DOC: Fix doc warning "Code input with no code"

[datapythonista]

xref #24173

When building the documentation ./doc/make.py html, among other warnings we find the next:

/home/travis/miniconda3/envs/pandas-dev/lib/python3.6/site-packages/IPython/sphinxext/ipython_directive.py:1020: UserWarning: Code input with no code at /home/travis/build/pandas-dev/pandas/doc/source/user_guide/advanced.rst, line 724
  warnings.warn(message)
/home/travis/miniconda3/envs/pandas-dev/lib/python3.6/site-packages/IPython/sphinxext/ipython_directive.py:1020: UserWarning: Code input with no code at /home/travis/build/pandas-dev/pandas/doc/source/user_guide/cookbook.rst, line 1300
  warnings.warn(message)

All code blocks seem to have code, so it's not obvious why we get the issues. Also, the reported lines aren't probably correct. We should research why this is failing, and fix the problem.

[thiviyanT]

I'm on it 👍

[jorisvandenbossche]

Also, the reported lines aren't probably correct. We should research why this is failing, and fix the problem.

You need to substract 15, then the number matches the line numbers in the rst files. That's because of the header that gets injected (something I mentioned at that point as annoying for the header approach, but I don't really have a good alternative)

[TomAugspurger]

I noticed a couple had %timeit, not sure about all of them.

In a tiny example, %timeit worked OK (but maybe that was at the top of the input).

[jorisvandenbossche]

• In case of the advanced.rst, it is due to a %timeit I think. The output in the online docs stops after the first block with timeit calls.
• In case of the cookbook one, it is a code block with a function, where the function is cutted after a first blank line within the function definition (compare http://pandas-docs.github.io/pandas-docs-travis/user_guide/cookbook.html#correlation with the actual rst source)

[TomAugspurger]

I think a custom directive may be able to handle the header throwing off of line numbers, but that's quite a bit more complex.

…

On Tue, Jun 11, 2019 at 11:07 AM Joris Van den Bossche < ***@***.***> wrote: Also, the reported lines aren't probably correct. We should research why this is failing, and fix the problem. You need to substract 15, then the number matches the line numbers in the rst files. That's because of the header that gets injected (something I mentioned at that point as annoying for the header approach, but I don't really have a good alternative) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#26788?email_source=notifications&email_token=AAKAOIQ4PPWQJMJLP3N2ENTPZ7EVHA5CNFSM4HW7XAV2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXNU4LA#issuecomment-500911660>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAKAOISVH4M7NSH62D7T6FLPZ7EVHANCNFSM4HW7XAVQ> .

[jorisvandenbossche]

BTW, the issue for the problem with the IPython directive is here: ipython/ipython#11362