feat(builtins): add shuf via codegen with helper-fn inlining#1538
Merged
feat(builtins): add shuf via codegen with helper-fn inlining#1538
Conversation
Implements `shuf` end-to-end. Adds the second genuinely-missing util from #1532's list (truncate landed in #1536); drops "command not found" for `shuf` invocations. Codegen tool extension: - bashkit-coreutils-port now scans uu_app() for plain identifier references (`parse_range`) and inlines any matching free fn defined at the source file's top level. The inlined fn runs through the same translate!() rewriter so its FTL keys resolve too. shuf's parse_range was the motivating case; the same path generalizes to any util whose value_parser refers to a local helper. - Generated preamble now always includes `clap::builder::ValueParser`, `std::ops::RangeInclusive`, and `std::str::FromStr` — three symbols utils in the wild reach for via `value_parser(...)` calls that pass either a path-style parser or a typed parser. `#![allow(unused_imports)]` keeps the imports inert for utils that don't reach for them. - syn dep gains the `visit` feature (in addition to existing `visit-mut`) for the read-only AST walk. - All existing generated files regenerated via `just regen-coreutils-args` to pick up the new imports — no semantic drift. Builtin behaviour (`crates/bashkit/src/builtins/shuf.rs`): - `-e ARG...` shuffles positional args. - `-i LO-HI` produces the integer range, then shuffles. - `-n N` caps output to N lines. - `-r` samples with replacement (requires `-n` in bashkit's safe mode; GNU loops forever without it — a non-starter inside an embedded VFS shell). - `-z` switches separator to NUL. - `-o FILE` writes to a VFS file. - `--random-seed`, `--random-source` rejected with explicit "not yet implemented" error, matching the existing tac -b/-r/-s pattern. - Tiny xorshift64* RNG inlined to avoid pulling rand into the always-on dep set (rand is currently feature-gated behind bot-auth). Quality is fine for line shuffling; never used cryptographically. Tests: - `crates/bashkit/tests/spec_cases/bash/shuf.test.sh` covers happy paths (`-e`, `-i`, `-n`, `-r -n`, `-z`), the `-r without -n` rejection, and unknown-flag rejection (with `### bash_diff` for the clap-vs-GNU exit-code divergence). - 3 unit tests in `shuf.rs` cover `split_separated` and the small RNG. Registration: - interpreter/mod.rs dispatch table. - compgen.rs allowlist (alphabetical). - every_builtin_handles_bogus_flag_cleanly list. - generated/mod.rs (`pub mod shuf_args;`). Spec: implementation-status.md lists the new test file with summary. Part of #1532.
Deploying with
|
| Status | Name | Latest Commit | Preview URL | Updated (UTC) |
|---|---|---|---|---|
| ✅ Deployment successful! View logs |
bashkit | 657495b | Commit Preview URL Branch Preview URL |
May 05 2026, 09:38 PM |
The strict bash_comparison_tests step in CI runs every spec test against real bash. GNU shuf -r without -n loops forever; bashkit's safe mode rejects it with exit 1. Without ### bash_diff the parity runner blocks until the GH Actions job timeout. Add the directive so bashkit's expected output (exit=1) is still asserted, but the test is excluded from real-bash comparison. Same pattern as `shuf_unknown_flag_rejected` and the existing tac unimplemented-flag rows. Part of #1532.
chaliy
added a commit
that referenced
this pull request
May 6, 2026
Minor release `0.4.1` → `0.5.0`. Two new builtins (`shuf`, `truncate`), a coreutils-codegen pipeline that ports uutils' `uu_app()` clap definitions for `cat`/`tac`/`truncate`/`shuf`/`readlink`, and `tool_def` flag-syntax improvements. ## Highlights - **Coreutils argument surface via codegen** — Ports uutils' `uu_app()` clap definitions into bashkit so builtins share the real coreutils argument shape; `cat`, `tac`, `truncate`, `shuf`, and `readlink` now flow through this surface, with a coreutils differential testing harness to catch parity drift. Pipeline reads a single pinned uutils revision so generated builtins, the differential harness, and CI all agree on the upstream source of truth (#1529, #1535, #1536, #1537, #1538, #1542). - **Site updates** — Bashkit agent skill is now published on the site, alongside rustdoc guides and content signal declarations for discoverability. ## What's Changed * refactor(builtins): migrate readlink to codegen-ported argument surface (#1542) * chore(site): publish bashkit agent skill (#1541) * chore(site): declare content signals * docs(site): publish rustdoc guides * feat(builtins): add shuf via codegen with helper-fn inlining (#1538) * chore(builtins): pin uutils revision as single source of truth (#1537) * feat(builtins): add truncate via codegen-ported argument surface (#1536) * test(builtins): add coreutils differential testing harness (#1535) * feat(builtins): port uutils argument surfaces via codegen (POC: cat, tac) (#1529) * feat(tool_def): accept --flag key=value... syntax for object/array flags (#1528) * fix(tool_def): coerce stringified JSON for array/object flag schemas (#1527) ## Publish-readiness report Per the updated `specs/release-process.md`: - [x] `cargo fmt --check` clean - [x] `cargo clippy --all-targets --all-features -- -D warnings` clean - [x] `cargo build` clean - [x] Versions synced across `Cargo.toml`, `crates/bashkit-cli/Cargo.toml`, `crates/bashkit-js/package.json`, `package-lock.json`, `Cargo.lock` - [x] `cargo publish --dry-run -p bashkit --allow-dirty` (after CI's monty-strip): **success** - [x] `cargo publish --dry-run -p bashkit-cli`: blocked only on ordering (`bashkit 0.5.0` not yet on crates.io) — resolves at real publish time when `publish-bashkit` runs first per `publish.yml`'s `needs:` chain. - [x] New `0.5.0` > latest published versions on crates.io / PyPI / npm (`0.4.1`). ## Companion change This branch also includes `chore(specs): document publish verification and post-merge monitoring`, codifying the verify-before-tag and watch-after-merge flow that this release follows. On merge, `release.yml` will create the GitHub Release `v0.5.0` and dispatch publish workflows for crates.io, PyPI, npm, and Homebrew. **Full Changelog**: v0.4.1...v0.5.0 --- _Generated by [Claude Code](https://claude.ai/code/session_01SvuLdA8pMAmP4woG2HqxKw)_
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Implements
shufend-to-end. The second genuinely-missing util from#1532's list (truncate landed in #1536); drops "command not found" for
shufinvocations.Codegen tool extension
bashkit-coreutils-portnow scansuu_app()for plain identifierreferences (
parse_range) and inlines any matching free fn defined atthe source file's top level. The inlined fn runs through the same
translate!()rewriter so its FTL keys resolve too. shuf'sparse_rangewas the motivating case; the same path generalizes toany util whose
value_parserrefers to a local helper.Generated preamble now always includes
clap::builder::ValueParser,std::ops::RangeInclusive, andstd::str::FromStr— three symbolsutils in the wild reach for via
value_parser(...)calls that passeither a path-style parser or a typed parser.
#![allow(unused_imports)]keeps the imports inert for utils thatdon't reach for them.
syndep gains thevisitfeature for the read-only AST walk.All existing generated files regenerated via
just regen-coreutils-argsto pick up the new imports — no semantic drift.
Builtin behaviour
-e ARG...shuffles positional args.-i LO-HIproduces the integer range, then shuffles.-n Ncaps output to N lines.-rsamples with replacement (requires-nin bashkit's safemode; GNU loops forever without
-n— a non-starter inside anembedded VFS shell).
-zswitches separator to NUL.-o FILEwrites to a VFS file.--random-seed,--random-sourcerejected with explicit"not yet implemented" error, matching the existing
tac -b/-r/-spattern.
Tiny xorshift64* RNG inlined to avoid pulling
randinto thealways-on dep set (
randis currently feature-gated behindbot-auth). Quality is fine for line shuffling; never usedcryptographically.
Test plan
cargo test -p bashkit --test spec_tests bash_spec_tests—includes new
shuf.test.sh(7 cases).cargo test -p bashkit --lib builtins::shuf— 3 unit testscover
split_separatedand the small RNG.every_builtin_handles_bogus_flag_cleanly— passes (shuf addedto the list).
generated_args_headers_match_pinned_uutils_revision— passes(all four
<util>_args.rsalign to the pinned rev).cargo clippy -p bashkit -p bashkit-coreutils-port --all-targets --all-features -- -D warnings— green.cargo fmt --check— green.Scope (
Part of #1532)After this PR, both genuinely-missing utils on #1532's list are
implemented (
truncatein #1536,shufhere). The remaining items inthat issue are codegen migrations of existing builtins (
tee,mktemp,realpath,readlink,stat,od) — separate PRs.Generated by Claude Code