REF: pandas/io/formats/format.py

[ivanovmg]

Is your feature request related to a problem?

class DataFrameFormatter seems mostly to target console formatting, while it plays a role of the mediator for to_latex and to_html conversions.
It is noticeable, that some kwargs, like table_id, render_links are pertaining to html only. Meanwhile dataframe truncation (present in init) is not applicable to latex formatting.
Probably needs refactoring.

Describe the solution you'd like

I suggest creating TableFormatterAbstract class.
Inherit from it: ConsoleFormatter, HTMLFormatter and LatexFormatter.
Put all the required input parameters into the corresponding init of each class (would not need to provide unnecessary kwargs).
Generic functionality like dataframe truncation can be shared via a mixin class TruncateMixin.
Drop DataFrameFormatter and instead of it call proper XFormatter where required.

API breaking implications

Should not affect public API, as will deal only with the internal implementation.

Describe alternatives you've considered

Separate ConsoleFormatter, similarly to HTMLFormatter and LatexFormatter and call it from DataFrameFormatter.to_string method.
This, however will leave DataFrameFormatter with only three public methods to_string, to_html and to_latex.
The problem here is that we still carry unnecessary kwargs, which are not applicable to a particular formatting.

Additional context

I would be interested to work on this. Would the team be interested in such refactoring?

[jreback]

sure glad to take internal refactorings here to make things more clear. only request is in general smaller PR's are better. If moving code only do it separately as well. e.g. first move / re-org, then follow with changes, otherwise very hard to tell the diffs.

[jbrockmendel]

@attack68 i think you're the most qualified to comment on this: is this issue still actionable or has it been refactored as much as it reasonably can be?

[attack68]

I removed the LatexFormatter and the to_latex method in #52844 , after re-routing DataFrame.to_latex via Styler.to_latex.

I think the end goal is really to do the same for HTML, i.e. rending it via Styler (and then remove HTMLFormatter), but it was difficult to get this through with disagreements on deprecation cycles and not be able to replicate every single keyword arg, etc.

If this did actually happen it would not leave this Issue with much to do really. So this Issue is probably still actionable but I think those actions might be quite different to what the OP actually suggests. Best to close?