API to style categorical variables

[jnothman]

It is often useful to style columns having a pd.CategoricalDtype, by either:

• being supplied a mapping from category name/code to background colour.
• being supplied a mapping from category name/code to Styler.
• selecting len(dtype.categories) colours from a matplotlib colormap (qualitative or otherwise).

So for example we might allow for:

StyleFrame(df).add_categorical_conditional_formatting(style_map={"a": Styler(font_color="red"), "b": Styler(font_color="green"), "c": Styler(font_color="blue")})

or

StyleFrame(df).add_categorical_conditional_formatting(style_bg_cmap="Pastel1"})

which would perform the following for each column

cmap = matplotlib.cm.get_cmap(style_bg_cmap)
col_style_map = dict(zip(col.dtype.categories, cmap(len(col.dtype.categories))))

Conditional formatting would be applied using a CellIsRule("=", category) for each category value in each column in range.

[DeepSpace2]

Can this not be achieved with the already available methods (albeit in a more explicit way)?

[jnothman]

I find this feature is possible but hard to achieve with both excel and StyleFrame. It takes several steps and is hard to find/imagine for the user. I think of StyleFrame not merely as facilitating the functionality it can provide, but also: * Providing a mapping between pandas concepts and excel concepts * Providing visibility to excel features that a pandas user might be interested in when presenting a dataframe to a spreadsheet user. In this vein, making it easier to style pandas Categoricals (or generally discrete values) makes sense. But maybe I'm just missing which existing API you refer to.

[jnothman]

How would you recommend styling categorical variables with the current API, @DeepSpace2?

[DeepSpace2]

Are you able to provide an example pd.DataFrame and the output sheet that you envision

StyleFrame(df).add_categorical_conditional_formatting(style_map={"a": Styler(font_color="red"), "b": Styler(font_color="green"), "c": Styler(font_color="blue")})

will output?

[jnothman]

Let's say that

style_map = {"a": Styler(font_color="red"), "b": Styler(font_color="green"), "c": Styler(font_color="blue")}
sf.add_categorical_conditional_formatting(style_map=style_map)

translates roughly to:

sf = StyleFrame(df)
for value, styler in style_map.items():
    sf = sf.apply_style_by_indexes(sf == value, styler)

[jnothman]

Ideally, however, it would be implemented with conditional formatting so that the styles do not need to be stored on each cell.

[jnothman]

And ideally, provisions would be made to use the categorical dtype's cardinality to select from a colour palette...

[DeepSpace2]

Ah, I see.

The current API does support conditional formatting, but only using ColorScaleRule (wrapped by StyleFrame.add_color_scale_conditional_formatting).

However, the above example can be implemented without it, using quite an explicit code:

style_map = {"a": Styler(font_color="red"), "b": Styler(font_color="green"), "c": Styler(font_color="blue")}
sf = StyleFrame({'1': ['a', 'b'],
                 '2': ['c', 'd']})
for value, styler in style_map.items():
    for col in sf.columns:
        sf.apply_style_by_indexes(sf[sf[col] == value], styler, cols_to_style=col)

generating

[jnothman]

I agree it's not hard to do; the point is to support:

• increased visibility of this styling approach for discrete-valued columns.
• using a matplotlib colormap or other arbitrary palette where the user is ambivalent about the specific colors used.

[stale]

This issue has been automatically marked as stale because it has not had activity in the last 60 days.