Record the actual NWB Inspector version in validation results

[CodyCBakerPhD]

User-visible effect

• NWB Inspector validation results now carry the real nwbinspector version. Every NWB Inspector record from dandi validate (and from validation during dandi upload) now has origin.validator_version set to the installed nwbinspector version, e.g. "0.7.2". Before this change it was always "". The archive can now match these records to the nwbinspector version that produced them.
• The "outdated NWB Inspector" warning can now actually appear. If etelemetry reports a newer nwbinspector release than the one installed, dandi validate logs the existing "NWB Inspector version X is installed - please use the latest release…" warning once per process. Before this change that code never ran.
  • If the etelemetry server can't be reached (e.g. offline), the existing Failed to retrieve NWB Inspector version due to … warning is now logged once per process.
  • Setting NO_ET (etelemetry's own opt-out) skips the check silently.

The bug

_get_nwb_inspector_version() in dandi/files/bases.py cached the version in a module global that started as "", but the cache check was is not None. Since "" is not None, the function always returned "" right away: the version was never looked up and the etelemetry check below it never ran.

Changes

• dandi/files/bases.py
  • The cache now starts as None, typed as Version | None. That matches what nwbinspector.utils.get_package_version() returns (packaging.version.Version), so the existing < max_version comparison still works. _get_nwb_inspector_version() is annotated -> Version, and the caller still does str(...).
  • The lookup runs under a Lock. dandi upload validates assets in up to 5 threads at once, and without the lock several threads could each run the etelemetry lookup and log duplicate warnings. Most of this diff is re-indentation from the lock; it reads best with whitespace ignored.
  • The etelemetry query moved into a small helper, _get_latest_nwb_inspector_version(). It returns None when etelemetry.get_project() does, which happens when NO_ET is set (as in our CI). Without that, users with NO_ET set (and CI) would get a confusing TypeError: 'NoneType' object is not subscriptable warning. The existing try/except around the check is unchanged.
• dandi/tests/fixtures.py: new autouse fixture no_nwbinspector_etelemetry that stubs out the helper. Tests that validate NWB files then never query etelemetry over the network and never log "please upgrade" or "failed to retrieve" warnings. It patches only this dandi-private helper, not etelemetry or NO_ET globally, so other projects that load the dandi pytest plugin aren't affected.
• dandi/tests/test_files.py: new tests, all marked @pytest.mark.ai_generated:
  • test_validate_nwb_records_nwbinspector_version: validates the real simple4_nwb fixture and checks that every NWB Inspector result's origin.validator_version equals the installed nwbinspector version. Both sides are normalized with packaging.version.Version.
  • test_get_nwb_inspector_version_cached[...]: runs in a fresh interpreter, so the cache starts in the same state as for dandi validate, with etelemetry.get_project replaced by a counting fake (no network). It checks that two calls return the same installed version and etelemetry is queried exactly once. It is parametrized over the NO_ET case (no warning) and a newer upstream release (exactly one outdated warning). Resetting the cache in-process wouldn't catch this bug, because the bug was in the initial value.

The synthetic validator_version="" placeholders in dandi/cli/tests/test_cmd_validate.py (formatter tests) are deliberately left alone.

Testing

• New tests fail on master ({''} == {'0.7.2'} and ['', ''] == ['0.7.2', '0.7.2']) and pass with the fix.
• dandi/cli/tests/test_cmd_validate.py, dandi/tests/test_files.py, and dandi/validate/: 134 passed, 3 skipped.
• mypy dandi: no issues. pre-commit (black, isort, codespell, flake8, reuse): all passed.
• End-to-end: dandi validate -f json_lines on an NWB file with best-practice violations now shows "validator_version": "0.7.2" on the NWBI.* records.

🤖 Generated with Claude Code

https://claude.ai/code/session_011mgA1x4jnEavPHXQkyrnEk

---

Generated by Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 97.67442% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 90.89%. Comparing base (0d5f96e) to head (57a2f15).

Files with missing lines	Patch %	Lines
dandi/files/bases.py	94.44%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1966      +/-   ##
==========================================
+ Coverage   90.76%   90.89%   +0.13%     
==========================================
  Files          92       92              
  Lines       14159    14190      +31     
==========================================
+ Hits        12851    12898      +47     
+ Misses       1308     1292      -16     

Flag	Coverage Δ
unittests	90.89% <97.67%> (+0.13%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.