Skip to content

v0.8.3

Latest
Latest

Choose a tag to compare

View all tags
@acgetchell acgetchell released this 26 Oct 03:22
· 12 commits to main since this release
Immutable release. Only release title and notes can be modified.
v0.8.3
3fa8a99
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

⚠️ Breaking Changes

  • Replace direct notebook-check invocations with
    just notebook-check or research-repo-tools notebooks lint. Configure
    example deadlines in tooling/examples.toml instead of EXAMPLE_TIMEOUT.
  • The minimum supported Rust version is now 1.99.0.
  • Retired benchmark baseline and threshold commands and
    legacy comparison readers are removed. New retained reports use shared
    .comparison.json and .evidence.json files; historical files remain records.
  • Python tooling now requires Python 3.15. The release benchmark workflow uses the Python publication adapter in place of scripts/release_benchmarks.sh.
  • Rust 1.99.1 or newer is now required.
  • Publish benchmark-equipped drafts with
    just release-publish TAG. Use bare just instead of help-workflows or
    perf-help, changelog-release instead of changelog-unreleased, and
    update-python-dependencies instead of update-python-deps.
  • Replace the directional safe_scalar_*_f64 and
    safe_coords_*_f64 helpers with try_finite_scalar and try_finite_coords.
  • Remove the public determinant helper and its geometry prelude export. Use Matrix::det() for a floating-point estimate with typed errors, or
    Matrix::det_sign_exact() for an exact sign.

Merged Pull Requests

  • Separate public API benchmarks from diagnostics #665
  • Make paper tooling opt-in and refresh project guides #663
  • Preserve scientific correctness across triangulation workflows #660
  • Delegate remaining numerical kernels to la-stack #658
  • Promote validated Cargo artifacts and align release commands #652
  • Adopt research-repo-tools v0.2.3 #651
  • Bump the github-actions group across 1 directory with 4 updates #646
  • Enforce geometry and recovery invariants #645
  • Automate reviewed releases and command discovery #644
  • Adopt la-stack filters on Rust 1.99.1 #643
  • Delegate spherical angles to la-stack #642
  • Retire local performance orchestration #641
  • Bump the github-actions group with 2 updates #640
  • Adopt research-repo-tools v0.1.8 #639
  • Bump acgetchell/research-repo-tools/.github/workflows/dependabot-approve.yml #638
  • Raise the minimum supported Rust version to 1.99 #637
  • Replace local helpers with research-repo-tools #635
  • Bump the github-actions group across 1 directory with 5 updates #624
  • Adopt shared maintenance and review workflows #623

Changed

  • Adopt shared maintenance and review workflows #623
    894c0b8

refactor(tooling): adopt shared maintenance and review workflows

  • Pin research-repo-tools 0.1.7 and use its managed toolchain for setup and CI.

  • Delegate dependency updates, release metadata, changelog rotation, and
    Semgrep fixture checks to the shared CLI.

  • Move completed changelog series while preserving release dates and links.

  • Verify live origin/main for branch reviews and preserve local-base
    overrides and uncommitted review.

  • Refresh Rust and Python dependency locks and retain updated tool pins.

  • [breaking] Replace local helpers with research-repo-tools #635
    af2f82d

    • Delegate file selection, subprocesses, archives, worktrees, Criterion
      parsing, and atomic publication to the pinned shared package.
    • Replace notebook and example runners with shared commands and
      declarative tooling configuration while preserving notebook policies.
    • Preserve exact Git patch bytes, publication recovery artifacts, and
      subprocess failure diagnostics.
    • Add just security with managed OSV and Gitleaks scans locally and in CI.
    • Document remaining v0.1.8 capabilities and dependent adoption issues.

  • [breaking] Adopt research-repo-tools v0.1.8 #639
    bde38f4

refactor!: adopt research-repo-tools v0.1.8

  • Replace reusable paper, example, notebook, scanning, metadata, and
    publication helpers with supported shared commands and APIs.

  • Keep consumer configuration and scientific Python modules under tooling,
    with only the release benchmark shell script remaining under scripts.

  • Adopt shared comparison JSON and evidence envelopes; retire legacy
    baseline caches, text/CSV readers, threshold commands, and their workflow.

  • Preserve scientific coverage, provenance, and publication checks while
    the remaining upstream performance capabilities are developed.

  • Refresh consumer dependency locks and declared tool versions.

  • [breaking] Retire local performance orchestration #641
    06284cf

    • Adopt research-repo-tools 0.1.9 for measurement execution, canonical Criterion inventory, saved-baseline verification, and release publication.
    • Preserve heterogeneous sampling and scientific provenance while rejecting incompatible retained measurements and colliding baseline names.
    • Keep the trusted publication runtime independent of release checkouts and revalidate the bound release before publishing.
    • Update the Python tooling environment to Python 3.15 and uv 0.13.1, and adopt la-stack 0.5.0.

  • Delegate spherical angles to la-stack #642
    ca5d337

refactor: delegate spherical angles to la-stack

  • Use la-stack's allocation-free slice angle API while retaining radius validation and arc-length scaling.
  • Preserve typed upstream numerical errors and document the quadratic ambient-dimension cost.
  • Add spherical distance benchmarks with independent correctness and allocation checks.
  • [breaking] Adopt la-stack filters on Rust 1.99.1 #643
    5c6bbaa

refactor!: adopt la-stack filters on Rust 1.99.1

  • Delegate relative in-sphere interval signs and circumsphere interval
    arithmetic to la-stack, removing the private interval kernel.

  • Reuse certified shared-vertex projections for Level 4 separation while
    preserving strict unit margins and exact geometric fallbacks.

  • Align the repository toolchain, minimum Rust version, and documentation
    with Rust 1.99.1.

  • [breaking] Delegate remaining numerical kernels to la-stack #658
    2c93eb2

    • Adopt la-stack v0.5.1 in the root crate and checkpoint fixture.
    • Use backend LU, Gram construction, checked dot products, and rational rounding while retaining exact geometric certification and fallback.
    • Refresh the locked protobuf and wrapt dependencies used by development tooling.

  • Separate public API benchmarks from diagnostics #665
    5d4bded

    • Use one curated public API measurement plan for routine CI and release evidence, with internal profiling selected explicitly.
    • Remove duplicate memory profiling jobs and post-merge comparisons while retaining instrumented allocation summaries.
    • Render summaries from retained measurement metadata and reject superseded benchmark plans during release promotion.
    • Document benchmark ownership and unpublished-release recovery, with regression coverage for measurement and reporting contracts.

Fixed

  • [breaking] Enforce geometry and recovery invariants #645
    460ef68

    • Make cospherical flip decisions consistent and periodic simplex identity
      translation-invariant across equality, ordering, hashing, and deduplication.
    • Evaluate periodic quality in lifted charts, preserve vertex-only TDS
      serialization, and return typed errors for oversized point requests.
    • Preserve unfinished builders, rollback configuration, predicate and lift
      causes, checkpoint payload failures, and strict-insphere witnesses.
    • Separate layer-local validation reports, consolidate finite-coordinate
      helpers and permutation parity, and remove unnecessary payload bounds.
    • Publish JSON files atomically and restore allocation-profiler state before
      propagating a measured panic.
    • Correct release packaging, profiling and Clippy failure reporting, artifact
      retention, Semgrep enforcement, and optional-feature test coverage.
    • Preserve benchmark confidence levels, derive conclusions from retained
      measurements, improve CLI and notebook diagnostics, and refresh tooling pins.
    • Align documentation, citations, immutable release evidence, and generated
      validation figures with the implementation.

  • Preserve scientific correctness across triangulation workflows #660
    e365c53

    • Complete exterior insertion conflict cavities and certify hull-corner
      deletion while preserving valid geometry and failure-atomic recovery.
    • Select periodic k=1 insertion images by exact containment, preserve caller
      coordinates and identities, and report ambiguity, range, and search limits.
    • Correct nonconvex point location, conservative spatial duplicate detection,
      and spatial-index restoration after failed bootstrap insertion.
    • Detect genuine nonflippable Delaunay violations, reject unsupported periodic
      location and exact intersection dimensions, and improve typed diagnostics
      for arity, incidence, orientation, and construction failures.
    • Relax snapshot and checkpoint payload bounds to the required Clone and
      serialization traits.
    • Use explicit benchmark fixtures and repair workloads, exclude owner
      destruction from operation timing, and correct profiling scope and quoting.
    • Require scientific report provenance, preserve artifacts on publication
      failure, and improve missing-tool diagnostics and cross-platform fixtures.
    • Prevent overflow in extreme-coordinate visualizations and keep notebook
      Semgrep checks within individual cell boundaries.
    • Remove vacuous and duplicate checks, strengthen invariant and API-contract
      coverage, and retain expensive exact checks with focused timeout allowances.
    • Align kernel documentation, scientific references, and development guidance;
      require production fixes for legitimate regressions and independent evidence
      before changing a test's contract.

  • Allow five-to-three benchmark plan transitions 5d85097

    • Accept the supported historical five-target baseline alongside the
      active three-target public API plan.
    • Preserve retired workloads as baseline-only evidence while retaining
      strict current-plan and shared-workload comparability requirements.
    • Reject incomplete, reordered, or unknown historical target plans.

Maintenance

ci!: automate reviewed releases and command discovery

  • Publish stable, immutable GitHub Releases through shared release gates,
    clean Cargo verification, deployment approval, and crates.io OIDC.
  • Leave verified benchmark assets on the draft for maintainer publication.
  • Generate complete Just help and align documentation ownership and release
    instructions with la-stack, including declared release dates.
  • Refresh Cargo dependencies, Python resolution, and managed tooling pins.
  • Bump the github-actions group across 1 directory with 4 updates #646
    bdde9d3

build(deps): bump the github-actions group across 1 directory with 4 updates

Bumps the github-actions group with 4 updates in the / directory: github/codeql-action/upload-sarif,
astral-sh/setup-uv, github/codeql-action/init and
github/codeql-action/analyze.

Updates github/codeql-action/upload-sarif from 4.38.2 to 4.38.3

Updates astral-sh/setup-uv from 10.0.1 to 10.3.0

Updates github/codeql-action/init from 4.38.2 to 4.38.3

Updates github/codeql-action/analyze from 4.38.2 to 4.38.3

  • Release notes

  • Changelog

  • Commits

  • Adopt research-repo-tools v0.2.3 #651
    d98d835

    • Pin shared tooling to v0.2.3 and refresh locked Python dependencies and managed tool versions.
    • Delegate Python dependency updates to the shared all-projects workflow and expose optional maintenance commands.
    • Use prebuilt formatters, prohibit notebook install cells, and report per-file coverage in CI.
    • Add an opt-in Actions updater policy while retaining the publishing restrictions and scientific validation contracts.

  • Promote validated Cargo artifacts and align release commands #652
    dba358a

ci: promote validated main Cargo artifacts for releases

  • Retain exact-commit Cargo packages after native and default/all-feature validation.
  • Revalidate immutable artifacts and frozen approval evidence before OIDC publication.
  • Use shared Actions output transport and bound archive inspection and diagnostics.
  • Document promotion recovery and refresh uv and locked Python tooling.
  • Make paper tooling opt-in and refresh project guides #663
    5fd85e0

build: make paper tooling opt-in and refresh project guides

  • Keep paper dependencies, tool pins, recipes, and checks opt-in while sharing the root Python environment and lockfile.
  • Move paper CI helpers under papers and run the paper workflow only by manual dispatch.
  • Refresh README scope, limitations, roadmap, security badges, contributor guidance, and project provenance for the v0.8.3 preparation.
  • Update the locked sse-starlette and starlette dependencies while retaining compatibility overrides and historical release evidence.

Refs acgetchell/Papers#1

ci: align Codacy checks with test and badge conventions

  • Apply the existing Codacy test exclusion to the optional paper tests while retaining local Python validation.
  • Scope the inline-HTML exception to the compact README badge rows.
  • Link contributor command discovery to the canonical README Quickstart.

docs: clarify the upstream tooling adoption condition

  • Describe the published upstream capabilities required before retiring local orchestration.
Assets 4
Loading