[MRG] Get multiple spaced events from single annotation

[massich]

This PR adds a parameter to events_from_annotation so that if this parameter
None, events correspond to the annotation onsets (as was happening until nos).
But if not, events_from_annotations returns as many events as they fit within
the annotation duration spaced according to this new parameter (right now is
called chunk_duraiton but we should discuss it before merging).

This PR needs to go in before #5718

---

Things that are not in this PR (it will be done in subsequent PRs):

• allow slicing of Annotations
• update the documentation of events and annotations

[codecov]

Codecov Report

Merging #5795 into master will decrease coverage by 0.11%.
The diff coverage is 20.83%.

@@            Coverage Diff             @@
##           master    #5795      +/-   ##
==========================================
- Coverage   88.53%   88.41%   -0.12%     
==========================================
  Files         369      369              
  Lines       68955    69041      +86     
  Branches    11614    11623       +9     
==========================================
- Hits        61048    61046       -2     
- Misses       5048     5131      +83     
- Partials     2859     2864       +5

[massich]

I would like to add slicing of the annotations so that this can be:
for onset, duration, description in raw.annotations[event_sel]:

[massich]

but this can be done afterwards

[larsoner]

Yes a __getitem__ that returns a copy and a __iter__ method (or whatever the correct Py3k one is named) would be good at some point

[massich]

@Slasnista you should be able to get the psg events just by doing the following:

raw = mne.io.read_raw_edf(psg_fname, stim_channel=False)
annotations = mne.read_annotations(hyp_fname)
desc2int = {'Sleep stage W': 5,
             'Sleep stage 1': 3,
             'Sleep stage 2': 2,
             'Sleep stage 3': 1,
             'Sleep stage 4': 1,
             'Sleep stage R': 4}
events, event_id = mne.events_from_annotations(raw, event_id=desc2int, chunk_duration=30)

[massich]

I've reverted the last commits where I refactored the code to be simpler because it was messing up with EDF and to fix it would complicate the PR. I'll get back to it once EDF #5741 is in and find_edf_events is out.

[agramfort]

@massich you have some pep8 errors.

[agramfort]

@massich you need to rebase

[massich]

@larsoner can you merge?

[massich]

After a talk IRL with @agramfort we decided to set back regexp to None instead of '.*'. And revoked the possibility of warning/raising to avoid the parameter at the API.

I see the value of on_missing (that was not the case for the None) but I agree that havning on_missing, verbose .. it keeps polluting the API. But if someone things that would rather have warning instead of error I don't mind revert 71623e8 ( and add a good first issue for someone to factor the code).

[massich]

Did I forgot to take them out from the signature? I'm AFK. Feel free to take them out and merge if you want. Other wise if you think that we should revert the last commit, feel free to do so. If not I'll remove the on_missing once I get back

…

On Wed, Dec 19, 2018, 19:20 Eric Larson ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In mne/tests/test_annotations.py <#5795 (comment)>: > @@ -530,7 +551,7 @@ def test_events_from_annot_in_raw_objects(): assert_array_equal(expected_events5, events5) with pytest.raises(ValueError, match='not find any of the events'): - events_from_annotations(raw, regexp='not_there') + events_from_annotations(raw, regexp='not_there', on_missing='error') on_missing is still here ------------------------------ In mne/annotations.py <#5795 (comment)>: > @@ -701,9 +701,47 @@ def _ensure_annotation_object(obj): 'mne.Annotations. Got %s.' % obj) +def _select_annotations_based_on_description(descriptions, event_id=None, + regexp=None, on_missing='ignore'): on_missing is still here — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#5795 (review)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AGt-44bbw-0QdHZPHibxayU0sFievb2Oks5u6oNSgaJpZM4ZV5Fb> .

[agramfort]

+1 for MRG when CIs are green