BUG display.max_colwidth do not accept -1 for unlimited width

[glemaitre]

Code Sample, a copy-pastable example if possible

import pandas as pd
pd.set_option("display.max_colwidth", -1)

Problem description

There is a regression with "display.max_colwidth". In the past, it was only accepting integer. The way to no limit the size was to pass -1. In pandas 1,0, this option becomes more consistent and not limiting the width should be set with None. However, the support for negative integer was removed.

Thus, one would need to either set to -1 or None depending on the pandas version. It would be best to support both options. Potentially, support for negative integer could be removed with a deprecation cycle.

[MarcoGorelli]

It would be best to support both options. Potentially, support for negative integer could be removed with a deprecation cycle.

@jorisvandenbossche Is this what we're going for then - add support for -1 back in, but raise a FutureWarning alerting the user that it'll no longer work in a future version?

It reminds me of the header argument in read_csv, whereby header=-1 used to work in 0.24, then in 0.25 it didn't (see #27779) and so we now raise

ValueError: Passing negative integer to header is invalid. For no header, use header=None instead

Perhaps the current ValueError message could just have an extra line specifically instructing to use None instead of -1, in the case that -1 is passed?

[glemaitre]

I made a PR to reverse the support and add a deprecation cycle.