mark special intervals in timeseries plots: e.g., ripple times

[magland]

This dataset provided by Eric has ripple times:

https://github.com/scratchrealm/sortingview-dev/blob/main/examples/load_jaq_03_12_data2.py

[jsoules]

Does this require special functionality that makes it distinct from issue #172?

[magland]

@jsoules - you're right - this is a duplicate. I will close this one and add my comment/link to the other one.

[magland]

Actually I decided to keep this one open and closed the other.

[jsoules]

Have been working on this. Testing some things now.

Testing using sha1://2102690caa2db411bc9cca4178270810e5a74f9a/Jaq_03_12_visualization_data2.npy?manifest=c38d157da18611aed00a092a938447a78d8292df for the jaq data with ripple events.

Loaded as per https://github.com/scratchrealm/sortingview-dev/blob/main/examples/load_jaq_03_12_data2.py and https://github.com/magland/sortingview/blob/main/sortingview/SpikeSortingView/create_spike_raster_plot.py to generate a raster plot with highlighted spans. The overall code is as follows:

import numpy as np
import figurl as fig
import kachery_client as kc
from math import ceil, floor

def create_spike_raster_plot_data(*, times: np.array, labels: np.array, label: str):
    unit_ids = np.sort(np.unique(labels))
    plots = []
    for unit_id in unit_ids:
        inds = np.where(labels == unit_id)[0]
        spike_times_sec = times[inds]
        plots.append({
            'unitId': unit_id,
            'spikeTimesSec': spike_times_sec.astype(np.float32)
        })

    data = {
        'type': 'RasterPlot',
        'startTimeSec': np.min(times),
        'endTimeSec': np.max(times),
        'plots': plots
    }
    return data

data_uri = 'sha1://2102690caa2db411bc9cca4178270810e5a74f9a/Jaq_03_12_visualization_data2.npy?manifest=c38d157da18611aed00a092a938447a78d8292df'
fname = kc.load_file(data_uri)
jaq_data = np.load(fname, allow_pickle=True)

spikes = jaq_data[2]
assert spikes['type'] == 'spikes'
spike_times = spikes['spike_times'].astype(np.float32)
cell_ids = spikes['cell_id'].astype(np.int32)

data = create_spike_raster_plot_data(
    times=spike_times,
    labels=cell_ids,
    label='Spikes'
)

# add ripples to data object
ripples = jaq_data[-1]
assert ripples['type'] == 'ripple_times'
np_starts = ripples['start_times'].to_numpy(dtype=np.float32)
np_ends = ripples['end_times'].to_numpy(dtype=np.float32)
assert len(np_starts) == len(np_ends)
for start, end in zip(np_starts, np_ends):
    assert start <= end

data['highlightIntervals'] = [{
    'intervalStarts': np_starts,
    'intervalEnds': np_ends
}]

figure = fig.Figure(
    view_url='http://localhost:3000/',
    data=data
)
print(figure.url(label='SPANS_TEST', channel='flatiron1'))

This yields https://figurl.org/f?v=http://localhost:3000/&d=3efdf652e24633ba8345fb6a02af9d01fc67e5d1&channel=flatiron1&label=SPANS_TEST as of this writing.

EDIT 2/23/22: Updated to ensure that the 'highlightIntervals' key points to a list, as I'd originally intended, not just an object. Regenerated FigURL.

[magland]

I've already merged the PR, but...

There is something off with the highlighted intervals. The location of the highlights changes relative to the spike events as you zoom in/out.

See: https://www.figurl.org/f?v=gs://figurl/spikesortingview-1&d=a5128490421d100de4f469ac[…]e042e2e&channel=flatiron1&label=Jaq_03_12_visualization_data2

[jsoules]

Looks like more is computed at the TimeScrollView level than I realized! That's where the offset for the toolbar (and in the case of the raster plot, the unit labels) is accounted for. So there was a margin alteration that I wasn't realizing.

Submitting a new PR on the same branch that should correct this issue.

In the process, I also discovered that I was painting the highlight bars too far downward in the composite view. This is also fixed.

[jsoules]

Example: Click on the 40-minute mark in https://www.figurl.org/f?v=gs://figurl/spikesortingview-1&d=a5128490421d100de4f469ac%5B%E2%80%A6%5De042e2e&channel=flatiron1&label=Jaq_03_12_visualization_data2 and zoom in. The first interval to the left of the 40-minute mark will move to the right (opposite of the direction it should move for a proper zoom) and eventually pass the 40-minute mark.

This corresponds to a span that begins at second 16799.615 and ends at 16799.732.

This example is now fixed.

[jsoules]

Closed by SpikeSortingView PRs #33 and #34.