BUG: multi-index output formatting is buggy (GH7174)

[bjonen]

Closes #7174

Adjusts the truncation in the notebook for multiindex dfs. Both rows and columns are affected.

http://nbviewer.ipython.org/gist/bjonen/492fea9559fd73edf579

[jreback]

can u show the results for the test cases
(df and df.T) with mi on an axis

[bjonen]

What do you mean by mi on an "axis"? I added mi on both index and columns in the notebook.

[jreback]

that's what I meant (has at least a multi index) on 1 axis

[jreback]

@cpcloud @jorisvandenbossche @TomAugspurger ?

[bjonen]

Seems like multiindex levels > 2 wasnt as bad as I thought (see updated notebook). I'll add some more tests to the notebook to make sure this is working.

[cpcloud]

looking good

[cpcloud]

bit o' merge conflicts there @bjonen let us know if u need help

[cpcloud]

looks like on the last 3 level MI there's a row missing in the last column

[bjonen]

Thx @cpcloud , fixed it.

[cpcloud]

nice. looking pretty good. just squash down and should be good to go. maybe put up a final notebook with all the fixes that we can use as a reference for the issue.

[bjonen]

Ok will merge and squash.

You mean add the notebook from above to the repo? If so, where?

[cpcloud]

no i just meant update the notebook, which it looks like you did, so nevermind. 👍

[bjonen]

ok great

[bjonen]

Added some more cases to the notebook and squashed. If you spot something strange, let me know, otherwise this is ok from my side.

[jreback]

Worth it to add a MultiIndex example In whatsnew? and/or docs/FAQ (where the display options are)

[bjonen]

For me it's fine like this. I think most users will just assume central truncation works for the different types of indices and backends (terminal, notebook).

But if you think it's better to showcase the way truncation works more, I'm happy to add an example.