using spike_vector on core/check_sortings_equal test
#4282
Conversation
for more information, see https://pre-commit.ci
|
@tayheau I think we actually needed the unit_index lex sort after all! The reason is that before we were testing spiketrain-by-spiketrain, so the order didn't matter if we had synchronous spikes. Now with the spike vector representation we need to ensure unit_index is sorted too! Sorry about that! :) |
for more information, see https://pre-commit.ci
…spike_vector_test_check_sorting_equal
…yheau/spikeinterface into spike_vector_test_check_sorting_equal
| spikes.append(spikes_in_seg) | ||
|
|
||
| spikes = np.concatenate(spikes) | ||
| spikes = spikes[np.lexsort((spikes["unit_index"], spikes["sample_index"], spikes["segment_index"]))] |
There was a problem hiding this comment.
@samuelgarcia I think it's better to lexsort the whole array at the end to make sure that units are also sorted within samples
|
|
||
| spikes = np.concatenate(spikes) | ||
| spikes = spikes[np.lexsort((spikes["sample_index"], spikes["segment_index"]))] | ||
| spikes = spikes[np.lexsort((spikes["unit_index"], spikes["sample_index"], spikes["segment_index"]))] |
There was a problem hiding this comment.
we were missing the unit index lexsort in generate
| spikes = spikes[np.lexsort((spikes["unit_index"], spikes["sample_index"], spikes["segment_index"]))] | ||
| self._cached_spike_vector = spikes |
There was a problem hiding this comment.
Doing this in the init is an enormous cost in term of perfs.
I am not sure this is a good idea to lexsort at each reading.
lets discuss
src/spikeinterface/core/testing.py
Outdated
| mask &= spikes["sample_index"] >= start_frame | ||
| if end_frame is not None: | ||
| mask &= spikes["sample_index"] <= end_frame |
There was a problem hiding this comment.
searchsorted should be faster no ?
|
thanks @tayheau Merging! |
3 participants
Following this todo