bug: animation indexing error

[jeremymanning]

It looks like there is an offset issue in the animation code whereby:

• When groups isn't specified, the first point in the group is transparent, even when chemtrails=False and precog=False.
• When groups is specified, and IndexError is raised

To replicate this issue:

import numpy as np
import hypertools as hyp

x = np.random.normal(np.ones([10, 5])) #generate some data

#things that work:
hyp.plot(x, 'o') #plotted correctly
hyp.plot(x, animate='spin') #plotted correctly

#things that don't work:
hyp.plot(x, 'o', animate='spin') #this runs, but the first dot is more transparent than the others, whereas everything should look the same

hyp.plot(x, 'o', animate='spin', group=np.arange(10)) #raises IndexError-- should plot each dot in a different color

[jeremymanning]

Note: to clarify the above example, this seems to be specific to when the marker style is specified, e.g. 'o' or similar.

Another weird behavior: when the line style is 'o-', it should look like the default plot (i.e. with smooth interpolation, as if the line style was '-', overlaid on a plot with line style of 'o'-- so with additional markers at the datapoints). instead, there's no interpolation, so straight lines connect each 'o' marker.

[andrewheusser]

hmmm, i'll look into this soon!

[jeremymanning]

Good news -- the remaining 'o-' interpolation gap from this thread is now fixed in the HyperTools 1.0 update (PR #272, #272).

Root cause: the line-smoothing interpolation step was gated on is_line(fmt), which is False for any format string containing a marker character -- so 'o-' skipped interpolation entirely and drew straight, unsmoothed segments between the raw points, while '-' alone drew the smoothed curve on identical data.

Before (still true on an unpatched checkout): 'o-' renders straight segments between raw points; '-' renders a smoothed spline through the same points.

After:

import numpy as np, hypertools as hyp
x = np.cumsum(np.random.default_rng(0).standard_normal((20, 3)), axis=0)

fig_dash = hyp.plot([x], '-', show=False)
fig_combo = hyp.plot([x], 'o-', show=False)

line = next(l for l in fig_combo.axes[0].get_lines() if l.get_linestyle() != 'None')
marker = next(l for l in fig_combo.axes[0].get_lines() if l.get_linestyle() == 'None')

# smoothed line now matches the pure '-' style's vertex count exactly
assert line.get_data_3d()[0].shape == fig_dash.axes[0].get_lines()[0].get_data_3d()[0].shape
# markers stay anchored to the true (pre-interpolation) sample points
assert marker.get_data_3d()[0].shape[0] == 20

'o-' now draws two artists: a smoothed/interpolated line (matching '-''s own smoothing exactly) plus markers at the original, un-interpolated sample points -- so markers never drift onto the dense interpolated curve. This applies to the static matplotlib backend; animated plots and the plotly backend get the same line-smoothing fix, though their markers still render at the interpolated points rather than only the original samples (documented as a follow-up in plot()'s docstring).

Closing as addressed in the 1.0 update -- please reopen if you hit anything else.