Conversation
0f981f5 to
7b4c587
Compare
Key changes: - Bump neuroconv from 0.6.1 to 0.9.1 across all environments - Sync ElectrodeColumns/UnitColumns metadata with all schema properties (incl. data_type inference) - Fix Table.js: value == '' was true for 0 (JS coercion bug) - Add metadata/conversion_options kwargs to temporally_align_data_interfaces (0.7.0 API change) - Add _ensure_dandi_staging_alias() for dandi >= 0.74.0 compat - Update automatic_dandi_upload param: staging → sandbox (0.9.0 API change) - Pin h5py < 3.13 on macOS arm64 (HDF5 symbol mismatch with pytables) - Remove deprecated macos-13 CI runners from all workflows - Update E2E tests: SpikeGLX now uses folder_path + stream_id (0.9.0 API change) - Loosen ndx-pose (>= 0.1.1) and nwbinspector (>= 0.6.2) pins - Remove neo and pandas pins (no longer needed with neuroconv >= 0.8.0)
ef5d1bc to
b070be2
Compare
for more information, see https://pre-commit.ci
- Update scikit-learn 1.4.0 → 1.6.1 (numpy 2.x compatibility) - Pin tqdm_publisher, tzlocal, ndx-pose, nwbinspector, numcodecs to specific versions - Remove stale scipy < 1.12.0 constraint on MAC-intel - Remove _ensure_dandi_staging_alias (no longer needed with neuroconv 0.9.1) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
The ExampleDataTests failures on Windows and macOS are caused by a regression in neuroconv's Fix PR: catalystneuro/neuroconv#1667 |
|
As far as I can tell, catalystneuro/neuroconv#1667 was the only blocking issue. When that is released in NeuroConv 0.9.3, these tests should pass |
|
fyi @h-mayorquin |
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The coerce_schema_compliance_recursive function used elif between patternProperties and properties checks, so schemas with both (like Ophys with Fluorescence/DfOverF pattern props alongside ImagingPlane regular props) would skip regular property coercion. This caused emission_lambda to remain as a string/None instead of being converted to float, breaking BrukerTiff conversions. Also fix resolve_references to recursively resolve nested $refs and handle patternProperties, and add str-to-float coercion as a safety net for number fields. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
for more information, see https://pre-commit.ci
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
BrukerTiffSinglePlaneConverter inherits from BaseDataInterface (not NWBConverter) and doesn't override get_metadata_schema(), so the converter's schema lacks Ophys properties. This caused the schema-based NaN coercion in replace_none_with_nan to miss emission_lambda, leaving it as the string "NaN" instead of float NaN. Add replace_nan_strings() as a schema-independent fallback that converts all string "NaN" values (JavaScript NaN artifacts) to float NaN. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Rename VideoInterface to ExternalVideoInterface, remove duplicate AlphaOmegaRecordingInterface, add WhiteMatterRecordingInterface schema, and regenerate all interface schemas with latest neuroconv. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
for more information, see https://pre-commit.ci
…-recording-interface Add WhiteMatterRecordingInterface
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
for more information, see https://pre-commit.ci
…ording-interface Add Plexon2RecordingInterface
Regenerate schemas to preserve pydantic's DirectoryPath and FilePath format types instead of resolving them all to generic "path". Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
for more information, see https://pre-commit.ci
These files are generated by `python generateInterfaceSchema.py` and don't need to be tracked. This eliminates noisy diffs across 40+ JSON files on every neuroconv version bump. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…at-types Restore file-path and directory-path format types in schemas
|
@rly 's comment from review: add back intel support, with runners, which look something like "macos15-intel" |
|
The WhiteMatter Recording interface shows this error on the source data input page: @bendichter Can you resolve this? I don't have test data and cannot find any in The Plexon2 interface works only on Windows. We should limit its availability in this PR or a separate one. The tutorial will need to be updated because the new SpikeGLX interface requires a stream name (both single session and multi-session tutorials) but that can be a separate PR. Otherwise looks good to me. Thanks @bendichter ! |
The trailing comma after CellExplorerRecordingInterface caused JSON parse failures, breaking all DevTests, LiveServices, ExampleDataTests, and E2ELiveServices CI jobs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The ephys-datasets-macos-15-intel cache has never been saved due to repeated race conditions between concurrent PR workflow runs. This empty commit triggers a clean run to populate the cache. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Deleted ~5GB of stale/duplicate caches to make room for the macos-15-intel ephys cache that has never been successfully saved. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The ephys-datasets-macos-15-intel cache reservation got permanently stuck from a race condition between concurrent workflow runs. Bumping the key prefix forces a fresh cache entry. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
Thanks @bendichter ! |

Bumps neuroconv from 0.6.1 to 0.9.2 (latest) in a single PR.
Key Changes
Backend (manage_neuroconv.py)
data_typeinference.temporally_align_data_interfaces: Addedmetadata=None, conversion_options=Nonekwargs (neuroconv 0.7.0 API change)_ensure_dandi_staging_alias(): Registersdandi-stagingas alias fordandi-sandboxin dandi >= 0.74.0automatic_dandi_upload:staging=→sandbox=parameter (neuroconv 0.9.0 API change)Frontend
value == ""was true for0due to JS type coercion; fixed tovalue === "" || value == nullEnvironment / CI
E2E Tests
file_path→folder_path+stream_id(neuroconv 0.9.0 API change)Supersedes #1043, #1049, #1053.