ENH?: __repr__ method for Index objects

[dengemann]

It often happens that I print a MultiIndex object (whether on purpose or not) which leads to flooding my interpreter and makes it unresponsive for a couple of seconds since often the length of my indexers exceeds 500-1000k. I was thinking about adding a neat __repr__ for Indexers that summarizes levels + length or just omits what's in the middle once a certain length is exceeded. It seems this has not been tackled so far when looking at the current master, but please feel free to close this issue if this is a dupe (checked but did not find anything related). Any pointers are welcome.

[jtratner]

I think you need to fix the Index representation of itself instead, because levels and labels are just Index objects that are getting pprinted in an extremely long way, here's the relevant code to intercept:

    def __repr__(self):
        encoding = get_option('display.encoding')
        attrs = [('levels', default_pprint(self.levels)),
                 ('labels', default_pprint(self.labels))]
        if not all(name is None for name in self.names):
            attrs.append(('names', default_pprint(self.names)))
        if self.sortorder is not None:
            attrs.append(('sortorder', default_pprint(self.sortorder)))

If you want to put something together that shortens extremely long indexes, you just need to alter FrozenNDArray.__str__ (and/or __repr__) in pandas/core/base.py. (because MI repr just pprints levels, labels, etc). There's been previous discussion about maximum sequence length that would go into this too.

[dengemann]

Thanks @jtratner I spotted this in index.py --- actually my idea was to modify this one. Thanks for the FrozenNDArray hint. I'm happy to look into this if the idea is appreciated ;-)

[jtratner]

I think the correct config option is display.max_seq_items. You can probably just pprint slices to make it match the length correctly.

[cpcloud]

fwiw you'll also be able adjust the "edge" items once i push that PR

[jreback]

@jtratner @cpcloud close this?

[jtratner]

I think it's reasonable. It just needs to be changed to what it actually is
"Make Index repr respect max_seq_len" or whatever the appropriate
option is.

[dengemann]

If we're not in a hurry I'm happy to look into this. Won't be before the weekend though, no time ....

[jtratner]

Push the milestone to 0.14. If you finish earlier we can incorporate it.

xref #3391 , #3347

[cancan101]

Also see #5734 I think

[jreback]

@y-p I belive this is should be closed?
@jtratner (because of fix to max_seq_items)

yep, I just tested post-#5753 and it limits the multiindex output (to something large) which
doesn't hang the terminal.
Also see that #4935 handled the issues I raised in #3347 a while back. great.