Add Pandas 2.0 support

[hoxbro]

This PR adds support for Pandas 2.0 and fixes a bug and an inconsistency.

Changes made in this PR:

• Internally holoviews/operation/datashader now uses datetime[ns] instead of datetime[us], for reasons see the comment below.
• The changes in the notebooks are simple changes where an argument has been changed to a keyword-only argument.
• Fixes Rasterize not working in combination with multiple rectangles with DateTime coordinates #5239.
• Fixes an inconsistency where dt_to_int could return a float in certain cases.
• Update setup.py/setup.cfg to work with Pandas 2.
• Add testing to both Pandas 2.0 and Pandas 1 in test CI.
• Move the logic of examples to ignore into examples/conftest,py.

Reference:

• Pandas 2.0 whats new

[codecov-commenter]

Codecov Report

Merging #5662 (d2ec764) into main (04ba54a) will increase coverage by 0.00%.
The diff coverage is 68.57%.

@@           Coverage Diff           @@
##             main    #5662   +/-   ##
=======================================
  Coverage   88.29%   88.29%           
=======================================
  Files         302      302           
  Lines       62576    62586   +10     
=======================================
+ Hits        55253    55263   +10     
  Misses       7323     7323           

Impacted Files	Coverage Δ
holoviews/operation/datashader.py	83.81% <52.17%> (+0.02%)	⬆️
holoviews/core/util.py	86.60% <100.00%> (ø)
holoviews/element/util.py	82.21% <100.00%> (ø)
holoviews/tests/core/data/test_daskinterface.py	88.09% <100.00%> (+1.42%)	⬆️
holoviews/tests/plotting/bokeh/test_callbacks.py	98.65% <100.00%> (-0.01%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[hoxbro]

The datetime change is because Pandas is now beginning to support non-nanosecond precision values:

import pandas as pd
from datetime import datetime

s0 = pd.Series([datetime(2020, 1, 1)])
s1 = s0.astype("datetime64[us]")
s2 = s0.astype("int64")
s3 = s1.astype("int64")


print(pd.__version__)

for s in [s0, s1, s2, s3]:
    print(f"{s[0]}, {s.dtype}")
    print(f"{s.values[0]}, {s.values.dtype}")
    print()

Outputs:

1.5.3
2020-01-01 00:00:00, datetime64[ns]
2020-01-01T00:00:00.000000000, datetime64[ns]

2020-01-01 00:00:00, datetime64[ns]
2020-01-01T00:00:00.000000000, datetime64[ns]

1577836800000000000, int64
1577836800000000000, int64

1577836800000000000, int64
1577836800000000000, int64

2.0.0rc0
2020-01-01 00:00:00, datetime64[ns]
2020-01-01T00:00:00.000000000, datetime64[ns]

2020-01-01 00:00:00, datetime64[us]
2020-01-01T00:00:00.000000, datetime64[us]

1577836800000000000, int64
1577836800000000000, int64

1577836800000000, int64
1577836800000000, int64

Also beginning to get this warning from xarray because of this:

/home/shh/Repos/holoviz/holoviews/holoviews/operation/datashader.py:848: UserWarning: Converting non-nanosecond precision datetime values to nanosecond precision. This behavior can eventually be relaxed in xarray, as it is an artifact from pandas which is now beginning to support non-nanosecond precision values. This warning is caused by passing non-nanosecond np.datetime64 or np.timedelta64 values to the DataArray or Variable constructor; it can be silenced by converting the values to nanosecond precision ahead of time.
  agg[ydim] = (agg[ydim]/1e3).astype('datetime64[us]')

Can be recreated with:

import datetime as dt

import holoviews as hv
from holoviews.operation.datashader import rasterize

rects = hv.Rectangles([(0, dt.datetime(2016, 1, 2), 4, dt.datetime(2016, 1, 3))])
agg = rasterize(rects, width=4, height=4, dynamic=False)

[hoxbro]

xarray does not yet support pandas 2.0: pydata/xarray#7716

[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.