Optimize downsample1d when data is shared between layers

[philippjfr]

Another optimization for downsample1d when multiple layers share the same data. Since the majority of time is spent slicing the data we can apply an optimization when all elements in an (Nd)Overlay that is being downsampled share the same underlying data and are indexed along the same key dimension(s).

In this case we simply get the first element slice it with the current x_range and then pass it down to each separate self._process call.

Here's the test case I was using:

import holoviews as hv
import numpy as np
import pandas as pd

from holoviews.operation.downsample import downsample1d

hv.extension('bokeh')

N = 1_000_000
index = pd._testing.makeDateIndex(k=N, freq='h', name='date')
df = pd.DataFrame({chr(65+i): np.random.randn(N).cumsum() for i in range(25)}, index=index).reset_index()

overlay = hv.NdOverlay({col: hv.Curve(df, 'date', col) for col in df.columns[1:]})

d = downsample1d(overlay).opts(responsive=True, height=300)
d

i = 2000
size_stream = d.streams[0]

%%timeit -r 10 -n 1
global i
i += 1
size_stream.event(width=i)

Results before:

1.41 s ± 60.8 ms per loop (mean ± std. dev. of 10 runs, 1 loop each)

Results after:

319 ms ± 40.8 ms per loop (mean ± std. dev. of 10 runs, 1 loop each)

Addresses part of #6058

Ping @droumis and @hoxbro

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (c800e52) 88.65% compared to head (0297598) 88.54%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6075      +/-   ##
==========================================
- Coverage   88.65%   88.54%   -0.12%     
==========================================
  Files         314      314              
  Lines       65942    65972      +30     
==========================================
- Hits        58462    58414      -48     
- Misses       7480     7558      +78     

Flag	Coverage Δ
ui-tests	?

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[hoxbro]

@philippjfr, this should be ready for review; I can't request you review it as you are the original author of the issue.

[philippjfr]

Can't review it, but your changes and the tests look good!

[philippjfr]

Probably would have used unittest.mock.patch on the Interface but this is fine.

[github-actions]

This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.