FIX: avoid API double select

[12rambau]

Fix #388

Suprisingly 1 extra CSS rule was sufficient (my recent discovery on inline-blocks led me to this test and lucky for us it works like a charm.
I think it should be ported to sphinx attention but until then it does the job on our side.

exemplified in our doc here: https://pydata-sphinx-theme--1015.org.readthedocs.build/en/1015/examples/api.html#exceptions

[AA-Turner]

On https://pydata-sphinx-theme.readthedocs.io/en/stable/examples/api.html#numpy.linalg.tensorinv

If I double click ind, only ind is highlighted. I am not sure of the behaviour difference of this PR, though maybe I misunderstand the bug.

A

[12rambau]

@AA-Turner on my side both ind and int are selected:

I'm using safari but that's the same on firefox (MacOS Monterey)

[choldgraf]

I just tried this on Chrome and both replicated the old bug and confirmed it works in this version:

[AA-Turner]

@choldgraf are you on mac or windows? I tried with FF 105 originally (no issue), but having just now checked on chromium 106 the issue did appear--all testing on Windows 10.

This seems like a HTML rendering engine difference -- with FF treating the strong and span tags as different words but blink/webkit based browsers may treat them as one word. I don't know what the HTML 5 spec says is the right thing to do here.

A

[12rambau]

I'll merge this PR as it's solving our little problem and it is not creating issues for blink/webkit based browser.
FYI: the problem is kind of the inverse of the one described here: https://css-tricks.com/fighting-the-space-between-inline-block-elements/

@AA-Turner if you want we can continue this conversation in sphinx-doc/sphinx#10815 to come up with a more robust solution.

[rok]

Thanks so much @12rambau ! :)

[choldgraf]

I'm on windows 10 as well!

[flying-sheep]

It’s creating problems for us though. Just making a inline node display:flex is a very risky hack.

[12rambau]

could you elaborate on the risk you face due to our hack ?

[flying-sheep]

I meant “risky” in the sense of “likely to break”. Numpydoc’s definition lists happen to have markup that survives this, other code emitting definition lists probably doesn’t happen to create them just that way.

[12rambau]

Sure but I think a more robust solution should come from Sphinx itself