imgmath: Fix relative file path

[jschueller]

Closes #10944

[return42]

@jschueller thanks a lot for the patch 👍

The patch fixed the issue for a .. math:: directive

.. math::
   :label: schroedinger general

    \mathrm{i}\hbar\dfrac{\partial}{\partial t} |\,\psi (t) \rangle =
          \hat{H} |\,\psi (t) \rangle.

But in my tests :math: inline examples are still broken::

``\tfrac`` **inline example** :math:`\tfrac{\tfrac{1}{x}+\tfrac{1}{y}}{y-z}`
``\dfrac`` **inline example** :math:`\dfrac{\dfrac{1}{x}+\dfrac{1}{y}}{y-z}`

[jschueller]

@return42 and now ?

[return42]

Perfect .. works like a charm:

---

IDK why the CI reports an flake8 issue not related to this patch.

[jschueller]

lets hope it can be merged soon

[return42]

lets hope it can be merged soon

Yes, I hope the maintainers pay attention to this bugfix.

BTW: the flake8 issues reported by CI in build of master branch are known and here is the patch

• fix flake8 warnings #10995

[jschueller]

maybe @marxin or @jfbu ?

[jfbu]

@jschueller I rebased your work on current master tip so that we have fresh linting.

@AA-Turner Should we consider fixing the #10944 as done here for 6.0.1 rather than 6.1.0?

[jfbu]

In reviewing this I realized #10888 modified the returned path from render_math() to be absolute rather than relative. But the docstring still says

sphinx/sphinx/ext/imgmath.py

Line 212 in a9b0f27

Return the filename relative to the built document and the "depth",

Could the #10888 have been achieved while keeping the behaviour of render_math() as formerly? And at same time avoid #10944? If not the docstring of render_math() should be updated.

[jfbu]

Actually #10888 also modified the returned type of render_math() but did not update accordingly the docstring, which still says that it also returns the temporary filepath as well as the absolute path of destination image file. I see no mention of this change in CHANGES or in our document. I think it is important that it should be documented somewhere (I don't know to what extent or if at all render_math() is public API. Edit: surely not public API, but my point is that docstrings should be kept up-to-date with code...).

[jschueller]

ok, I updated the docstring

[jfbu]

LGTM, thanks
ping @AA-Turner about perhaps adding some brief summary of overall architecture at start of module as the #10888 discussion was somewhat terse and it is hard to follow it a posteriori. In particular at 4cb857d the extension internal dynamics were somewhat refactored and it would be nice to have some brief overview in plain terms of what this all does.