FIX: Replace np.fromstring with np.frombuffer in GDF reader for NumPy… #13415
Conversation
… ≥1.24 Note that .encode() was added because np.frombuffer requires bytes.
|
Hello! 👋 Thanks for opening your first pull request here! ❤️ We will try to get back to you soon. 🚴 |
mne/io/edf/edf.py
Outdated
| etmode = fid.read(1).decode() | ||
| if etmode != "": | ||
| etmode = np.fromstring(etmode, UINT8).tolist()[0] | ||
| etmode = np.frombuffer(etmode.encode(), dtype=UINT8).tolist()[0] |
There was a problem hiding this comment.
We decode it just to encode it? Maybe we could do instead
etmode = fid.read(1)
if len(etmode):
etmode = np.frombuffer(etmode, dtype=UINT8)[0]
|
Also needs a changelog entry in |
|
would be great if the test coverage gap @scott-huberty mentioned in #13414 (comment) could also be closed here. @devparikh0506 let us know if you want help with that part. |
…string (NumPy ≥1.24) - Now reading a single byte directly from the file handle (fid.read(1)), and parsing it using np.frombuffer with dtype=UINT8. - Added changelog entry (doc/changes/devel/13415.bugfix.rst) and contributor link in doc/changes/names.inc.
|
Hi @larsoner, thanks for pointing out the improvement! I’ve updated the PR with the suggested change. Also @drammock, I’d be happy to add test coverage for this. Could you please provide a brief outline of the steps or the best place in the test suite to add it? That would help me make sure I implement it correctly. |
@devparikh0506 we are trying to figure that part out. Stand by! |
|
Needs mne-tools/mne-testing-data#124 @devparikh0506 I submitted a PR to add a test file to mne-testing-data . Once it is merged and @larsoner cuts a new release, You will need to update this PR to point to the newly released testing dataset, by following the directions here: mne-python/mne/datasets/config.py Lines 85 to 94 in a8e2fe8 mne-python/mne/datasets/config.py Lines 115 to 122 in a8e2fe8 Once this is updated, you can run: import mne
testing_fpath = mne.datasets.testing.data_path(force_update=True)
# FYI The new test file will be located at:
testing_fpath / "GDF" / "test_gdf_1.99.gdf"Which should download the updated testing dataset. At this point you can add a unit test to mne/io/edf/tests/test_gdf.py, that reads Let us know if you have any questions along the way or need any help. |
|
https://github.com/mne-tools/mne-testing-data/releases/tag/0.165 release: |
… ≥1.24
Note that .encode() was added because np.frombuffer requires bytes.
Reference issue (if any)
Fixes #13414.
What does this implement/fix?
np.fromstringin binary mode was removed in NumPy ≥1.24.np.frombuffer(etmode.encode(), dtype=UINT8)in_read_gdf_header.How was this tested?
pytest mne/io/edf/testssuccessfully (all relevant tests passed)..gdffiles using Python 3.11.7 + NumPy 2.3.3.read_raw_gdf("B0101T.gdf", preload=True)now works without errors.Additional information