off by one error in stimulus file

[danielsf]

A large number of the ecephys sessions for the VBN release provoke an error when current_frame exceeds the length of draw_log in this block of code

https://github.com/AllenInstitute/AllenSDK/blob/master/allensdk/brain_observatory/behavior/stimulus_processing.py#L388-L400

    while current_frame <= stop_frame:
        epoch_length = 0
        while current_frame < stop_frame and draw_log[current_frame] == 1:
            epoch_length += 1
            current_frame += 1
        else:
            current_frame += 1

        if epoch_length:
            draw_epochs.append(
                (current_frame - epoch_length - 1, current_frame - 1,)
            )

Probably we need to do something more clever to set the value of stop_frame.

Tasks

• [ ] Determine why stop_frame is larger than len(draw_log)
• [ ] Determine solution that does not change the result of VBO NWB generation
• [ ] Implement fix

Validation

• [ ] We are able to generate VBN ecephys NWB files without provoking this error

[danielsf]

This will likely be addressed as part of

https://app.zenhub.com/workspaces/allensdk-10-5c17f74db59cfb36f158db8c/issues/alleninstitute/allensdk/2346

[aamster]

It is not correct to load the stimulus presentations table via the behavior stimulus pkl file as was done in VBO. The stimulus presentations table for VBN is read directly from a precomputed csv. This code block is only executed when we try to load the stimulus presentations table from the behavior pkl file. Therefore this is a nonissue.