PrecomputedCompactionPlanner {
+    plan: CompactionPlan,
+}
+
+#[async_trait::async_trait]
+impl CompactionPlanner for PrecomputedCompactionPlanner {
+    async fn plan(&self, dataset: &lance::Dataset) -> lance::Result<CompactionPlan> {
+        debug_assert_eq!(dataset.manifest().version, self.plan.read_version());
+        Ok(self.plan.clone())
+    }
+}
+
+fn sq_vector_dimension(data_type: &DataType) -> Option<i32> {
+    match data_type {
+        DataType::FixedSizeList(_, dimension) => Some(*dimension),
+        DataType::List(field) => match field.data_type() {
+            DataType::FixedSizeList(_, dimension) => Some(*dimension),
+            _ => None,
+        },
+        _ => None,
+    }
+}
+
+fn segment_touches_fragments(
+    fragment_ids: Option<impl IntoIterator<Item = u32>>,
+    affected_fragments: &HashSet<u64>,
+) -> bool {
+    fragment_ids.is_none_or(|fragment_ids| {
+        fragment_ids
+            .into_iter()
+            .any(|fragment| affected_fragments.contains(&(fragment as u64)))
+    })
+}
+
+fn oversized_sq_partition(
+    partition_sizes: impl IntoIterator<Item = u64>,
+    dimension: u64,
+) -> Option<u64> {
+    partition_sizes.into_iter().find(|partition_size| {
+        partition_size
+            .checked_mul(dimension)
+            .is_none_or(|child_len| child_len > MAX_ARROW_FIXED_SIZE_LIST_CHILD_INDEX)
+    })
+}
+
+async fn validate_sq_index_remapping(
+    dataset: &lance::Dataset,
+    options: &CompactionOptions,
+    has_custom_remapper: bool,
+) -> Result<CompactionPlan> {
+    let plan = plan_compaction(dataset, options).await?;
+    if options.defer_index_remap || has_custom_remapper || dataset.manifest().uses_stable_row_ids()
+    {
+        return Ok(plan);
+    }
+
+    if plan.tasks.is_empty() {
+        return Ok(plan);
+    }
+    let affected_fragments: HashSet<u64> = plan
+        .tasks
+        .iter()
+        .flat_map(|task| task.fragments.iter().map(|fragment| fragment.id))
+        .collect();
+
+    for segment in dataset.load_indices().await?.iter() {
+        if !segment_touches_fragments(
+            segment
+                .fragment_bitmap
+                .as_ref()
+                .map(|fragment_bitmap| fragment_bitmap.iter()),
+            &affected_fragments,
+        ) {
+            continue;
+        }
+        let Some(field_id) = segment.fields.first() else {
+            continue;
+        };
+        let Some(field) = dataset.schema().field_by_id(*field_id) else {
+            continue;
+        };
+        let data_type = field.data_type();
+        let Some(dimension) = sq_vector_dimension(&data_type) else {
+            continue;
+        };
+        let dimension = u64::try_from(dimension).map_err(|_| Error::InvalidInput {
+            message: format!(
+                "SQ index '{}' has an invalid vector dimension of {}",
+                segment.name, dimension
+            ),
+        })?;
+
+        let field_path = dataset.schema().field_path(*field_id)?;
+        let vector_index = match dataset
+            .open_vector_index(&field_path, &segment.uuid, &NoOpMetricsCollector)
+            .await
+        {
+            Ok(vector_index) => vector_index,
+            Err(error) => {
+                // The default remapper also treats an index it cannot open as
+                // unusable and drops it, so it cannot reach the SQ take kernel.
+                debug!(
+                    "Skipping remap preflight for index segment {} because it could not be opened: {}",
+                    segment.uuid, error
+                );
+                continue;
+            }
+        };
+        if !matches!(
+            vector_index.index_type(),
+            IndexType::IvfSq | IndexType::IvfHnswSq
+        ) {
+            continue;
+        }
+
+        let partition_sizes = (0..vector_index.total_partitions())
+            .map(|partition_id| vector_index.partition_size(partition_id) as u64);
+        if let Some(partition_size) = oversized_sq_partition(partition_sizes, dimension) {
+            let max_partition_size = MAX_ARROW_FIXED_SIZE_LIST_CHILD_INDEX / dimension;
+            return Err(Error::InvalidInput {
+                message: format!(
+                    "Cannot compact table because SQ index '{}' segment {} has a partition with {} vectors of dimension {}. Arrow's fixed-size-list take kernel cannot remap partitions whose child array exceeds {} values. Recreate the index with default IVF partitioning, or enough partitions to keep each partition at or below {} vectors, before compacting.",
+                    segment.name,
+                    segment.uuid,
+                    partition_size,
+                    dimension,
+                    MAX_ARROW_FIXED_SIZE_LIST_CHILD_INDEX,
+                    max_partition_size,
+                ),
+            });
+        }
+    }
+
+    Ok(plan)
+}

 /// Optimize the dataset.
 ///
@@ -144,7 +287,7 @@ pub(crate) async fn cleanup_old_versions(
 /// This can be run after making several small appends to optimize the table
 /// for faster reads.
 ///
-/// This calls into [lance::dataset::optimize::compact_files].
+/// This calls into [lance::dataset::optimize::compact_files_with_planner].
 pub(crate) async fn compact_files_impl(
    table: &NativeTable,
    options: CompactionOptions,
@@ -152,7 +295,9 @@ pub(crate) async fn compact_files_impl(
 ) -> Result<CompactionMetrics> {
    table.dataset.ensure_mutable()?;
    let mut dataset = (*table.dataset.get().await?).clone();
-    let metrics = compact_files(&mut dataset, options, remap_options).await?;
+    let plan = validate_sq_index_remapping(&dataset, &options, remap_options.is_some()).await?;
+    let planner = PrecomputedCompactionPlanner { plan };
+    let metrics = compact_files_with_planner(&mut dataset, remap_options, &planner).await?;
    table.dataset.update(dataset);
    Ok(metrics)
 }
@@ -214,13 +359,14 @@ pub(crate) async fn execute_optimize(

 #[cfg(test)]
 mod tests {
+    use super::{oversized_sq_partition, segment_touches_fragments, sq_vector_dimension};
    use arrow_array::{
        Array, FixedSizeListArray, Float32Array, Int32Array, RecordBatch, StringArray,
    };
    use arrow_schema::{DataType, Field, Schema};
    use lance_arrow::FixedSizeListArrayExt;
    use rstest::rstest;
-    use std::sync::Arc;
+    use std::{collections::HashSet, sync::Arc};

    use crate::connect;
    use crate::database::listing::OPT_NEW_TABLE_ENABLE_STABLE_ROW_IDS;
@@ -230,6 +376,172 @@ mod tests {
    use crate::table::{CompactionOptions, OptimizeAction, OptimizeStats};
    use futures::TryStreamExt;

+    /// Regression test for https://github.com/lancedb/lancedb/issues/2866.
+    #[test]
+    fn test_detect_oversized_sq_partition() {
+        const DIMENSION: u64 = 4095;
+        let safe_size = u32::MAX as u64 / DIMENSION;
+
+        assert_eq!(
+            oversized_sq_partition([safe_size, safe_size + 1], DIMENSION),
+            Some(safe_size + 1)
+        );
+        assert_eq!(
+            oversized_sq_partition([safe_size / 2, safe_size / 2 + 1], DIMENSION),
+            None
+        );
+    }
+
+    #[test]
+    fn test_segment_touches_fragments_includes_unknown_coverage() {
+        let affected_fragments = HashSet::from([7]);
+
+        assert!(segment_touches_fragments(
+            Some([7_u32]),
+            &affected_fragments
+        ));
+        assert!(!segment_touches_fragments(
+            Some([8_u32]),
+            &affected_fragments
+        ));
+        assert!(segment_touches_fragments(
+            None::<[u32; 0]>,
+            &affected_fragments
+        ));
+    }
+
+    #[test]
+    fn test_sq_vector_dimension_includes_multivectors() {
+        const DIMENSION: i32 = 4095;
+        let vector = DataType::FixedSizeList(
+            Arc::new(Field::new("item", DataType::Float32, false)),
+            DIMENSION,
+        );
+        let multivector = DataType::List(Arc::new(Field::new("item", vector.clone(), false)));
+
+        assert_eq!(sq_vector_dimension(&vector), Some(DIMENSION));
+        assert_eq!(sq_vector_dimension(&multivector), Some(DIMENSION));
+        assert_eq!(sq_vector_dimension(&DataType::Float32), None);
+    }
+
+    #[tokio::test]
+    async fn test_compact_legacy_vector_index_with_unknown_fragment_coverage() {
+        use lance::Dataset;
+        use lance::index::vector::{IndexFileVersion, VectorIndexParams};
+        use lance::index::{DatasetIndexExt, DatasetIndexInternalExt};
+        use lance_index::IndexType;
+        use lance_linalg::distance::MetricType;
+        use lance_table::io::commit::write_manifest_file_to_path;
+        use object_store::ObjectStoreExt;
+
+        const INDEX_NAME: &str = "legacy_ivf_pq";
+        const ROWS: i32 = 128;
+        const DIMENSION: i32 = 8;
+
+        let tmpdir = tempfile::tempdir().unwrap();
+        let conn = connect(tmpdir.path().to_str().unwrap())
+            .execute()
+            .await
+            .unwrap();
+        let vectors = FixedSizeListArray::try_new_from_values(
+            Float32Array::from_iter_values((0..ROWS).flat_map(|row| {
+                (0..DIMENSION).map(move |offset| (row * DIMENSION + offset) as f32)
+            })),
+            DIMENSION,
+        )
+        .unwrap();
+        let schema = Arc::new(Schema::new(vec![Field::new(
+            "vector",
+            vectors.data_type().clone(),
+            false,
+        )]));
+        let batch = RecordBatch::try_new(schema, vec![Arc::new(vectors)]).unwrap();
+        let table = conn
+            .create_table("test_legacy_vector_compact", batch.clone())
+            .execute()
+            .await
+            .unwrap();
+        table.add(batch.clone()).execute().await.unwrap();
+        table.add(batch).execute().await.unwrap();
+        let mut dataset = (*table.dataset().unwrap().get().await.unwrap()).clone();
+        let mut params = VectorIndexParams::ivf_pq(1, 8, 1, MetricType::L2, 10);
+        params.version(IndexFileVersion::Legacy);
+        dataset
+            .create_index(
+                &["vector"],
+                IndexType::Vector,
+                Some(INDEX_NAME.to_string()),
+                &params,
+                true,
+            )
+            .await
+            .unwrap();
+        table.dataset().unwrap().update(dataset);
+
+        let dataset = table.dataset().unwrap().get().await.unwrap();
+        let mut segments = dataset.load_indices_by_name(INDEX_NAME).await.unwrap();
+        assert_eq!(segments.len(), 1);
+        let original = segments.pop().unwrap();
+        let field_path = dataset.schema().field_path(original.fields[0]).unwrap();
+        let vector_index = dataset
+            .open_vector_index(
+                &field_path,
+                &original.uuid,
+                &lance_index::metrics::NoOpMetricsCollector,
+            )
+            .await
+            .unwrap();
+        assert_eq!(vector_index.index_type(), IndexType::IvfPq);
+        assert_eq!(
+            vector_index.statistics().unwrap()["index_file_version"],
+            "Legacy"
+        );
+        let mut legacy = original.clone();
+        legacy.fragment_bitmap = None;
+        let object_store = dataset.object_store(None).await.unwrap();
+        let mut manifest = dataset.manifest().clone();
+        manifest.index_section = None;
+        manifest.transaction_section = None;
+        let manifest_path = dataset.manifest_location().path.clone();
+        object_store.inner.delete(&manifest_path).await.unwrap();
+        write_manifest_file_to_path(
+            object_store.as_ref(),
+            &mut manifest,
+            Some(vec![legacy]),
+            &manifest_path,
+            None,
+        )
+        .await
+        .unwrap();
+        let legacy_dataset = Dataset::open(dataset.uri()).await.unwrap();
+        table.dataset().unwrap().update(legacy_dataset);
+
+        let dataset = table.dataset().unwrap().get().await.unwrap();
+        let description_error = dataset
+            .describe_indices(None)
+            .await
+            .err()
+            .expect("legacy coverage should be unknown to index descriptions");
+        assert!(
+            description_error
+                .to_string()
+                .contains("Fragment bitmap is required")
+        );
+
+        let stats = table
+            .optimize(OptimizeAction::Compact {
+                options: CompactionOptions {
+                    target_rows_per_fragment: 1_000,
+                    ..Default::default()
+                },
+                remap_options: None,
+            })
+            .await
+            .unwrap();
+        assert!(stats.compaction.unwrap().fragments_removed > 0);
+        assert_eq!(table.count_rows(None).await.unwrap(), ROWS as usize * 3);
+    }
+
    #[tokio::test]
    async fn test_optimize_compact_simple() {
        let conn = connect("memory://").execute().await.unwrap();
@@ -10,7 +10,7 @@ use arrow_array::{
 use arrow_schema::{DataType, Field, Fields, Schema};
 use futures::TryStreamExt;
 use lance::Dataset;
-use lance_file::version::LanceFileVersion;
+use lance_encoding::version::LanceFileVersion;
 use lancedb::{
    Connection, Error, Result, Table,
    blob::{BlobRangeRequest, blob},
Author	SHA1	Message	Date
Gatefixer	74c003af59	fix(rust): reuse SQ preflight compaction plan	2026-08-06 11:30:18 +00:00
Gatefixer	695e7ae956	fix(rust): safely classify legacy vector indexes	2026-08-06 11:13:04 +00:00
Gatefixer	67532dc4ce	test(rust): use valid legacy compaction fixture	2026-08-06 10:31:30 +00:00
Gatefixer	04836120d4	fix(rust): preserve legacy index remap coverage	2026-08-06 10:12:16 +00:00
Gatefixer	160b72bbd9	Merge remote-tracking branch 'origin/main' into gatekeeper/fix-2866-1 # Conflicts: # rust/lancedb/src/table/optimize.rs	2026-08-06 09:14:00 +00:00
Gatefixer	53c6495a32	fix(rust): guard oversized SQ index remaps	2026-08-06 04:14:40 +00:00