Spans and Lines don’t support half-open hard ranges

[alcrene]

ALL software version info

HoloViews: 1.18.1
Pandas: 2.1.4
NumPy: 1.26.3
Bokeh: 3.3.3
Matplotlib: 3.8.2
Python: 3.10.12

Description of error

The *Spans and *Lines annotations don’t support hard ranges unless both ends are specified.
Singular versions (*Span and *Line) are fine.

The issue seems due to converting extents to a Pandas array:

holoviews/holoviews/plotting/bokeh/annotation.py

Lines 70 to 82 in e243fe0

	def get_extents(self, element, ranges=None, range_type='combined', **kwargs):
	extents = super().get_extents(element, ranges, range_type)
	if isinstance(element, HLines):
	extents = np.nan, extents[0], np.nan, extents[2]
	elif isinstance(element, VLines):
	extents = extents[0], np.nan, extents[2], np.nan
	elif isinstance(element, HSpans):
	extents = pd.array(extents)
	extents = np.nan, extents[:2].min(), np.nan, extents[2:].max()
	elif isinstance(element, VSpans):
	extents = pd.array(extents)
	extents = extents[:2].min(), np.nan, extents[2:].max(), np.nan
	return extents

which converts nan to pandas.NA. Then these lines filter out NaNs, but not NAs:

holoviews/holoviews/core/util.py

Lines 1091 to 1092 in e243fe0

	lower = [v for v in arr[lidx] if v is not None and not is_nan(v)]
	upper = [v for v in arr[uidx] if v is not None and not is_nan(v)]

A search for pd.array(extents) shows this is done in both the Bokeh and Matplotlib renderers.
I also tested that the error occurs with both. I did not test Plotly.

Complete, minimal, self-contained example code that reproduces the issue

The following will raise the exception printed above

import holoviews as hv
hv.extension("bokeh")

dim_prob = hv.Dimension('p', label="prob", range=(0, None))
fig = hv.Curve([(0, 0.6), (1, 0.3), (2, 0.4), (3, 0.45)], kdims="x", vdims=dim_prob) \
      * hv.HSpans([(0, .2), (0.4, 0.6)], kdims=["x", dim_prob])
display(fig)   # Error occurs when the figure is displayed

With any of the following changes, the example does work:

• range=(0, 1) instead of range=(0, None)
• soft_range=(0, None) instead of range=(0, None)
• HSpan instead of HSpans

Possible workaround

Use soft_range for half-open ranges.

Stack traceback and/or browser JavaScript console output

Last bit of traceback:

...
File [~/usr/venv/comp/lib/python3.10/site-packages/holoviews/plotting/plot.py:1988](http://localhost:33091/home/rene/usr/venv/comp/lib/python3.10/site-packages/holoviews/plotting/plot.py#line=1987), in GenericOverlayPlot.get_extents(self, overlay, ranges, range_type, dimension, **kwargs)
   1986 subplot_extents = self._get_subplot_extents(overlay, ranges, range_type, dimension)
   1987 zrange = isinstance(self.projection, str) and self.projection == '3d'
-> 1988 extents = {k: util.max_extents(rs, zrange) for k, rs in subplot_extents.items()}
   1989 if range_type != 'combined':
   1990     return extents[range_type]

File [~/usr/venv/comp/lib/python3.10/site-packages/holoviews/plotting/plot.py:1988](http://localhost:33091/home/rene/usr/venv/comp/lib/python3.10/site-packages/holoviews/plotting/plot.py#line=1987), in <dictcomp>(.0)
   1986 subplot_extents = self._get_subplot_extents(overlay, ranges, range_type, dimension)
   1987 zrange = isinstance(self.projection, str) and self.projection == '3d'
-> 1988 extents = {k: util.max_extents(rs, zrange) for k, rs in subplot_extents.items()}
   1989 if range_type != 'combined':
   1990     return extents[range_type]

File [~/usr/venv/comp/lib/python3.10/site-packages/holoviews/core/util.py:1121](http://localhost:33091/home/rene/usr/venv/comp/lib/python3.10/site-packages/holoviews/core/util.py#line=1120), in max_extents(extents, zrange)
   1119 for lidx, uidx in inds:
   1120     lower = [v for v in arr[lidx] if v is not None and not is_nan(v)]
-> 1121     upper = [v for v in arr[uidx] if v is not None and not is_nan(v)]
   1122     if lower and isinstance(lower[0], datetime_types):
   1123         extents[lidx] = np.min(lower)

File [~/usr/venv/comp/lib/python3.10/site-packages/holoviews/core/util.py:1121](http://localhost:33091/home/rene/usr/venv/comp/lib/python3.10/site-packages/holoviews/core/util.py#line=1120), in <listcomp>(.0)
   1119 for lidx, uidx in inds:
   1120     lower = [v for v in arr[lidx] if v is not None and not is_nan(v)]
-> 1121     upper = [v for v in arr[uidx] if v is not None and not is_nan(v)]
   1122     if lower and isinstance(lower[0], datetime_types):
   1123         extents[lidx] = np.min(lower)

File missing.pyx:419, in pandas._libs.missing.NAType.__bool__()

TypeError: boolean value of NA is ambiguous

[github-actions]

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