REF: type change escape in Styler.format to str to allow "html" and "latex"

[attack68]

Change is made ahead of 1.3.0 to pre-empt the addition of escape_latex and to avoid name ambiguity.

This is not user facing since the escape kwarg was added for 1.3.0 only recently.

Alternative

If instead of having separate escape_html and escape_latex kwargs we could have escape={None, 'html', 'latex'} since it is likely they will be used exclusively??

[jreback]

can u move this to the to_html or to_latex method instead?

[attack68]

can u move this to the to_html or to_latex method instead?

no, because it's mixed in as part of the formatting routine, and the idea is that DataFrames should be formatted before calling to_html or to_latex, i.e.

df.style.format(...).to_xxx()

if escape is added to the render method then format needs to be called within the render method overwriting the previous chained application and negating it all and restricting its flexibility.

what about expanding the keyword inputs to:

escape=None,
escape='html'
escape='latex'
escape=['html', 'latex']

?

[jreback]

yeah the last is fine i guess

but is there a way to just escape things that work in both html and latex?

[attack68]

yeah the last is fine i guess

but is there a way to just escape things that work in both html and latex?

not currently because of operation order. the format method is ignorant of how the styler will ultimately be rendered, and it yields a one-arg-callable for each data element decoupled from render method. if it becomes dependent it introduces operation order complexity and possible performance degradation.

(and of course html and latex have different escaped chars)

will propose the former alternative and can review again.

[attack68]

@jreback it turns out some of the escaped characters overlap so it is impossible to escape 'html' and 'latex' simultaneously.

I've edited this PR to allow escape in {None, 'html', 'latex'} and added the necessary latex escape function and tests.

[jreback]

where is eacape_html?

[attack68]

imported from markupsafe which is a jinja2 dependency

[attack68]

it is a very simple function, can write here instead of having dependency to markupsafe if preferred?

[jreback]

oh its fine, just not completely obvious where it is (but its already there so its fine)

[jreback]

oh its fine, just not completely obvious where it is (but its already there so its fine)