pl.render_labels cannot render when color is mapped to a categorical variable. Returns "ValueError: buffer source array is read-only"

[MiTPenguin]

In general, sd.pl.render_labels returns a ValueError if color is set to a categorical variable in the attached anndata.obs table

Using the given cosmx data in the tutorial:

cosmx_sdata.pl.render_labels(elements = "2_labels", outline = True, color = 'Area', cmap = 'magma').pl.show(coordinate_systems=["2"])

gives the expected output: segmentation map colored by size of the area in the 'magma' color map

However, the following:

cosmx_sdata.pl.render_labels(elements = "2_labels", outline = True, color = 'fov_labels').pl.show(coordinate_systems=["2"])

Does not return the expected plot (segmentation map with a single color to represent the fov label). Instead it returns the following:

ValueError Traceback (most recent call last)
Cell In[418], [line 1](vscode-notebook-cell:?execution_count=418&line=1)
----> [1](vscode-notebook-cell:?execution_count=418&line=1) cosmx_sdata.pl.render_labels(color = 'fov_labels').pl.show(coordinate_systems=["2"])

File ~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:937, in PlotAccessor.show(self, coordinate_systems, legend_fontsize, legend_fontweight, legend_loc, legend_fontoutline, na_in_legend, colorbar, wspace, hspace, ncols, frameon, figsize, dpi, fig, title, share_extent, pad_extent, ax, return_ax, save)
 [925](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:925) _maybe_set_colors(
 [926](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:926) source=sdata[table],
 [927](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:927) target=sdata[table],
 [928](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:928) key=params_copy.color,
 [929](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:929) palette=params_copy.palette,
 [930](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:930) )
 [932](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:932) rasterize = (params_copy.scale is None) or (
 [933](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:933) isinstance(params_copy.scale, str)
 [934](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:934) and params_copy.scale != "full"
 [935](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:935) and (dpi is not None or figsize is not None)
 [936](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:936) )
--> [937](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:937) _render_labels(
 [938](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:938) sdata=sdata,
 [939](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:939) render_params=params_copy,
 [940](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:940) coordinate_system=cs,
 [941](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:941) ax=ax,
 [942](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:942) fig_params=fig_params,
 [943](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:943) scalebar_params=scalebar_params,
 [944](https://file+.vscode-resource.vscode-cdn.net/Users/a.tu/dev_env/Translational_Analysis_Multi-program_mIF/~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:944) legend_params=legend_params,
...
File <stringsource>:663, in View.MemoryView.memoryview_cwrapper()

File <stringsource>:353, in View.MemoryView.memoryview.__cinit__()

ValueError: buffer source array is read-only

identical error is found using other spatialdata objects (public, or custom constructed)

Environment is on m2 macbook pro, on python 3.11

spatialdata is version 0.2.2
xarray is version 2023.12.0
skimage is version 0.24.9
anndata is version 0.10.8
pandas is version 2.2.2

[clwgg]

which version of spatialdata-plot are you using? some of this was "fixed" recently by copying the input array during plotting, although a proper fix within skimage would be preferable; see also the discussion here: #344

[MiTPenguin]

spatialdata-plot is 0.2.4

I tried the saving to a zarr and re-read, but that did not work.

[MiTPenguin]

I installed the version of skimage with the Cython update, and now I have a different error

{
	"name": "ValueError",
	"message": "all the input arrays must have same number of dimensions, but the array at index 0 has 4 dimension(s) and the array at index 1 has 3 dimension(s)",
	"stack": "---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[11], line 1
----> 1 cosmx_sdata.pl.render_labels(elements = \"2_labels\", outline = True, color = 'fov_labels', cmap = 'magma').pl.show(coordinate_systems=[\"2\"])

File ~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/basic.py:937, in PlotAccessor.show(self, coordinate_systems, legend_fontsize, legend_fontweight, legend_loc, legend_fontoutline, na_in_legend, colorbar, wspace, hspace, ncols, frameon, figsize, dpi, fig, title, share_extent, pad_extent, ax, return_ax, save)
 925 _maybe_set_colors(
 926 source=sdata[table],
 927 target=sdata[table],
 928 key=params_copy.color,
 929 palette=params_copy.palette,
 930 )
 932 rasterize = (params_copy.scale is None) or (
 933 isinstance(params_copy.scale, str)
 934 and params_copy.scale != \"full\"
 935 and (dpi is not None or figsize is not None)
 936 )
--> 937 _render_labels(
 938 sdata=sdata,
 939 render_params=params_copy,
 940 coordinate_system=cs,
 941 ax=ax,
 942 fig_params=fig_params,
 943 scalebar_params=scalebar_params,
 944 legend_params=legend_params,
 945 rasterize=rasterize,
 946 )
 948 if title is None:
 949 t = cs

File ~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/render.py:779, in _render_labels(sdata, render_params, coordinate_system, ax, fig_params, scalebar_params, legend_params, rasterize)
 765 color_source_vector, color_vector, categorical = _set_color_source_vec(
 766 sdata=sdata_filt,
 767 element=label,
 (...)
 774 table_name=table_name,
 775 )
 777 if (render_params.fill_alpha != render_params.outline_alpha) and render_params.contour_px is not None:
 778 # First get the labels infill and plot them
--> 779 labels_infill = _map_color_seg(
 780 seg=label.values,
 781 cell_id=instance_id,
 782 color_vector=color_vector,
 783 color_source_vector=color_source_vector,
 784 cmap_params=render_params.cmap_params,
 785 seg_erosionpx=None,
 786 seg_boundaries=render_params.outline,
 787 na_color=render_params.cmap_params.na_color,
 788 )
 789 _cax = ax.imshow(
 790 labels_infill,
 791 rasterized=True,
 (...)
 796 zorder=render_params.zorder,
 797 )
 798 _cax.set_transform(trans_data)

File ~/miniforge3/envs/spatial/lib/python3.11/site-packages/spatialdata_plot/pl/utils.py:781, in _map_color_seg(seg, cell_id, color_vector, color_source_vector, cmap_params, seg_erosionpx, seg_boundaries, na_color)
 779 seg = np.squeeze(seg, axis=0)
 780 seg_bound: ArrayLike = np.clip(seg_im - find_boundaries(seg)[:, :, None], 0, 1)
--> 781 return np.dstack((seg_bound, np.where(val_im > 0, 1, 0))) # add transparency here
 783 if len(val_im.shape) != len(seg_im.shape):
 784 val_im = np.expand_dims((val_im > 0).astype(int), axis=-1)

File ~/miniforge3/envs/spatial/lib/python3.11/site-packages/numpy/lib/shape_base.py:715, in dstack(tup)
 713 if not isinstance(arrs, list):
 714 arrs = [arrs]
--> 715 return _nx.concatenate(arrs, 2)

ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 4 dimension(s) and the array at index 1 has 3 dimension(s)"
}

This time it's a dimension issue with numpy.dstack. Again data replicated using cosmx example data.

[timtreis]

Could you try upgrading to 0.2.6 and verify that the issue persists?

[MiTPenguin]

@timtreis

Upgraded to 0.2.6 and the hotfixed skimage (installed from git) seems to have solved the problem (I also updated cython in my conda environment, for good measure)

Also fixed some odd behavior with color mapping of discrete integer labels.

[timtreis]

Happy to hear that!