BUG/API: Catch exceptions in _ipython_display_

[TomAugspurger]

We raise an UnserializeableWarning when we fail to generate
the table schema and do not publish a table_schema repr.

closes #15996

[TomAugspurger]

cc @rgbkrk so users get a pandas.errors.UnserializableWarning (open to better names here) if there's something we can't handle. The data published won't contain a application/vnd.dataresource+json key.

[codecov]

Codecov Report

Merging #16132 into master will decrease coverage by 0.02%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #16132      +/-   ##
==========================================
- Coverage   90.86%   90.84%   -0.03%     
==========================================
  Files         159      159              
  Lines       50771    50777       +6     
==========================================
- Hits        46133    46126       -7     
- Misses       4638     4651      +13

Flag	Coverage Δ
#multiple	88.62% <100%> (-0.03%)	⬇️
#single	40.33% <28.57%> (-0.01%)	⬇️

Impacted Files	Coverage Δ
pandas/errors/__init__.py	100% <100%> (ø)	⬆️
pandas/core/generic.py	91.31% <100%> (+0.02%)	⬆️
pandas/plotting/_converter.py	63.54% <0%> (-1.82%)	⬇️
pandas/core/common.py	90.68% <0%> (-0.35%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8d122e6...37c7ad0. Read the comment docs.

[rgbkrk]

Looks good to me, this is completely appropriate behavior and informs the user as to what happened better than before. 👍

[jreback]

lgtm merge away

[jorisvandenbossche]

Small question: is it worth to add a special Warning for this (instead of using just UserWarning)?

In the sense that, if we intend to actually solve this issue in an upcoming release (certainly if we want this as the default repr this will be needed), then it is maybe not needed to add the special class (which we have to keep for compat afterwards).
If we foresee that we have to keep this warning for a longer time, then I am totally fine with it!

[TomAugspurger]

Small question: is it worth to add a special Warning for this (instead of using just UserWarning)?

Yeah, I went back and forth on this. In the end I went with a special warning since we may always have some things that can't be represented with the spec (maybe), and it's a bit easier to filter just this warning if you want. I'm open to changing it to a builtin warning though.