Serve hgvs UTA lookups from a recorded cache in tests

[theferrit32]

Closes #659.

Tests no longer connect to the public UTA database or fetch sequences from NCBI. The hgvs library's data provider lookups (UTA queries and sequence fetches) are replayed from a recorded cache, tests/data/hgvs_cache.json, using hgvs's built-in cache modes. Library code is unchanged; the cache is configured only in the test suite.

• A session fixture in tests/conftest.py gives every HgvsTools one shared, cache-backed hgvs data provider. VRS_HGVS_CACHE_MODE selects run (default, never opens a UTA connection), learn, or verify, or empty to use UTA directly. These variables are read only by the test suite.
• hgvs stores its cache as a pickle, which can run code when loaded and can't be reviewed in a diff. The cache is committed as JSON instead, and tests/hgvs_cache_json.py converts it to a temporary pickle for hgvs when tests start.
• make record-hgvs-cache re-records the cache against UTA and a seqrepo-rest-service: hgvs records into the pickle named by VRS_HGVS_CACHE_FILE (learn mode only), and the JSON is written only if the test run passes. VCR cassettes are not touched.
• Re-recorded the hgvs-related cassettes, dropping NCBI requests that are now served from the cache.
• A test checks that a full hgvs translation in run mode makes no UTA queries or sequence fetches.
• The README now describes recording both VCR cassettes and the hgvs cache.

In CI the pytest step drops from about 60 s to about 10 s per Python version, and tests no longer hang or fail when the public UTA server is unavailable.

[theferrit32]

Previously I was just checking in the pickle file that biocommons/hgvs creates. But that has some security concerns because it's binary serialized python that cannot be easily inspected in a pr/commit diff and can execute arbitrary code when deserialized. So now this declarative json form gets written instead, and conftest.py converts it into the pickle form that biocommons/hgvs expects, on the fly.

[Copilot]

Copilot review overview

🟡 Changes recommended

The documented HgvsTools environment-variable behavior is implemented only by the pytest fixture.

Review effort: Balanced
Findings: 1

Open (1)

• Implement advertised cache environment variables in HgvsTools · New

What changed in this PR

Adds a recorded HGVS data-provider cache so tests can run without live UTA or NCBI access.

Changes:

• Adds JSON-to-pickle HGVS cache conversion and a shared cached provider fixture.
• Adds offline-cache coverage and removes superseded NCBI cassette interactions.
• Documents and automates cache re-recording.

File	Description
.gitignore	Ignores interrupted cache recordings.
Makefile	Adds the cache recording target.
README.md	Documents recorded test data workflows.
tests/​conftest.py	Configures the shared HGVS cache provider.
tests/​hgvs_cache_json.py	Converts HGVS caches between JSON and pickle.
tests/​data/​hgvs_cache.json	Stores recorded UTA and sequence results.
tests/​extras/​test_hgvs_tools.py	Verifies provider sharing.
tests/​extras/​test_allele_translator.py	Tests offline HGVS translation.
tests/​extras/​cassettes/​test_hgvs_cache_run_mode_is_offline.yaml	Records local SeqRepo interactions.
tests/​extras/​cassettes/​test_rle_seq_limit.yaml	Removes cached NCBI requests.
tests/​extras/​cassettes/​test_hgvs[NM_181798.1:c.1007G>T-expected7].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NM_001331029.1:c.722A>G-expected6].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000019.10:g.289485_289500del-expected9].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000019.10:g.289464_289465insCACA-expected8].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000013.11:g.32936732=-expected0].yaml	Removes an NCBI sequence fetch.
tests/​extras/​cassettes/​test_hgvs[NC_000013.11:g.32331093_32331094dup-expected4].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000013.11:g.32316467dup-expected5].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000007.14:g.55181320A>T-expected1].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000007.14:g.55181230_55181231insGGCT-expected3].yaml	Removes NCBI sequence fetches.
tests/​extras/​cassettes/​test_hgvs[NC_000007.14:g.55181220del-expected2].yaml	Removes NCBI sequence fetches.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.