LaTeX: Fix #12735 about siglines with typing but no parameter list

[jfbu]

Fix #12735

Related #11444, #12561

[jfbu]

All the added CR's are cosmetic they will put one argument per line for the various \pysig... support macros, which helps when examining LaTeX file, or when setting up tests. Cause of the bug was that in the case of presence of typing of class name but no parameters the LaTeX macro was fed with only 3 arguments but it is designed to fetch 4. Note that for LaTeX commands declared via \newcommand with N parameters you can separate the {...} by spaces (but not by blank lines), inclusive of EOL space, it is ignored by TeX.

[picnixz]

I'll need a laptop for that PR just to confirm that the comments are also correct. The logic should be correct though. I think I first wanted a three arg macro parameter but then I observed that I could just have an empty parameter instead and then I forget to update one case. Thanks JF! I also forgot about the CR for cosmetic purposes. By the way do you think we should have some CRI which also indents the arguments like

\macro
  {}
  {}
  {}
  {}

Instead of

\macro
{}
{}
{}
{}

[jfbu]

@picnixz +1 for indenting by two spaces as you display, it is helpful for those who like me like to browse the latex files...

[picnixz]

Let's do it in a separate PR when I return.

[jfbu]

Let's do it in a separate PR when I return.

ok.

@AA-Turner Can we merge this now? (test of tests/test_builders/test_build_latex.py::test_one_parameter_per_line will still be failing on DocUtils HEAD, but passes on current DocUtils).

[AA-Turner]

Should this have a CHANGES entry?

[jfbu]

This is a bugfix but there was no bug report... I asked for merge because if we do a 8.0.x bugfix release it appeared to belong to it. I am +0 on adding a CHANGES but documentation is always good, so I can do it if you prefer.

[picnixz]

I think it's a fix of a fix but it's kind of a separate issue. So a CHANGELOG should be added (or you can just move my changelog entry (since we didn't really solve the issue) and credit yourself as well).

[picnixz]

Huh, I forgot about this one. Tomorrow afternoon, I'll have a last look and we can merge it I think. Sorry for that JF.

[picnixz]

Ok, three is one single case where I'm wondering whether we should actually handle it or not.

.. py:function:: generic_foo[R] -> R

This renders as

both in HTML and in LaTeX. I don't expect people to forget about the () if they have a return type. But should we actually do something for that instead? (I also think it's a separate issue though). Otherwise I'm good.

[jfbu]

Ok, three is one single case where I'm wondering whether we should actually handle it or not.

.. py:function:: generic_foo[R] -> R

Yes good find this looks a bit anomalous:

• latex code will have no associated \index command, and function does not show in the PDF index indeed
• but the latex code itself, using \pysigline with one argument, at least does not work only per accident, it does fetch one legit argument and renders it,
• everything is bold including brackets in output, contrasting with output for correct input.

It seems, afaict, that this not correct source anyhow. It does not cause a build crash, it produces in PDF a reasonable output, main deficiency is lacking an index entry (which is actually a help perhaps to locate such problems).

I would say it is a separate issue!

I need to go for some outdoor activity before upcoming bad weather, so I will not do anything yet until a few more hours to give you a chance to respond... +1 for merge-as-is on my side.

[picnixz]

I would say it is a separate issue!

I'll file one after this. Otherwise, I'm good to go. You can merge whenever you want!

[jfbu]

Here we go, thanks @picnixz!