FIX: fix for empty event durations #8384
Conversation
|
It works. Thank you! |
|
@cbrnr Thanks for the comments, I'll return to this tomorrow. |
|
@mmagnuski can you add a what's new? thx +1 for MRG |
|
Can we add a test for this please? :) |
|
@agramfort Sure! |
|
Great, looking forward! |
|
Friendly poke @mmagnuski :) |
|
Thanks for reminding @hoechenberger, I'm on low time budget recently, sorry for delays (but, hey, it's been only 18 days! ;) ) |
|
@mmagnuski Great! And no worries, I can very much relate. |
mmagnuski
force-pushed
the
fix_eeglab_empty_durations
branch
from
4f055d6
to
88ccbc6
Compare
|
I don't understand why codecov is angry at me... :( |
mmagnuski
force-pushed
the
fix_eeglab_empty_durations
branch
from
ff37287
to
8359319
Compare
|
@larsoner - less smoke now and we can see the green CI fields. :) |
| shutil.copyfile(op.join(base_dir, 'test_raw.fdt'), | ||
| empty_dur_fname.replace('.set', '.fdt')) | ||
| raw = read_raw_eeglab(input_fname=empty_dur_fname, preload=True) | ||
| assert (raw.annotations.duration == 0).all() |
There was a problem hiding this comment.
Shouldn't some or all of these been np.nan based on the conditional in eeglab.py and the ev.duration = np.array([], dtype='float') above?
There was a problem hiding this comment.
No, the eeglab reader turns np.nan somewhere later into zeros. I can use zeros instead of nans if you think this would be better. But we may also decide to expose the nan durations in annotations at some time in the future.
|
I killed most CIs as we can safely ignore them (I just changed |
|
Thanks for the corrections! |
|
Thanks @mmagnuski ! |
|
Thanks @larsoner @cbrnr @agramfort @hoechenberger ! |
@ociepkam - Try if this works for you.
Reference issue
Fixes #8383.
What does this implement/fix?
It seems some eeglab files contain empty durations - these are read as empty errays from matlab files and cause error when trying to set the
durationsarray.I'll add a test later.