PositionGroup.fetch_position_info() breaks if merge ids order does not match chronological order of the positions

[gshvarts]

I'm running into an issue where PositionGroup.fetch_position_info() returns an empty dataframe and have narrowed it down to what I think is a bug at the end of the fetch_position() function.

The two pos_merge_ids that I'm trying to combine into a position group are as follows:
run session 04_r2 (happens first): UUID('a7585350-6b4b-8f46-b4c9-e350ecffc1b6')
sleep session 05_s3 (happens second): UUID('141c4c08-ec23-114f-ebbf-0168fc3b2b42')

Although I pass these keys into PositionGroup().create_group() via a list of dicts in the above order, they end up getting iterated through in alphabetically-sorted order in PositionGroup.fetch_position() which results in the position_info dataframe containing the sleep position info (chronologically later) first and the run position info (chronologically earlier) second. Because of this, once the dataframes are concatenated and indexed from [min_time: max_time] (see line 231 in spyglass.decoding.v1.core.py), this returns an empty dataframe.

Instead, I think the concatenated dataframe should be sorted by the time index prior to doing the min_time, max_time indexing.

I've replaced the original line 231 with
position_info = pd.concat(position_info, axis=0).sort_index().loc[min_time:max_time] in my own code and that solved the issue for me!

[edeno]

That makes sense. Thank for filing the issue!