CLN FloatArrayFormatter

[nbonnotte]

I'm working on #12164

This is a work in progress.

[nbonnotte]

Ok, that should do.

What I have done:

• Instead of differentiating on the one hand FloatArrayFormatter.get_result (which comes from GenericArrayFormatter), and on the other hand FloatArrayFormatter.get_formated_data, according to their destination (two usages, two functions, two independent pieces of code), I differentiate with the type of the output: .get_result keeps returning a list like before, and the new .get_result_as_array returns a numpy array like .get_formatted_data used to.
• The question "do we want a fixed-width output?", which is a formatting question, is translated like all formatting matters into a parameter fixed_width for the GenericArrayFormatter constructor (boolean, default True, because that's the main usage as of today)

This seems (to me) the simplest way to know which method should be used.

I have merged the two independent pieces of code that we used to have, and which were doing basically the same thing. But FloatArrayFormatter.get_formated_data was more refined, as it was using a mask to deal separately with the NaN values. I've kept that approach, but I don't know if it is the best option in term of performances.

[jreback]

you can always call this, to_dense() is a generic method (doesn't do anything to non-dense things)

[nbonnotte]

AttributeError: 'numpy.ndarray' object has no attribute 'to_dense'

:)

[nbonnotte]

All green

[jreback]

this looked ok last time. pls rebase and repush.

[nbonnotte]

@jreback all set

[jreback]

does this actually change any features? e.g. implement the decimal fix? (or is that another PR), this just cleanup?

[jreback]

thanks!

[nbonnotte]

No, that was just a cleanup, that's why I haven't touched the tests. Now I should be able to move on to the decimal parameter (issue #12031)