Fixed LatexFormatter with MultIndex index

[agartland]

Used MultiIndex.formatter() to correctly sparsify the index for to_latex().

I was having an issue with printing Latex tables. This can be seen in the comparison of to_string() and to_latex() for the following table. Notice that the first row of Vol8 should not be missing any labels because it marks a new label in the outermost level. This is correct in to_string and incorrect in to_latex.

Output of to_string():

Volunteer Visit Stim Marker Vol6 2 Long CD137+ 96 IFNg+ 192 Short CD154+ 96 14 Long IFNg+ 288 Short CD137+ 192 Vol7 2 Long CD69- 96 IFNg+ 96 Short CD137+ 96 4 Short CD137+ 192 IFNg+ 96 14 Short CD137+ 96 CD154+ 96 IFNg+ 96 Vol8 2 Short CD137+ 288 4 Long IFNg+ 192 

Output of to_latex():

\begin{tabular}{llllr} \toprule & & & & Total cells \\ Volunteer & Visit & Stim & Marker & \\ \midrule Vol6 & 2 & Long & CD137+ & 96 \\ & & & IFNg+ & 192 \\ & & Short & CD154+ & 96 \\ & 14 & Long & IFNg+ & 288 \\ & & Short & CD137+ & 192 \\ Vol7 & 2 & Long & CD69- & 96 \\ & & & IFNg+ & 96 \\ & & Short & CD137+ & 96 \\ & 4 & & & 192 \\ & & & IFNg+ & 96 \\ & 14 & & CD137+ & 96 \\ & & & CD154+ & 96 \\ & & & IFNg+ & 96 \\ Vol8 & 2 & & CD137+ & 288 \\ & 4 & Long & IFNg+ & 192 \\ \bottomrule \end{tabular} 

I've tested the relevant pieces of this code using this example, but I would be grateful for help with adding tests to the repository. I'm not set up to develop the pandas codebase currently and am not exactly sure how to add the right tests in the right places. Thanks!

• Addresses issue Bug with sparsify in DataFrame.to_latex() #15480
• tests added / passed
• passes git diff upstream/master | flake8 --diff

[TomAugspurger]

Tests will go in pandas/tests/formats/test_formats.py, you'll see some other latex tests there.

Could you also post screen shots of the output before and after your fix? A release note too in doc/source/whatsnw/v0.20.0.txt

[jreback]

this is going to need tests

[jreback]

can you rebase / update

[jreback]

can you add the test from the issue? and a whatsnew note (bug fixes is fine).

tests are now in pandas/tests/formats/test_to_latex.py

[agartland]

I merged from upstream/master where upstream is pandas-dev/pandas. Is that what you meant? I apologize that I have done nothing with this recently. I'll try to get to it soon.

[agartland]

Yes, I should be able to generate the test pretty easily. I'll let you know if I need more help.

[jreback]

@agartland yep, that's what I meant.

[jreback]

can you rebase / update

[TomAugspurger]

@agartland we're releasing 0.20.2 in a few days, if you want to rebase and update.

[jreback]

needs a test & whatsnew note.

[agartland]

OK I've added a test and a whatsnew note. I will wait to see if the new test passes the CI.
Let me know if I need to do anything else. Sorry for the delay.

Before patch:

After patch:

[agartland]

Looks like m patch broke these two tests:

TestToLatex.test_to_latex_multicolumnrow 

They both fail because they expect a black value to look like:
5 & & 4
but instead it is:
5 & {} & 4

I'm not sure if there's a difference in Latex for these two. The {} comes from code that I didn't write, but now they are inserted in different places. Should I update the tests with {} so they pass?

[TomAugspurger]

@agartland the outputs do look the same to be. Could you do a bit of digging to see who's inserting the {} though? At a glance, it seems to be from LateFormatter.write_result. If that's right, then we should just update the tests.

I think https://travis-ci.org/pandas-dev/pandas/jobs/238131999#L1144 is a real failure though? Here it is, using pandoc, assuming I copy-pasted correctly:

--- header-includes: - \usepackage{booktabs} - \usepackage{multirow} --- # Old \begin{tabular}{ll} \toprule {} & x \\ {} & y \\ \midrule 0 & a \\ \bottomrule \end{tabular} # New \begin{tabular}{lll} \toprule & & 0 \\ \midrule x & y & a \\ \bottomrule \end{tabular} 

[jreback]

closing as stale. if you want to update. pls comment.