Install Cargo tools

.github/actions/setup-prebuild/action.yml

@@ -39,11 +39,18 @@ runs:
 
     # Fallback path: full setup for forks
     - name: Full Rust setup
-      if: github.repository != 'vortex-data/vortex'
+      if: github.repository == 'massive-com/vortex'
       uses: ./.github/actions/setup-rust
       with:
         repo-token: ${{ inputs.repo-token }}
         toolchain: ${{ inputs.toolchain }}
         components: ${{ inputs.components }}
         targets: ${{ inputs.targets }}
         enable-sccache: ${{ inputs.enable-sccache }}
+
+    # uv is preinstalled on the prebuilt images; install it for forks.
+    - name: Install uv
+      if: github.repository == 'massive-com/vortex'
+      uses: spiraldb/actions/.github/actions/setup-uv@a746510eafaa926484c354541cfc49b2ec06cc63  # 0.18.6
+      with:
+        sync: false

.github/actions/setup-rust/action.yml

@@ -80,6 +80,11 @@ runs:
       shell: bash
       run: sccache --start-server
 
+    - name: Install cargo tools (cargo-hack, cargo-nextest, cargo-minimal-versions)
+      uses: taiki-e/install-action@9e1e5806d4a4822de933115878265be9aaa786d9  # v2
+      with:
+        tool: cargo-hack,cargo-nextest,cargo-minimal-versions
+
     - name: Install Protoc (for lance-encoding build step)
       if: runner.os != 'Windows'
       uses: ./.github/actions/setup-protoc

.github/workflows/ci.yml

@@ -108,16 +108,23 @@ jobs:
           sccache: s3
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7
       - uses: ./.github/actions/setup-prebuild
+      # doxygen is preinstalled on the prebuilt images; install it for forks.
+      - name: Install Doxygen
+        if: github.repository == 'massive-com/vortex'
+        uses: ssciwr/doxygen-install@329d88f5a303066a5bd006db7516b1925b86350e  # v2
 
       - name: Pytest - Vortex
         run: |
           uv run --all-packages pytest --benchmark-disable -n auto test/
         working-directory: vortex-python/
 
+      # Benchmark tuning requires the self-hosted prebuilt runner; skip on forks.
       - name: Setup benchmark environment
+        if: github.repository == 'vortex-data/vortex'
         run: sudo bash scripts/setup-benchmark.sh
 
       - name: Pytest Benchmarks - Vortex
+        if: github.repository == 'vortex-data/vortex'
         run: |
           bash ../scripts/bench-taskset.sh uv run --all-packages pytest --benchmark-only benchmark/
         working-directory: vortex-python/
@@ -193,7 +200,8 @@ jobs:
 
   build-rust:
     name: "Rust build (${{matrix.config.name}})"
-    timeout-minutes: 30
+    # Forks build without sccache, so give the uncached build more headroom.
+    timeout-minutes: ${{ github.repository == 'vortex-data/vortex' && 30 || 60 }}
     runs-on: >-
       ${{ github.repository == 'vortex-data/vortex'
           && format('runs-on={0}/runner={1}/image=ubuntu24-full-x64-pre-v2/tag={2}', github.run_id, matrix.config.runner, matrix.config.name)
@@ -533,6 +541,9 @@ jobs:
           sccache: s3
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0  # v7
       - uses: ./.github/actions/setup-prebuild
+      # flatc is preinstalled on the prebuilt images; install it for forks.
+      - uses: ./.github/actions/setup-flatc
+        if: github.repository == 'massive-com/vortex'
       - name: Install nightly for cbindgen macro expansion
         run: rustup toolchain install $NIGHTLY_TOOLCHAIN
       - name: "regenerate all .fbs/.proto Rust code"

.github/workflows/codspeed.yml

@@ -24,6 +24,8 @@ env:
 
 jobs:
   bench-codspeed:
+    # CodSpeed requires an upstream account/token; skip on forks.
+    if: github.repository == 'vortex-data/vortex'
     strategy:
       matrix:
         include:

.github/workflows/publish-dry-runs.yml

@@ -24,6 +24,9 @@ env:
 jobs:
   python-wheel-build:
     name: "Python (wheel build)"
+    # Publish validation is upstream-only (git deps are unpublishable, and some
+    # jobs require the upstream self-hosted runners); skip on forks.
+    if: github.repository == 'vortex-data/vortex'
     runs-on: ubuntu-latest
     timeout-minutes: 30
     steps:
@@ -64,6 +67,7 @@ jobs:
         working-directory: vortex-python/
 
   check-java-publish-build:
+    if: github.repository == 'vortex-data/vortex'
     timeout-minutes: 30
     strategy:
       fail-fast: false
@@ -96,12 +100,14 @@ jobs:
 
   compat-check:
     name: "Compat check"
+    if: github.repository == 'vortex-data/vortex'
     uses: ./.github/workflows/compat-validation.yml
     with:
       mode: last
 
   rust-publish-dry-run:
     name: "Rust publish dry-run"
+    if: github.repository == 'vortex-data/vortex'
     timeout-minutes: 30
     runs-on: >-
       ${{ github.repository == 'vortex-data/vortex'

.github/workflows/rust-instrumented.yml

@@ -43,6 +43,10 @@ jobs:
 
   rust-coverage:
     name: "Rust tests (coverage) (${{ matrix.suite }})"
+    # Coverage instruments and runs the whole test suite, which OOMs the
+    # GitHub-hosted runners forks use (upstream runs it on larger self-hosted
+    # runners); it also uploads to Codecov via OIDC that forks lack. Skip on forks.
+    if: github.repository == 'vortex-data/vortex'
     needs: duckdb-ready
     timeout-minutes: 30
     permissions:
@@ -130,7 +134,8 @@ jobs:
       ${{ github.repository == 'vortex-data/vortex'
           && format('runs-on={0}/pool=amd64-medium-pre-v2/tag=rust-test-sanitizer', github.run_id)
           || 'ubuntu-latest' }}
-    timeout-minutes: 30
+    # Forks build instrumented without sccache; give the uncached run more headroom.
+    timeout-minutes: ${{ github.repository == 'vortex-data/vortex' && 30 || 90 }}
     env:
       ASAN_OPTIONS: "symbolize=1:check_initialization_order=1:detect_leaks=1:leak_check_at_exit=1"
       LSAN_OPTIONS: "report_objects=1"
@@ -154,6 +159,27 @@ jobs:
         run: |
           rustup toolchain install $NIGHTLY_TOOLCHAIN
           rustup component add --toolchain $NIGHTLY_TOOLCHAIN rust-src rustfmt clippy llvm-tools-preview
+      # The prebuilt image ships /usr/bin/llvm-symbolizer; GitHub-hosted runners
+      # do not. Without it TSan/ASan cannot symbolize frames, so the
+      # tsan_suppressions.txt entries (benign crossbeam/oneshot fence races)
+      # never match and fail the job. Symlink an available llvm-symbolizer.
+      - name: Ensure llvm-symbolizer (forks)
+        if: github.repository == 'massive-com/vortex'
+        shell: bash
+        run: |
+          if [ ! -x /usr/bin/llvm-symbolizer ]; then
+            sym="$(command -v llvm-symbolizer \
+              || ls -1 /usr/bin/llvm-symbolizer-* 2>/dev/null | sort -V | tail -1 \
+              || ls -1 /usr/lib/llvm-*/bin/llvm-symbolizer 2>/dev/null | sort -V | tail -1)"
+            if [ -z "$sym" ]; then
+              sudo apt-get update
+              sudo apt-get install -y llvm
+              sym="$(ls -1 /usr/bin/llvm-symbolizer-* 2>/dev/null | sort -V | tail -1 \
+                || ls -1 /usr/lib/llvm-*/bin/llvm-symbolizer 2>/dev/null | sort -V | tail -1)"
+            fi
+            sudo ln -sf "$sym" /usr/bin/llvm-symbolizer
+          fi
+          /usr/bin/llvm-symbolizer --version
       - name: Build tests with sanitizer
         run: |
           RUSTFLAGS="${RUSTFLAGS} ${{ matrix.sanitizer_flags }}" \

Cargo.toml

@@ -137,7 +137,9 @@ cudarc = { version = "0.19.0", features = [
 ] }
 custom-labels = "0.4.4"
 dashmap = "6.1.0"
-datafusion = { git = "https://github.com/massive-com/arrow-datafusion", rev = "6363fbeb2a593e78fb6d73ad07ddba13c0406949", default-features = false, features = ["sql"] }
+datafusion = { git = "https://github.com/massive-com/arrow-datafusion", rev = "6363fbeb2a593e78fb6d73ad07ddba13c0406949", default-features = false, features = [
+    "sql",
+] }
 datafusion-catalog = { git = "https://github.com/massive-com/arrow-datafusion", rev = "6363fbeb2a593e78fb6d73ad07ddba13c0406949" }
 datafusion-common = { git = "https://github.com/massive-com/arrow-datafusion", rev = "6363fbeb2a593e78fb6d73ad07ddba13c0406949" }
 datafusion-common-runtime = { git = "https://github.com/massive-com/arrow-datafusion", rev = "6363fbeb2a593e78fb6d73ad07ddba13c0406949" }

deny.toml

@@ -16,7 +16,10 @@ ignore = [
     # Paste is no longer maintained because its essentially "finished".
     "RUSTSEC-2024-0436",
     # proc-macro-error-2 is unmaintained, only used by the `test_with` test dependency
-    "RUSTSEC-2026-0173"
+    "RUSTSEC-2026-0173",
+    # quick-xml is a transitive dependency that we cannot upgrade ourselves
+    "RUSTSEC-2026-0195",
+    "RUSTSEC-2026-0194"
 ]
 
 [licenses]

vortex-array/src/array/erased.rs

@@ -265,7 +265,7 @@ impl ArrayRef {
             .optimize()
     }
 
-    /// Wraps the array in a [`ReversedArray`](crate::arrays::ReversedArray) such that it is
+    /// Wraps the array in a [`ReversedArray`] such that it is
     /// logically yielded in reverse element order. The optimizer is run immediately, which may
     /// eliminate the wrapper (for example, for `Reversed(Reversed(x))` or `Reversed(Dict(...))`).
     pub fn reverse(&self) -> VortexResult<ArrayRef> {

vortex-array/src/arrays/patched/mod.rs

@@ -43,7 +43,7 @@
 //! `indices` and `values` are aligned and accessed together.
 //!
 //! ```text
-//!
+//! 
 //!                  chunk 0      chunk 0      chunk 0     chunk 0       chunk 0     chunk 0
 //!                  lane  0      lane 1       lane  2     lane 3        lane  4     lane  5
 //!              ┌────────────┬────────────┬────────────┬────────────┬────────────┬────────────┐

vortex-array/src/arrays/reversed/mod.rs

@@ -15,7 +15,7 @@
 //!   dictionary-encoded, so the per-chunk reversal cost is O(n_codes) rather than O(n_rows).
 //!
 //! For encodings that have no reduce rule the `ReversedArray` wrapper survives to
-//! decode time, where [`execute.rs`](self::execute) reverses the canonical form
+//! decode time, where `execute.rs` reverses the canonical form
 //! directly:
 //!
 //! * `Primitive`: iterates the typed buffer backwards — O(n), fully sequential.