px.imshow requires int for facet_col while px.scatter can be str

px.imshow produces an error if the dimension of a facet plot has strings as their labels instead of integers. I ran across this with xarray, however, I believe it is present regardless of the input data structure. The other px graphs, e.g. px.scatter, do allow strings in the facet dimension, so this seems like a bug with imshow. https://plotly.com/python/facet-plots/ has many plotting examples with strings as the facet_col, e.g. "sex=Female" and "sex=Male" from fig = px.scatter(df, x="total_bill", y="tip", color="smoker", facet_col="sex") and many others from px.histogram, px.choropleth, etc.

Versions in use:
python=3.9.1
plotly=4.14.3

Here is an example of the error.

>>> import numpy as np
>>> import xarray as xr
>>> import plotly.express as px
>>> row = xr.Variable('img_row', range(4))
>>> col = xr.Variable('img_col', range(4))
>>> fd = xr.Variable('facet_dim', 'QA QB'.split())
>>> a = [xr.DataArray(np.random.rand(4,4), coords=[row, col]) for i in range(2)]
>>> da_str = xr.concat(a, "facet_dim")
>>> da_var = xr.concat(a, fd)
>>> da_str
<xarray.DataArray (facet_dim: 2, img_row: 4, img_col: 4)>
array([[[0.17020614, 0.23346368, 0.55878844, 0.58773312],
        [0.80092718, 0.25899341, 0.12681188, 0.77129175],
        [0.85068779, 0.48145364, 0.72109667, 0.51325248],
        [0.2659319 , 0.62613397, 0.09588715, 0.59035821]],

       [[0.32991787, 0.59718269, 0.60165123, 0.63523444],
        [0.11699783, 0.19545503, 0.22478829, 0.56217593],
        [0.59725   , 0.34207063, 0.29841437, 0.30079022],
        [0.21422747, 0.74626584, 0.86025186, 0.61071694]]])
Coordinates:
  * img_row  (img_row) int32 0 1 2 3
  * img_col  (img_col) int32 0 1 2 3
Dimensions without coordinates: facet_dim
>>> da_var
<xarray.DataArray (facet_dim: 2, img_row: 4, img_col: 4)>
array([[[0.17020614, 0.23346368, 0.55878844, 0.58773312],
        [0.80092718, 0.25899341, 0.12681188, 0.77129175],
        [0.85068779, 0.48145364, 0.72109667, 0.51325248],
        [0.2659319 , 0.62613397, 0.09588715, 0.59035821]],

       [[0.32991787, 0.59718269, 0.60165123, 0.63523444],
        [0.11699783, 0.19545503, 0.22478829, 0.56217593],
        [0.59725   , 0.34207063, 0.29841437, 0.30079022],
        [0.21422747, 0.74626584, 0.86025186, 0.61071694]]])
Coordinates:
  * img_row    (img_row) int32 0 1 2 3
  * img_col    (img_col) int32 0 1 2 3
  * facet_dim  (facet_dim) <U2 'QA' 'QB'
>>> fig = px.imshow(da_str, facet_col='facet_dim', aspect='equal')
>>> [d.__class__.__name__ for d in fig.data]
['Heatmap', 'Heatmap']
>>> fig = px.imshow(da_var, facet_col='facet_dim', aspect='equal')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Users\<NAME>\Miniconda3\envs\jupyter\lib\site-packages\plotly\express\_imshow.py", line 525, in imshow
    col_labels = ["%s=%d" % (slice_label, i) for i in facet_slices]
  File "C:\Users\<NAME>\Miniconda3\envs\jupyter\lib\site-packages\plotly\express\_imshow.py", line 525, in <listcomp>
    col_labels = ["%s=%d" % (slice_label, i) for i in facet_slices]
TypeError: %d format: a number is required, not numpy.str_

I think the culprit is the different string conversions in play. Can the px.imshow col_labels use a more forgiving conversion?
From File "C:\Users\<NAME>\Miniconda3\envs\jupyter\lib\site-packages\plotly\express\_imshow.py", line 525, in imshow
col_labels = ["%s=%d" % (slice_label, i) for i in facet_slices]
From File "C:\Users\<NAME>\Miniconda3\envs\jupyter\lib\site-packages\plotly\express\_core.py", line 1885, in make_figure
col_labels = [prefix + str(s) for s in sorted_group_values[m.grouper]]

Commencez dans plotly/express/_imshow.py à la ligne signalée 525, où les libellés de facet sont formatés, et comparez cela avec la gestion des libellés présentée dans plotly/express/_core.py. Reproduisez l’exemple px.imshow avec un facet_dim dont les valeurs sont des chaînes, puis vérifiez que les libellés de facet sous forme de chaînes ne lèvent plus de TypeError et produisent toujours les facets de heatmap attendus.