Bars plot crashes with AttributeError when using datetime x-axis with single or duplicate values

[ghostiee-11]

Description

hv.Bars crashes with AttributeError: 'int' object has no attribute 'astype' when the datetime x-axis has either:

1. A single data point
2. All identical x-values

This happens because the bar width calculation in BarPlot.get_data() replaces xdiff (a NumPy array) with the plain Python int 1 when the diff array is empty or all-zero, but then tries to call .astype('timedelta64[ms]') on it — which only works on NumPy arrays.

Reproduction

import holoviews as hv
import pandas as pd

hv.extension('bokeh')

# Case 1: Single datetime bar — crashes
hv.render(hv.Bars([(pd.Timestamp('2024-01-01'), 5)]))

# Case 2: Duplicate datetime x-values — crashes
hv.render(hv.Bars([(pd.Timestamp('2024-01-01'), 5), (pd.Timestamp('2024-01-01'), 3)]))

# Works fine: multiple bars with distinct datetime values
hv.render(hv.Bars((pd.date_range('2024-01-01', periods=3), [5, 3, 7])))

# Works fine: single bar with numeric (non-datetime) x-axis
hv.render(hv.Bars([(1, 5)]))

Traceback

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "holoviews/util/__init__.py", line 878, in render
    return renderer_obj.get_plot_state(obj)
  File "holoviews/plotting/renderer.py", line 273, in get_plot_state
    obj = self_or_cls.get_plot(obj=obj, renderer=renderer, **kwargs)
  File "holoviews/plotting/bokeh/renderer.py", line 70, in get_plot
    plot = super().get_plot(obj, doc, renderer, **kwargs)
  File "holoviews/plotting/renderer.py", line 239, in get_plot
    plot.update(init_key)
  File "holoviews/plotting/plot.py", line 957, in update
    return self.initialize_plot()
  File "holoviews/plotting/bokeh/element.py", line 2221, in initialize_plot
    self._init_glyphs(plot, element, ranges, source)
  File "holoviews/plotting/bokeh/element.py", line 2121, in _init_glyphs
    data, mapping, style = self.get_data(element, ranges, style)
  File "holoviews/plotting/bokeh/chart.py", line 1065, in get_data
    width *= xdiff.astype('timedelta64[ms]').astype(np.int64)
             ^^^^^^^^^^^^
AttributeError: 'int' object has no attribute 'astype'

Screenshots

Full Traceback

Crash vs Working Comparison

Root Cause

In holoviews/plotting/bokeh/chart.py lines 1060-1065:

xdiff = np.abs(np.diff(xvals[xslice]))
diff_size = len(np.unique(xdiff))
if diff_size == 0 or (diff_size == 1 and xdiff[0] == 0):
    xdiff = 1                          # <-- becomes a plain int
if is_dt:
    width *= xdiff.astype('timedelta64[ms]').astype(np.int64)  # <-- crash: int has no .astype()

When there is only one x-value, np.diff() returns an empty array, so diff_size == 0 and xdiff is reassigned to the plain Python int 1. The subsequent .astype() call then fails because int doesn't have that method.

The non-datetime path (else branch on line 1067) also uses np.min(xdiff) on the scalar, but NumPy's np.min() handles scalars gracefully so that path doesn't crash.

Version Info

• HoloViews: main branch (1.23.0b1.dev)
• Bokeh: 3.x
• Python: 3.12
• OS: macOS

---

I've identified the root cause and have a fix ready. If this issue is confirmed, I'd be happy to open a PR with the fix along with test coverage for these edge cases.