hgvs-weaver-data 0.1.0__tar.gz

hgvs_weaver_data-0.1.0/.claude/rules/design-docs.md

@@ -0,0 +1,6 @@
+---
+paths:
+  - "docs/design/**/*.md"
+---
+
+Before writing or substantially rewriting a design doc, load the `writing-design-docs` skill.

hgvs_weaver_data-0.1.0/.claude/rules/python.md

@@ -0,0 +1,6 @@
+---
+paths:
+  - "**/*.py"
+---
+
+@../../docs/style/python.md

hgvs_weaver_data-0.1.0/.claude/rules/tests.md

@@ -0,0 +1,9 @@
+---
+paths:
+  - "**/tests/**"
+  - "**/test_*.py"
+  - "**/conftest.py"
+  - "src/weaver_data_provider/testing.py"
+---
+
+@../../docs/style/writing-tests.md

hgvs_weaver_data-0.1.0/.claude/skills/writing-design-docs/SKILL.md

@@ -0,0 +1,55 @@
+---
+name: writing-design-docs
+description: "Write or substantially rewrite a design doc under `docs/design/`. The procedure: what to read before drafting, what stays in the doc versus what moves to the code, the shape to draft in, the self-read, and the checks. Use when authoring a new design doc, rewriting an existing one, splitting or folding one, or bringing a doc up to the guide."
+---
+
+# Writing a design doc
+
+The guidance is `docs/style/design-docs.md` — the reader, the style, where low-level detail goes, the default shape, the
+policy. This is the procedure; it does not restate the guide.
+
+## Read first
+
+- The guide.
+- What the doc's reader has already read: `docs/PRODUCT.md`, `GLOSSARY.md`.
+- The doc as it stands, where one exists, and the Overview of each doc that will appear under `Related`.
+- The code and contract files the doc describes — the proto, the module entry points, the tests. "The code states it" is
+  verified there before a fact is left out, never assumed.
+
+## Separate what stays from what restates code
+
+Stays: the decisions, the named interfaces and what each promises, the consequences, the alternatives and why each was
+rejected, open questions.
+
+Goes: per-field paraphrase of a proto or schema, env-var names, paths beyond an entry point, function, class and test
+names, error strings, constants.
+
+For each code-owned fact the doc does not carry, note where it lives — the comment on the proto field, the docstring,
+the test — or that it has no home yet because that code does not exist on this branch. Keep the list; it is part of the
+report. No such fact goes into an inline comment beside the implementation (`docs/style/general.md`, Comments).
+
+## Draft
+
+The guide's default shape, deviating where the design reads better another way. A doc covering a surface, a flow, or
+several interfaces uses the concrete forms the guide names: a mockup of the surface, a request diagram, a plain
+statement of what is stored where, one subsection per interface in one consistent shape.
+
+## Re-read as the reader
+
+Read the draft as the maintainer the guide describes — has read `PRODUCT.md` and `GLOSSARY.md`, knows nothing about the
+area, one read on GitHub. Can they state each decision and the reason for it afterwards? Where not, fix the passage, not
+the reader.
+
+## Checks
+
+- `python3 tools/check_links.py` from the repo root — it scans all tracked Markdown, so read only the lines naming the
+  doc.
+- `pre-commit run mdformat --files <doc>`.
+- After the push, check a mermaid block in the branch's file view on GitHub; the PR diff will not render it, so a
+  reviewer opens the file.
+- The header line carries `**Related:**`.
+
+## Report
+
+Where each code-owned fact lives, calling out the ones with no home yet so the author can push them onto the code's
+documentation surface when it lands. Any question the writing raised for the author.

hgvs_weaver_data-0.1.0/.gitattributes

@@ -0,0 +1,5 @@
+# Generated protobuf stubs: collapse in GitHub diffs and drop from language stats.
+# Collapse only — GitHub still counts these lines in a PR's +/- total, as with any
+# lockfile; to inspect one, expand it in the diff.
+**/*_pb2.py linguist-generated=true
+**/*_pb2.pyi linguist-generated=true

hgvs_weaver_data-0.1.0/.github/workflows/lint.yml

@@ -0,0 +1,34 @@
+name: lint
+
+on:
+  pull_request:
+    types: [opened, ready_for_review, synchronize]
+  push:
+    branches: [main]
+
+jobs:
+  pre-commit:
+    # Runs every hook in .pre-commit-config.yaml — the whole static-check gate. Skips draft PRs.
+    if: github.event_name != 'pull_request' || !github.event.pull_request.draft
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v7
+      - uses: astral-sh/setup-uv@v10.2.0
+        with:
+          enable-cache: true
+      # buf on PATH for the buf-lint hook.
+      - uses: bufbuild/buf-setup-action@v1.50.0
+        with:
+          version: "1.73.0"
+          # No default token: the release lookup that resolves the download URL would go out
+          # anonymous, against a per-IP rate limit shared across the hosted runner pool.
+          github_token: ${{ github.token }}
+      # Pre-sync the env so the pyright hook's `uv run` is a no-op.
+      - run: uv sync --locked --python 3.13
+      - uses: actions/cache@v6
+        with:
+          path: ~/.cache/pre-commit
+          key: pre-commit-${{ hashFiles('.pre-commit-config.yaml') }}
+      - run: uv run pre-commit run --all-files --show-diff-on-failure

hgvs_weaver_data-0.1.0/.github/workflows/release.yml

@@ -0,0 +1,61 @@
+name: release
+
+# A `v*` tag builds, tests, publishes to PyPI through the trusted publisher registered for this
+# workflow (environment `pypi`), and creates the GitHub release with the built distributions.
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v7
+      - uses: astral-sh/setup-uv@v10.2.0
+        with:
+          enable-cache: true
+      # The tag is the release's name; the wheel's version comes from pyproject.toml. They must agree,
+      # or PyPI shows one version and the repo another.
+      - name: Check the tag matches the package version
+        run: |
+          set -euo pipefail
+          version="$(uv version --short)"
+          if [ "v${version}" != "${GITHUB_REF_NAME}" ]; then
+            echo "::error::tag ${GITHUB_REF_NAME} does not match pyproject version ${version}"
+            exit 1
+          fi
+      - run: uv sync --locked --python 3.13
+      - run: uv run pytest
+      - run: uv build
+      - uses: actions/upload-artifact@v7
+        with:
+          name: dist
+          path: dist/
+          if-no-files-found: error
+
+  publish:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi
+      url: https://pypi.org/project/hgvs-weaver-data/${{ github.ref_name }}
+    permissions:
+      # OIDC token for PyPI trusted publishing; contents for the GitHub release.
+      id-token: write
+      contents: write
+    steps:
+      - uses: actions/download-artifact@v8
+        with:
+          name: dist
+          path: dist/
+      - uses: pypa/gh-action-pypi-publish@v1.14.2
+      - name: Create the GitHub release
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |-
+          gh release create "${GITHUB_REF_NAME}" dist/* \
+            --repo "${GITHUB_REPOSITORY}" \
+            --title "hgvs-weaver-data ${GITHUB_REF_NAME#v}" \
+            --generate-notes

hgvs_weaver_data-0.1.0/.github/workflows/schema-compat.yml

@@ -0,0 +1,49 @@
+name: schema-compat
+
+on:
+  pull_request:
+    types: [opened, ready_for_review, synchronize]
+  push:
+    branches: [main]
+
+jobs:
+  backward-compatible:
+    # `buf breaking` under the FILE rule set (buf.yaml): stores and genomes already written sit in
+    # other projects' buckets, so a renumber or type change of a committed field fails here.
+    # Skips draft PRs.
+    if: github.event_name != 'pull_request' || !github.event.pull_request.draft
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      # Full history: the baseline is a commit, not the checkout.
+      - uses: actions/checkout@v7
+        with:
+          fetch-depth: 0
+      - uses: bufbuild/buf-setup-action@v1.50.0
+        with:
+          version: "1.73.0"
+          # No default token: the release lookup that resolves the download URL would go out
+          # anonymous, against a per-IP rate limit shared across the hosted runner pool.
+          github_token: ${{ github.token }}
+      # Baseline = what main held before this change. For a PR that is the base branch tip; on a push
+      # it is the commit the push started from, so every commit a multi-commit push brings is compared.
+      # A push that creates the branch has no before, and falls back to the previous commit; a root
+      # commit has no baseline and is skipped.
+      - name: Resolve baseline ref
+        id: baseline
+        env:
+          BEFORE: ${{ github.event.before }}
+        run: |
+          set -euo pipefail
+          if [ "${{ github.event_name }}" = "pull_request" ]; then
+            echo "ref=origin/${{ github.base_ref }}" >> "$GITHUB_OUTPUT"
+          elif [ -n "${BEFORE:-}" ] && [ "${BEFORE}" != "0000000000000000000000000000000000000000" ]; then
+            echo "ref=${BEFORE}" >> "$GITHUB_OUTPUT"
+          elif git rev-parse --verify -q HEAD^ >/dev/null; then
+            echo "ref=HEAD^" >> "$GITHUB_OUTPUT"
+          else
+            echo "ref=" >> "$GITHUB_OUTPUT"
+          fi
+      - if: steps.baseline.outputs.ref != ''
+        run: buf breaking --against ".git#ref=${{ steps.baseline.outputs.ref }}"

hgvs_weaver_data-0.1.0/.github/workflows/schema-freshness.yml

@@ -0,0 +1,38 @@
+name: schema-freshness
+
+on:
+  pull_request:
+    types: [opened, ready_for_review, synchronize]
+  push:
+    branches: [main]
+
+jobs:
+  regen-is-fresh:
+    # Skips draft PRs.
+    if: github.event_name != 'pull_request' || !github.event.pull_request.draft
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v7
+      # buf exports the protos + deps for protoc. No remote plugins — codegen is grpcio-tools' protoc.
+      - uses: bufbuild/buf-setup-action@v1.50.0
+        with:
+          version: "1.73.0"
+          # No default token: the release lookup that resolves the download URL would go out
+          # anonymous, against a per-IP rate limit shared across the hosted runner pool.
+          github_token: ${{ github.token }}
+      - uses: astral-sh/setup-uv@v10.2.0
+        with:
+          enable-cache: true
+      - run: uv run --locked --group codegen python scripts/regen.py
+      # The committed stubs must equal what regen just produced. `git add -A` stages edits, new
+      # files, and deletions; a non-empty staged diff means a .proto (or a tool) changed without
+      # regenerating.
+      - name: Fail if committed stubs are stale
+        run: |-
+          git add -A
+          if ! git diff --cached --exit-code; then
+            echo "::error::Generated stubs are stale. Run 'uv run --group codegen python scripts/regen.py' and commit the result."
+            exit 1
+          fi

hgvs_weaver_data-0.1.0/.github/workflows/tests.yml

@@ -0,0 +1,23 @@
+name: tests
+
+on:
+  pull_request:
+    types: [opened, ready_for_review, synchronize]
+  push:
+    branches: [main]
+
+jobs:
+  pytest:
+    # Skips draft PRs.
+    if: github.event_name != 'pull_request' || !github.event.pull_request.draft
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    steps:
+      - uses: actions/checkout@v7
+      - uses: astral-sh/setup-uv@v10.2.0
+        with:
+          enable-cache: true
+      - run: uv sync --locked --python 3.13
+      # No path arg: collect the testpaths configured in pyproject.toml.
+      - run: uv run pytest

hgvs_weaver_data-0.1.0/.gitignore

@@ -0,0 +1,11 @@
+.venv/
+__pycache__/
+*.pyc
+.pytest_cache/
+.ruff_cache/
+# Anchored: an unanchored `build/` also matches the builder package, src/weaver_data_provider/build.
+/dist/
+/build/
+*.egg-info/
+.claude/worktrees/
+.claude/settings.local.json

hgvs_weaver_data-0.1.0/.pre-commit-config.yaml

@@ -0,0 +1,58 @@
+# `pre-commit run --all-files` is the lint gate, locally and in CI. A hook added here is gated with no
+# separate workflow edit. pytest stays a separate job (too slow for the per-commit path).
+default_language_version:
+  python: python3.13
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+      - id: check-yaml
+      # Generated stubs are committed verbatim from protoc; a hygiene rewrite would drift them from
+      # the generator and break the freshness gate.
+      - id: end-of-file-fixer
+        exclude: '^src/(weaver_data_provider/v1|buf)/'
+      - id: trailing-whitespace
+        exclude: '^src/(weaver_data_provider/v1|buf)/'
+      - id: check-case-conflict
+      - id: check-merge-conflict
+      - id: detect-private-key
+      - id: debug-statements
+      - id: check-added-large-files
+        args: [--maxkb=1024]
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.15.12
+    hooks:
+      - id: ruff
+      - id: ruff-format
+  - repo: local
+    hooks:
+      # pyright needs the project deps to resolve imports, so it runs in the lint env via `uv run`.
+      # Whole-project (pass_filenames: false); fires on any staged .py.
+      - id: pyright
+        name: pyright
+        entry: uv run --group lint pyright
+        language: system
+        files: \.py$
+        pass_filenames: false
+      # buf lint: the hand-authored protos' structural and protovalidate discipline (buf.yaml).
+      # Whole-module; fires on any staged .proto. Needs buf on PATH.
+      - id: buf-lint
+        name: buf lint
+        entry: buf lint
+        language: system
+        files: \.proto$
+        pass_filenames: false
+  # Markdown formatter. CommonMark-strict, so intraword `_` (snake_case) stays literal and fenced
+  # code is untouched. gfm = tables/task-lists/autolinks; frontmatter = YAML headers. Wrap at 120.
+  - repo: https://github.com/hukkin/mdformat
+    rev: 1.0.0
+    hooks:
+      - id: mdformat
+        args: ["--wrap", "120"]
+        additional_dependencies:
+          - mdformat-gfm==1.0.0
+          - mdformat-frontmatter==2.1.2
+  - repo: https://github.com/google/yamlfmt
+    rev: v0.21.0
+    hooks:
+      - id: yamlfmt

hgvs_weaver_data-0.1.0/.yamlfmt

@@ -0,0 +1,10 @@
+# YAML formatter config; the YAML analogue of ruff-format / biome.
+# retain_line_breaks_single: yamlfmt can't selectively strip blank lines
+# between block-sequence items, so the no-blank-line-between-steps layout in
+# .github/workflows is kept by hand — this setting only stops yamlfmt from
+# reintroducing or multiplying blank lines elsewhere.
+formatter:
+  type: basic
+  retain_line_breaks_single: true
+  trim_trailing_whitespace: true
+  eof_newline: true

hgvs_weaver_data-0.1.0/CLAUDE.md

@@ -0,0 +1,95 @@
+# hgvs-weaver-data development notes
+
+## Product
+
+See [`docs/PRODUCT.md`](docs/PRODUCT.md) for the product north star — what the package is, the load-bearing principles
+(the publisher's placement, never a derived one; the files at rest are a contract; an answer or an error, never a
+guess), and what is out of scope. Read it before proposing designs or plans. Shared terminology lives in
+[`GLOSSARY.md`](GLOSSARY.md).
+
+## Working norms
+
+Operating directives for Claude (and any agent) in this repo; they counteract default model dispositions.
+
+- **Resist the minimal-diff reflex.** Don't reach for the smallest change that hides the symptom (special-casing,
+  papering over root causes). Aim for the correct fix at the right complexity level — not the smallest, not gold-plated.
+- **Fail loudly and early.** Raise on a missing expected input or precondition; never fall back to a default/placeholder
+  to limp along. A placeholder is an explicit caller input, never a code default.
+- **Never instruct around a defect — fix the defect.** Don't write prose telling readers to work around broken code —
+  "pass it as a string, the converter loses precision". Prose is untested, and callers who didn't read it stay broken.
+- **Push back; don't just comply.** When a design, name, or approach seems worse — including a shortcut you're asked to
+  take — say so with reasoning, unprompted. The author owns the final call.
+- **Offer better alternatives with trade-offs.** When a materially better approach than the proposed one exists, present
+  it and the trade-offs — don't just execute the ask.
+- **Investigate before producing.** Read the code and verify constraints first. Don't treat a training-pattern
+  convention as load-bearing unchecked; don't speculate about what you can read.
+- **Explain non-obvious changes first.** For a change whose rationale isn't self-evident, give the why before showing or
+  applying the diff.
+- **Ask when unsure** rather than assume intent.
+- **No intensifiers or emphasis filler.** Drop words and phrases that add emphasis but no information — "that's the
+  key", "crucially", "importantly", "the key insight", "it's worth noting". State the point plainly. Applies to all
+  prose: chat replies, PR/review comments, commit messages, and docs.
+
+## Code style
+
+@docs/style/general.md
+
+Tests — what one asserts, what it may depend on, what its data may contain — follow
+[`docs/style/writing-tests.md`](docs/style/writing-tests.md); it loads when a test file is touched.
+
+## Schema
+
+The protos under `proto/` are the source of truth for the files at rest; the committed Python stubs are generated. After
+editing a proto: `buf lint`, `buf breaking --against .git#branch=main`, then
+`uv run --group codegen python scripts/regen.py` and commit the stubs with the proto. Evolution is additive only — CI's
+`schema-compat` job fails a renumber or type change. A change a reader cannot ignore bumps `FORMAT_VERSION`, so an old
+reader refuses the new files rather than misreading them.
+
+## Dependency direction
+
+At runtime this package depends on hgvs-weaver and never the reverse ([`docs/PRODUCT.md`](docs/PRODUCT.md)). Don't
+propose a change that makes weaver's runtime or CLI import this package; its test suite may.
+
+## Docs
+
+Two audiences, two registers:
+
+- **Instruction files** are prompts and rules — `CLAUDE.md`, `.claude/rules/`, `.claude/skills/`: model-only, only what
+  changes behavior, no maintainer notes, no harness mechanics (which rules load when, where files live). A token there
+  is paid on every run that loads it; human-facing explanation belongs in `docs/` or code.
+- **Everything under `docs/`** is written for a human first — a maintainer who has read
+  [`docs/PRODUCT.md`](docs/PRODUCT.md) and [`GLOSSARY.md`](GLOSSARY.md) but not this area, and has to get the take-aways
+  from one read on GitHub. Explain with the clarity and style of Martin Kleppmann — motivation before mechanism,
+  specifics out of the argument's way. Detail that restates code — field lists, paths, env vars, test names — stays in
+  the code and is linked, never transcribed. A model reads what a human reads. Design docs are the durable design
+  record: one living doc per area under `docs/design/`, rewritten in place; no ADRs — rationale lives in the doc,
+  chronology in git. The guide is [`docs/style/design-docs.md`](docs/style/design-docs.md); to write or rewrite one,
+  load the `writing-design-docs` skill.
+
+## Committing
+
+- **Stage explicit paths**, not `git add -A` / `.`; explicit staging avoids sweeping in an untracked file.
+- **Pre-commit runs lint/format/hygiene** (`.pre-commit-config.yaml`); CI runs the same hooks plus pytest. Ensure hooks
+  are installed (`pre-commit install`) — if not, install or ask the author; never bypass with `--no-verify`.
+- **Correct a pushed branch with a new commit on top**, not amend + force-push. PRs squash-merge, so `main` history
+  stays linear regardless and intermediate fixups vanish on merge. Reserve force-push for rebasing a branch onto `main`.
+
+## Worktrees
+
+Worktrees go in `.claude/worktrees/` (gitignored), never `../` siblings.
+
+- **New branch** → the Claude Code worktree command.
+- **Existing branch** → `git worktree add .claude/worktrees/<name> <branch>` (the command only cuts fresh branches).
+
+## CI and review
+
+- **Adversarially review before opening a PR.** For any change with non-trivial code or logic, run adversarial review
+  passes in subagents with fresh context — the reviewer sees only the diff, not the authoring conversation — and fix the
+  findings autonomously; repeat until a pass surfaces only diminishing findings, then open the PR. Exempt: trivial
+  changes, doc-only changes, resource/asset changes.
+- **A PR description is written for the human reviewer**: what the change is and why, the take-aways, and where to look
+  — the altitude of a design doc's Overview, shorter. The diff carries the detail; don't narrate it. Same style:
+  [`docs/style/design-docs.md` § Style](docs/style/design-docs.md#style).
+- **Pin third-party GitHub Actions to the latest stable release**: the moving major tag (`@v7`) where the action
+  publishes one, else the exact latest version (`@v10.2.0`). Verify against the action's releases when adding or bumping
+  one.

hgvs_weaver_data-0.1.0/GLOSSARY.md

@@ -0,0 +1,52 @@
+# Glossary
+
+Shared terms across hgvs-weaver-data docs and code.
+
+## Reference data
+
+- **Assembly** — one version of the reference genome, GRCh38 or GRCh37. Coordinates mean nothing without the assembly
+  they are on, so every alignment and every store names its assembly.
+- **Annotation release** — one publication of a publisher's gene and transcript models against an assembly, named by the
+  publisher (`RS_2024_08` for RefSeq).
+- **Transcript record** — the publisher's own sequence of a transcript (an `NM_` or `NR_` accession with its version).
+  It usually matches the genome spliced at the transcript's exons, but not always.
+- **Alignment** (or **placement**) — where a transcript sits on one sequence of an assembly: its exons, strand, and the
+  per-exon CIGAR that says how the record and the genome differ. Published by NCBI; never derived here. A transcript can
+  have several — on X and on Y, on a chromosome and on a patch, or on each alternate locus that carries its gene.
+- **Alternate locus** — a scaffold holding another haplotype of a region too variable for one sequence (the MHC, the KIR
+  cluster). A gene the primary assembly's haplotype lacks is placed only on alternate loci. Alternate loci are `NT_`
+  accessions, as are the unlocalized and unplaced scaffolds of the primary assembly.
+- **Patch** — an `NW_` scaffold released between assembly versions: a fix patch corrects the primary assembly (and is
+  folded into the next major release), a novel patch adds alternate sequence.
+- **MANE** — the NCBI/EMBL-EBI project that picks one representative transcript per protein-coding gene (MANE Select)
+  and matches it across RefSeq and Ensembl. It ranks transcripts; it does not place them.
+- **HGNC** — the committee that assigns human gene symbols; the source of each gene's current, previous and alias
+  symbols.
+- **refget digest** — a sequence's identity computed from its residues (sha512t24u, written `SQ.…`), so two copies of
+  the same sequence share an identifier whatever they are called.
+
+## Project
+
+- **Gene bundle** (or **bundle**) — the unit of storage: one gene with every transcript, protein, alignment and sequence
+  that belongs to it (`GeneBundle` in `proto/`).
+- **Shard** — one annotation release's bundles in one bagz file, immutable once written and named for its contents.
+- **Store** — an index over an ordered list of shards: the key tables (symbol, accession, protein, digest), the interval
+  tables (bundles by genome position), and the **manifest** that names the shards, written last.
+- **Genome** — a derived copy of an assembly: the **blocks file**, every sequence cut into fixed-size compressed blocks,
+  and the **catalogue** of names, lengths and refget digests, written last.
+- **Format version** — the number every manifest and catalogue carries; a reader refuses one it does not know.
+- **Provider** — the object weaver reads through: weaver's `DataProvider` over one assembly's store and genome.
+
+## Testing
+
+- **Test double** — anything standing in for a production dependency in a test. The kinds below differ in fidelity and
+  in what a test can do with them
+  ([`docs/style/writing-tests.md`](docs/style/writing-tests.md#test-doubles-in-order-of-fidelity)).
+- **Fake** — a working, lightweight implementation of an interface, unsuitable for production and held to the real
+  implementation's contract.
+- **Stub** — returns canned values to put the code under test in a state; makes no claim about being called.
+- **Mock** — carries expectations about how it is called and fails the test when they are not met; the only double a
+  test verifies interactions on.
+- **Change detector** — a test that pins the code's current shape rather than an invariant, so it fails on every
+  legitimate change and catches no defect.
+- **Flaky test** — one that passes and fails on the same code; a defect to root-cause, never a reason to rerun.

hgvs_weaver_data-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Centre for Population Genomics
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

hgvs_weaver_data-0.1.0/PKG-INFO

@@ -0,0 +1,62 @@
+Metadata-Version: 2.5
+Name: hgvs-weaver-data
+Version: 0.1.0
+Summary: Reference data for the hgvs-weaver HGVS engine: a builder that cuts RefSeq and Ensembl releases into gene bundles, and the DataProvider that reads them.
+Author-email: Tobias Sargeant <toby.sargeant@populationgenomics.org.au>
+License-Expression: MIT
+License-File: LICENSE
+Requires-Python: >=3.12
+Requires-Dist: bagz[gcs]>=0.3.8
+Requires-Dist: google-cloud-storage>=2.18
+Requires-Dist: google-crc32c>=1.5
+Requires-Dist: hgvs-weaver>=0.7
+Requires-Dist: protobuf<7,>=6.33.5
+Requires-Dist: protovalidate>=1.2
+Requires-Dist: pysam>=0.23
+Description-Content-Type: text/markdown
+
+# hgvs-weaver-data
+
+Reference data for [hgvs-weaver](https://github.com/populationgenomics/hgvs-weaver): a builder that turns a publisher's
+release into a local store, and the `DataProvider` that feeds weaver from it.
+
+An HGVS engine is only as correct as what it is told about each transcript. A RefSeq transcript's sequence is not always
+the genome's — about one record in forty differs — and where they differ by an insertion or deletion, where the
+difference sits is a choice. NCBI publishes its choice: the alignment of every RefSeq transcript to the assembly. The
+builder takes that alignment exon by exon rather than deriving one, so the positions weaver computes are the ones
+ClinVar's names and VariantValidator's projections follow.
+
+Every placement NCBI publishes is kept — on a chromosome, an alternate locus or a patch — so a gene the primary assembly
+lacks is served where it is. An Ensembl transcript is placed at its exons, after a check that its record is the genome
+spliced there. [`docs/design/placements.md`](docs/design/placements.md) has the reasons. A retired RefSeq version — the
+one an older variant name cites — is served from NCBI's historical set, stacked under the current release
+([`docs/design/retired-versions.md`](docs/design/retired-versions.md)).
+
+A store is one record per gene — every transcript, its protein, its alignments and the sequences they cite — in
+[bagz](https://github.com/google-deepmind/bagz) files with sorted key and interval tables that are read into memory at
+open. A lookup is a search over bytes in memory and one ranged read, from local disk or `gs://`.
+
+```sh
+weaver-data-build refseq ...     # one RefSeq release's bundles, as a shard
+weaver-data-build historical ... # NCBI's historical set of RefSeq versions, each with its Entrez status
+                                 # (scripts/fetch_refseq_status.py), as a shard to stack under a release's
+weaver-data-build ensembl ...    # one Ensembl release's, placed at its exons and checked against the genome
+weaver-data-build index ...      # the index and manifest over an ordered list of shards
+weaver-data-build genome ...     # the assembly, cut into compressed blocks
+```
+
+The at-rest format is defined by the protos under `proto/`, and `buf breaking` gates every change: stores sit in other
+projects' buckets, so the format only grows.
+
+## Installing
+
+`pip install hgvs-weaver-data`; the module is `weaver_data_provider`.
+
+## Platforms
+
+bagz publishes wheels for Linux x86_64 and macOS arm64; anywhere else, installing this package builds bagz from source.
+
+## Development
+
+`uv sync`, then `uv run pytest`. The Python stubs under `src/` are generated from `proto/` by
+`uv run python scripts/regen.py` and committed.