BUG: fix to_latex() when using MultiIndex with NaN in (#14249) (v2)

[tomneep]

This is an improved version of #19910 cleaning up MultiIndex handling in the process.

• closes BUG: NaN in multi-index and to_latex() #14249
• tests added / passed
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

I quickly looked at some other issues that this didn't originally aim to address and found that this PR closes #19981 and #18326 too, but I didn't add a tests or whats new entries for those (yet).

Edit: tests and whatsnew entries added for:

• closes to_latex() errors if index level name is not string #19981
• closes to_latex bug for midrule location #18326
• closes DataFrame.to_latex() outputs empty cells if MultiIndex names are empty strings #18669

[codecov]

Codecov Report

Merging #20797 into master will decrease coverage by 0.07%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #20797      +/-   ##
==========================================
- Coverage   91.85%   91.77%   -0.08%     
==========================================
  Files         153      153              
  Lines       49310    49257      -53     
==========================================
- Hits        45292    45207      -85     
- Misses       4018     4050      +32

Flag	Coverage Δ
#multiple	90.16% <100%> (-0.08%)	⬇️
#single	41.87% <0%> (-0.02%)	⬇️

Impacted Files	Coverage Δ
pandas/io/formats/latex.py	100% <100%> (ø)	⬆️
pandas/plotting/_converter.py	63.2% <0%> (-3.61%)	⬇️
pandas/core/nanops.py	95.13% <0%> (-1.17%)	⬇️
pandas/core/resample.py	96.06% <0%> (-0.37%)	⬇️
pandas/core/indexes/base.py	96.63% <0%> (-0.06%)	⬇️
pandas/core/common.py	92.01% <0%> (-0.03%)	⬇️
pandas/core/indexes/datetimelike.py	96.7% <0%> (-0.02%)	⬇️
pandas/core/arrays/base.py	84.14% <0%> (ø)	⬆️
pandas/core/frame.py	97.17% <0%> (ø)	⬆️
pandas/core/series.py	93.9% <0%> (ø)	⬆️
... and 5 more

---

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 31e77b0...2a05733. Read the comment docs.

[toobaz]

The patch looks good, and thanks for investigating the related issues. If you could also add tests for them too, it would be great.

[jreback]

lgtm. @toobaz merge when satisfied.

[toobaz]

@tomneep I just found out this PR fixes #18669 too: the original testcase cannot be reproduced (because of #18882 ), but the following can:

In [2]: mi = pd.MultiIndex.from_product([[1, 2]])

In [3]: df = pd.DataFrame(-1, index=mi, columns=range(4))

In [4]: df.index.names = ['']

In [5]: print(df.to_latex())
\begin{tabular}{lrrrr}
\toprule
  &  0 &  1 &  2 &  3 \\
\midrule
1 &    &    &    &    \\
2 & -1 & -1 & -1 & -1 \\
\bottomrule
\end{tabular}

Would you mind adding a test/whatsnew mention for this? Then I think we're ready to go.

[tomneep]

@toobaz Thanks for spotting that one too. I've added a test and whatsnew entry. The output is slightly different than your expected output in #18669 as there is an extra empty row for the index name (which is empty) but I've checked and this is consistent with what to_string() and to_html() do.

[toobaz]

@tomneep thanks!