ferro-hgvs 0.7.1__tar.gz

ferro_hgvs-0.7.1/.claudeignore

@@ -0,0 +1,26 @@
+# Build artifacts (never need to scan)
+target/
+
+# Mutalyzer cache (562k+ files causes freeze)
+mutalyzer_cache_new/
+mutalyzer_cache/
+
+# Coverage and profiling
+coverage/
+tarpaulin-report.json
+dhat-heap.json
+*.profraw
+*.profdata
+
+# Fuzz artifacts
+fuzz/artifacts/
+fuzz/corpus/
+fuzz/target/
+
+# Large binary/reference files (can still read specific files when asked)
+reference_data/*.fa
+reference_data/*.fa.gz
+reference_data/*.fasta
+reference_data/*.fasta.gz
+*.bgz
+*.tbi

ferro_hgvs-0.7.1/.codecov.yml

@@ -0,0 +1,26 @@
+# Codecov configuration
+# https://docs.codecov.com/docs/codecovyml-reference
+
+coverage:
+  status:
+    project:
+      default:
+        target: 70%
+        threshold: 5%
+        informational: true
+    patch:
+      default:
+        target: 70%
+        threshold: 10%
+        informational: true
+
+  # Don't require coverage on these paths
+  ignore:
+    - "src/bin/*"
+    - "benches/*"
+    - "tests/*"
+
+comment:
+  layout: "reach,diff,flags,files"
+  behavior: default
+  require_changes: false

ferro_hgvs-0.7.1/.coderabbit.yaml

@@ -0,0 +1,104 @@
+# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json
+#
+# CodeRabbit review configuration for ferro-hgvs (HGVS variant normalizer).
+#
+# This repo is output-stability-critical: parsing, normalization, and coordinate
+# projection must be idempotent, round-trip stable, and byte-identical to the
+# external oracles we conform to (biocommons hgvs, Mutalyzer, VariantValidator).
+# The `assertive` profile + the path instructions below aim the review at the
+# failure modes that actually bite here (output divergence, non-idempotent
+# normalization, parser/fast-path disagreement, coordinate off-by-one, panics on
+# adversarial input, spec-conformance drift) rather than generic style. CI
+# (proptests, fuzz targets, idempotency tests, conformance oracles) is the
+# primary correctness net; assertive makes CodeRabbit surface the borderline tier
+# that net can still miss — a false positive here is cheaper than a missed
+# output divergence.
+#
+# Scope: this tunes the auto-read GitHub CodeRabbit app. The local `coderabbit`
+# CLI does NOT reliably inherit this file via `-c .coderabbit.yaml` (it can miss
+# findings the bot catches), so treat the CLI as a lighter pre-push smoke check,
+# not a substitute for the bot review.
+
+language: en-US
+
+# Terse, technical, no praise or filler — match the codebase's review style.
+# NB: CodeRabbit caps tone_instructions at 250 characters — keep this terse.
+tone_instructions: >-
+  Concise and technical. Skip praise and restating the code. Lead with the risk
+  and a one-line fix. Favor correctness, idempotency, and byte-identity with the
+  conformance oracles over style. Skip clippy/rustfmt nits; CI enforces them.
+
+reviews:
+  # Surface borderline correctness findings, not just high-confidence ones. This
+  # codebase prefers a false positive to a missed output divergence.
+  profile: assertive
+  high_level_summary: true
+  poem: false
+  collapse_walkthrough: false
+  # Reviews are triggered on demand, not automatically. CodeRabbit enforces a
+  # per-developer hourly review rate limit; auto-review on PR-open (`enabled`)
+  # and on every push (`auto_incremental_review`) fire extra reviews into that
+  # one shared budget, so a burst of PRs/commits exhausts it and later reviews
+  # stall for hours. Disabling both keeps the budget under deliberate control —
+  # reviews run only when requested. On-demand `@coderabbitai review` still works
+  # with `enabled: false`, and the assertive profile + path instructions above
+  # still apply to those reviews.
+  auto_review:
+    enabled: false
+    auto_incremental_review: false
+    drafts: false
+  # Never auto-block merges on a review verdict; findings are advisory.
+  request_changes_workflow: false
+
+  # Don't spend review budget on build output, vendored trees, lockfiles, or the
+  # generated spec fixture (regenerated from code, never hand-edited).
+  path_filters:
+    - "!**/target/**"
+    - "!**/*.lock"
+    - "!vendor/**"
+    - "!assets/**"
+    - "!tests/fixtures/grammar/hgvs_spec_normalization.json"
+
+  path_instructions:
+    - path: "src/hgvs/parser/**/*.rs"
+      instructions: >-
+        This is the HGVS grammar parser (nom-based). The fast path in
+        `fast_path.rs` is the DEFAULT parse route and is only sound if it is
+        observationally identical to the generic `parse_variant` for EVERY input
+        — both reject, or both accept and yield the same `HgvsVariant`. Treat any
+        change to fast-path eligibility or output as potentially behavior-changing
+        and confirm the differential test (`tests/fast_path_differential.rs`,
+        both its table and its proptest) still covers it. Flag: inputs the spec
+        rejects being accepted (or vice-versa); panics / unwraps / indexing on
+        attacker-controlled input (this parser is fuzzed — `fuzz/fuzz_targets/`);
+        loss of parse→Display round-trip stability; and `usize` arithmetic on
+        positions/offsets that can wrap or narrow.
+    - path: "src/normalize/**/*.rs"
+      instructions: >-
+        This is normalization: 3'-shifting, boundary handling, repeat shuffling,
+        edit merging, and overlap resolution. The core invariant is IDEMPOTENCY —
+        normalize(normalize(x)) == normalize(x) — and shift correctness
+        (3'-most placement, exact repeat-unit boundaries). Confirm changes are
+        covered by `tests/idempotency_tests.rs` and `tests/normalize_property_tests.rs`.
+        Flag: off-by-one in shuffle/boundary windows; wrong shift direction or
+        strand handling; mishandled circular/mitochondrial wraparound; overlap
+        merges that silently drop or reorder edits; and any path that can make a
+        second normalization pass change the output.
+    - path: "src/{project,convert,coords,spdi,liftover}/**/*.rs"
+      instructions: >-
+        This is coordinate projection and cross-system conversion (c/g/n/r/p,
+        SPDI, liftover, CDS/intron math, protein consequence + frameshift
+        classification). These are the off-by-one minefields. Flag: intronic
+        offset sign errors on the minus strand; CDS start/end clamp boundaries;
+        UTR/exon-junction edge cases; frame/codon arithmetic; and any projection
+        that emits coordinates without validating them against transcript/contig
+        bounds. Require coverage by the relevant `tests/projection*.rs`,
+        `tests/spdi_tests.rs`, or `issue_*`/`*_matrix` regression test.
+    - path: "tests/**/*.rs"
+      instructions: >-
+        Generate test data programmatically; do not add committed fixtures.
+        Conformance assertions must check against an INDEPENDENT oracle
+        (biocommons hgvs, Mutalyzer, VariantValidator) or a different internal
+        code path than the one under test — not a self-consistency check. Flag
+        assertions weaker than the stated contract, and `issue_*` regression
+        tests that assert the buggy output instead of the fixed one.

ferro_hgvs-0.7.1/.config/nextest.toml

@@ -0,0 +1,8 @@
+[profile.default]
+fail-fast = false
+
+[profile.ci]
+test-threads = "num-cpus"
+fail-fast = true
+slow-timeout = { period = "60s", terminate-after = 2 }
+retries = 1

ferro_hgvs-0.7.1/.gitattributes

@@ -0,0 +1,2 @@
+tests/fixtures/bulk/*.gz filter=lfs diff=lfs merge=lfs -text
+tests/fixtures/validation/*_exhaustive.json.gz filter=lfs diff=lfs merge=lfs -text

ferro_hgvs-0.7.1/.gitignore

@@ -0,0 +1,85 @@
+# Generated by Cargo
+/target/
+
+# Cargo.lock is included for binary crates
+# Cargo.lock
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+*~
+
+# macOS
+.DS_Store
+
+# Test output
+*.profraw
+*.profdata
+
+# Coverage
+coverage/
+tarpaulin-report.html
+tarpaulin-report.json
+cobertura.xml
+
+# Fuzz
+fuzz/artifacts/
+fuzz/corpus/
+fuzz/target/
+
+# Memory profiling
+dhat-heap.json
+
+# External validation data (large files)
+data/clinvar/
+data/exac/
+data/gnomad/
+data/lovd/
+data/civic/
+data/uta/
+data/seqrepo/
+data/ferro/
+data/mutalyzer/
+external-repos/
+reference_data/
+results/
+
+# Local configuration (generated via `ferro-web config`)
+config/service.toml
+
+# Local design specs and implementation plans (not tracked)
+docs/superpowers/
+
+# Python
+__pycache__/
+*.pyc
+*.egg-info/
+.venv/
+*.so
+
+# Manuscript (benchmark workflow, figures, drafts)
+manuscript/
+
+# Pixi
+.pixi/
+
+# Claude
+.claude/
+
+# Generated HGVS spec-normalization fixture (regenerated by the
+# `generate_spec_fixture` example; see CLAUDE.md). Tracking it made every
+# parser PR a conflict magnet, so it is produced on demand instead.
+/tests/fixtures/grammar/hgvs_spec_normalization.json
+/tests/fixtures/grammar/.hgvs_spec_normalization.*.tmp
+
+# Generated reference caches (cdot rkyv, TranscriptDb)
+*.tdb
+*.rkyv
+
+# Local, untracked third-party corpus checkouts (e.g. the full hgvs-rs
+# projection corpus consumed by scripts/refresh-hgvs-rs-projection-fixtures.py).
+# Only the curated tests/fixtures/hgvs-rs-projection/cases.json is committed;
+# the bulk corpus must never be added to git.
+/vendor/

ferro_hgvs-0.7.1/.gitmodules

@@ -0,0 +1,3 @@
+[submodule "assets/hgvs-nomenclature"]
+	path = assets/hgvs-nomenclature
+	url = https://github.com/HGVSnomenclature/hgvs-nomenclature.git

ferro_hgvs-0.7.1/.pre-commit-config.yaml

@@ -0,0 +1,77 @@
+# Pre-commit hooks for ferro-hgvs
+# Install: pip install pre-commit && pre-commit install
+# Run manually: pre-commit run --all-files
+#
+# `default_install_hook_types` makes the bare `pre-commit install`
+# wire up BOTH `pre-commit` (per-commit fast checks) and `pre-push`
+# (the spec-fixture drift guard, which requires a Rust build and
+# would be too slow per commit).
+default_install_hook_types: [pre-commit, pre-push]
+
+repos:
+  # Rust formatting
+  - repo: local
+    hooks:
+      - id: cargo-fmt
+        name: cargo fmt
+        entry: cargo fmt --all --
+        language: system
+        types: [rust]
+        pass_filenames: false
+
+      # Spec-fixture drift guard (#397 item 7). CI already runs this
+      # check (search for `generate_spec_fixture --check` in
+      # `.github/workflows/ci.yml`), but a push that drifts the
+      # fixture wastes a full CI cycle to discover the drift. The
+      # pre-push stage runs once per `git push` rather than per
+      # commit, so the build cost is paid only when actually pushing.
+      - id: generate-spec-fixture-check
+        name: generate_spec_fixture --check
+        entry: cargo run --features dev --example generate_spec_fixture -- --check
+        language: system
+        # always_run (not types: [rust]) — a push that only touches the
+        # spec submodule or test fixtures still needs the drift guard,
+        # and pre-commit otherwise skips type-filtered hooks when no
+        # matching files are staged in the push range.
+        always_run: true
+        pass_filenames: false
+        stages: [pre-push]
+
+  # Rust linting (optional - can be slow)
+  # Uncomment if you want clippy in pre-commit
+  # - repo: local
+  #   hooks:
+  #     - id: cargo-clippy
+  #       name: cargo clippy
+  #       entry: cargo clippy --features dev -- -D warnings
+  #       language: system
+  #       types: [rust]
+  #       pass_filenames: false
+
+  # Python linting with ruff
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.15.12
+    hooks:
+      - id: ruff
+        args: [--fix]
+      - id: ruff-format
+
+  # Python type checking with mypy
+  - repo: https://github.com/pre-commit/mirrors-mypy
+    rev: v1.10.0
+    hooks:
+      - id: mypy
+        files: ^python/ferro_hgvs/
+        args: [--config-file=pyproject.toml]
+
+  # General file hygiene
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v4.6.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: check-yaml
+      - id: check-toml
+      - id: check-merge-conflict
+      - id: check-added-large-files
+        args: [--maxkb=500]