Remove the unstable encodings feature

.github/workflows/ci.yml

@@ -517,21 +517,15 @@ jobs:
       - uses: ./.github/actions/setup-prebuild
         with:
           enable-sccache: "true"
-      # The golden suite pins the compressor's decisions per feature variant, and the
-      # variants are mutually exclusive at compile time: the `default` variant only exists
-      # without `unstable_encodings` (which changes ALL_SCHEMES), so the `--all-features`
-      # workspace test jobs cannot run it. Run each feature combination explicitly.
-      - name: Golden corpus (default features)
+      # The golden suite pins compressor decisions for the core-only and explicitly enabled
+      # edition configurations. Compact compression additionally needs its optional schemes.
+      - name: Golden corpus (edition configurations)
         run: |
           cargo nextest run --cargo-profile ci --locked --no-fail-fast -p vortex-btrblocks --test golden
-      - name: Golden corpus (unstable_encodings)
+      - name: Golden corpus (compact, zstd + pco)
         run: |
           cargo nextest run --cargo-profile ci --locked --no-fail-fast -p vortex-btrblocks --test golden \
-            --features unstable_encodings
-      - name: Golden corpus (compact, unstable_encodings + zstd + pco)
-        run: |
-          cargo nextest run --cargo-profile ci --locked --no-fail-fast -p vortex-btrblocks --test golden \
-            --features unstable_encodings,zstd,pco
+            --features zstd,pco -E 'test(golden_compact)'
 
   build-java:
     name: "Java"
@@ -729,6 +723,9 @@ jobs:
         with:
           sccache: s3
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1  # v7
+        with:
+          # check-editions compares against the merge base, so it needs real history.
+          fetch-depth: 0
       - uses: ./.github/actions/setup-prebuild
         with:
           enable-sccache: "true"
@@ -738,6 +735,16 @@ jobs:
         run: |
           cargo run --profile ci -p xtask -- generate-fbs
           cargo run --profile ci -p xtask -- generate-proto
+      - name: "regenerate the edition records"
+        run: |
+          cargo run --profile ci -p xtask -- generate-editions
+      - name: "check frozen edition records never change"
+        # Independent of the regeneration above: a stale record must not mask a frozen one
+        # being edited, nor the other way round.
+        if: "!cancelled()"
+        run: |
+          BASE="${{ github.event.pull_request.base.sha || 'HEAD^' }}"
+          cargo run --profile ci -p xtask -- check-editions --base "$BASE"
       - name: "regenerate FFI header file"
         run: |
           cargo +$NIGHTLY_TOOLCHAIN build --profile ci -p vortex-ffi

Cargo.toml

@@ -171,6 +171,7 @@ geo-types = "0.7.19"
 geoarrow = "0.8.0"
 geoarrow-cast = "0.8.0"
 get_dir = "0.5.0"
+git2 = { version = "0.21", default-features = false }
 glob = "0.3.2"
 goldenfile = "1"
 half = { version = "2.7.1", features = ["std", "num-traits"] }
@@ -276,6 +277,7 @@ thiserror = "2.0.3"
 tokio = { version = "1.52" }
 tokio-stream = "0.1.17"
 tokio-util = "0.7.17"
+toml = "0.9"
 tpchgen = "3.0.0"
 tpchgen-arrow = "3.0.0"
 tracing = { version = "0.1.41", default-features = false }

bench-orchestrator/README.md

@@ -364,6 +364,6 @@ Benchmarks are built with:
 
 - Profile: `release_debug`
 - RUSTFLAGS: `-C target-cpu=native -C force-frame-pointers=yes`
-- Features: `unstable_encodings`
+- Features: none
 
 This enables native CPU optimizations while preserving debug symbols for profiling.

bench-orchestrator/bench_orchestrator/config.py

@@ -282,7 +282,7 @@ class BuildConfig:
 
     profile: str = "release_debug"
     rustflags: str = "-C target-cpu=native -C force-frame-pointers=yes"
-    features: tuple[str, ...] = ("unstable_encodings",)
+    features: tuple[str, ...] = ()
 
 
 def get_workspace_root() -> Path:

benchmarks/compress-bench/Cargo.toml

@@ -41,7 +41,6 @@ vortex-cuda = { workspace = true, optional = true }
 [features]
 cuda = ["dep:tempfile", "dep:vortex-cuda"]
 lance = ["dep:lance-bench"]
-unstable_encodings = ["vortex/unstable_encodings", "vortex-cuda?/unstable_encodings"]
 
 [[bin]]
 name = "compress-bench"

benchmarks/compress-bench/README.md

@@ -41,11 +41,11 @@ Both sides therefore decode all the way to device-resident arrays, which is what
 
 ```bash
 cargo run -p compress-bench --profile release_debug \
-  --features cuda,unstable_encodings -- --gpu-decompress
+  --features cuda -- --gpu-decompress
 
 # pick the Parquet page codec the GPU file is written with (default: snappy)
 cargo run -p compress-bench --profile release_debug \
-  --features cuda,unstable_encodings -- --gpu-decompress --gpu-parquet-codec zstd
+  --features cuda -- --gpu-decompress --gpu-parquet-codec zstd
 ```
 
 ### cuDF
@@ -116,7 +116,7 @@ still publishes comparable numbers — it just takes considerably longer:
 
 ```bash
 cargo run -p compress-bench --profile release_debug \
-  --features cuda,unstable_encodings -- --gpu-decompress --gpu-verify --iterations 1
+  --features cuda -- --gpu-decompress --gpu-verify --iterations 1
 ```
 
 Any `--gpu-decompress` run reports on every dataset rather than stopping at the first failure, so

benchmarks/datafusion-bench/Cargo.toml

@@ -58,7 +58,6 @@ custom-labels = { workspace = true }
 
 [features]
 cuda = ["dep:vortex-cuda"]
-unstable_encodings = ["vortex/unstable_encodings"]
 
 [lints]
 workspace = true

benchmarks/duckdb-bench/Cargo.toml

@@ -37,7 +37,6 @@ vortex-duckdb = { workspace = true }
 
 [features]
 cuda = ["dep:vortex-cuda"]
-unstable_encodings = ["vortex/unstable_encodings"]
 
 [lints]
 workspace = true

benchmarks/lance-bench/Cargo.toml

@@ -30,9 +30,6 @@ tokio = { workspace = true, features = ["full"] }
 tracing = { workspace = true }
 vortex-bench = { workspace = true }
 
-[features]
-unstable_encodings = ["vortex-bench/unstable_encodings"]
-
 [lib]
 test = false
 

benchmarks/random-access-bench/Cargo.toml

@@ -27,7 +27,6 @@ vortex-bench = { workspace = true }
 
 [features]
 lance = ["dep:lance-bench"]
-unstable_encodings = ["vortex-bench/unstable_encodings"]
 
 [[bin]]
 name = "random-access-bench"

benchmarks/string-bench/Cargo.toml

@@ -32,15 +32,8 @@ vortex-btrblocks = { workspace = true }
 vortex-fsst = { workspace = true }
 vortex-onpair = { workspace = true }
 
-[features]
-unstable_encodings = [
-    "vortex/unstable_encodings",
-    "vortex-btrblocks/unstable_encodings",
-]
-
 [[bin]]
 name = "string-bench"
-required-features = ["unstable_encodings"]
 
 [lints]
 workspace = true

benchmarks/string-bench/README.md

@@ -80,22 +80,22 @@ Input preparation is outside benchmark timing.
 
 ```bash
 # Tracked metrics: size, write, read for every configured column and encoder.
-cargo run -p string-bench --profile release_debug --features unstable_encodings
+cargo run -p string-bench --profile release_debug
 
 # Focus on selected columns or encoders.
-cargo run -p string-bench --profile release_debug --features unstable_encodings -- \
+cargo run -p string-bench --profile release_debug -- \
   --columns URL --encoders onpair
 
 # Add the direct codec microbenchmark.
-cargo run -p string-bench --profile release_debug --features unstable_encodings -- \
+cargo run -p string-bench --profile release_debug -- \
   --suite both
 
 # Emit benchmark-comparator JSONL.
-cargo run -p string-bench --profile release_debug --features unstable_encodings -- \
+cargo run -p string-bench --profile release_debug -- \
   --display-format gh-json --output-path results.json
 ```
 
-Run `cargo run -p string-bench --features unstable_encodings -- --help` for all
+Run `cargo run -p string-bench -- --help` for all
 filters and tuning options.
 
 Before timing, the benchmark checks that each requested encoding was produced

benchmarks/string-bench/src/lib.rs

@@ -1,8 +1,6 @@
 // SPDX-License-Identifier: Apache-2.0
 // SPDX-FileCopyrightText: Copyright the Vortex contributors
 
-#![cfg(feature = "unstable_encodings")]
-
 //! String-compression benchmarks for Vortex.
 //!
 //! Two intentionally separate suites: