html documentation: modules table and prev-next links

[GBillotey]

In my recent builds of the html documentation (sphynx 1.1.3, numpydoc 0.4) I have
noticed 2 issues (And thanks to @NelleV online builds, I know that I am at
least not the only one):

• 2 modules links appears in the top toolbar (on the left), for 2 domains
  py-modindex and np-modindex (which in fact are identical).
  If I am correct, this is related with sphynx new keyword
  http://sphinx-doc.org/config.html?highlight=html_use_modindex#confval-html_domain_indices
  Explicitely setting html_domain_indices = ["py-modindex"] in the build
  configuration file config.py (line 162) seems to solve this.
• The "previous topic", "next topic" sometimes (often) disappear. Compare for instance
  http://matplotlib.org/api/ticker_api.html#module-matplotlib.ticker with
  http://cbio.ensmp.fr/~nvaroquaux/matplotlib/doc/api/ticker_api.html#module-matplotlib.ticker
  It seems that some prev-next relationships are no more included in the sphynx toc tree ; however this goes far beyond my understanding of the doc build.

[pelson]

Added to the 1.3 blocker milestone. We should either fix or close before cutting v1.3.

[NelleV]

Thanks for reporting these problems.

I've investigated the next-previous link problem without success so far. I think we may have stumbled upon either a sphinx issue or a numpydoc issue.

[GBillotey]

@NelleV : Thanks for the feedback !

By any chance, did you try adding numpydoc_show_class_members = False in config.py ?
Seems it did help, at least in this case:
phn/pytpm#3 (comment)

(Should also solve #1896 - it works on my computer but not with current master as I am for the moment unable to build from the last source : a new windows issue.)

[NelleV]

@GBillotey works for me (I haven't checked it got rid of the warnings). Do you want to do the PR?

[pelson]

works for me (I haven't checked it got rid of the warnings). Do you want to do the PR?

Funny, I've just found this solution independently. Lets go for it - it removes thousands of warnings!

[mdboom]

Thanks! I had meant to look into this myself, but pleasantly pleased to find a solution has already been found!

[mdboom]

Just to confirm before I close this: @NelleV, @pelson: #1918 resolves this bug?

[NelleV]

Yes, but unfortunately we have a whole bunch of other problems on our hand (but that aren't roadblocks for the release IMO)