Improving Fortran integration (f2py)

[stefanv]

Critical issues

Input required

• numpy.f2py produces wrong results for arrays of strings #4519

High priority work

Quick fixes

• DOC: DOC: f2py docs outdated: extension module has no attribute 'as_column_major_storage' #14865
• DOC: DOC: f2py docs outdated: extension module has no attribute 'has_column_major_storage' #14862
• DOC: f2py user's guide uses python 2 syntax for print #14812

Other issues

All f2py issues

[melissawm]

Thanks, @stefanv . I'm preparing a PR with doc fixes for what I can see now (mainly checking that each example works as expected and bringing the python 3 syntax to the docs.)

[stefanv]

Thank you very much, @melissawm, and please feel free to edit the description of this issue to better describe the current state of Fortran issues.

[melissawm]

I can't seem to be able to edit this issue description, but I have another one:

#14960

Any input appreciated!

[rgommers]

I can't seem to be able to edit this issue description

looks like I told you the wrong thing, sorry. let's just add new issues in comments here

[melissawm]

New issues: #15325, #15311, #15308

[pearu]

Here is a list of topics where f2py requires improvements:

• Fortran 90 derived types: this is the only Fortran feature that support is missing in f2py. It has been (too) long in my todo list and the features have been asked by many since the beginning of f2py. In the past, one attempt was made by Jeffrey Hagelberg to implement the derived support but it was never completed. Unfortunately, the corresponding patch is lost. Originally, I planned tackled this issue in another rewrite of f2py, see https://github.com/pearu/f2py, but it is also possible to use the current numpy.f2py approach to implement the Fortran 90 derived types support.
• Character/string array support: currently, f2py supports character arrays as byte arrays that support was implemented when Numeric/numpy did not have unicode string support and f2py has not been updated since then with respect to this feature.
• Maintenance: over the years, some f2py features (for instance, the common blocks support, see BUG: f2py does not print names of COMMON data block items #15308) have regressed due to incorrect maintenance patches and the lack of tests that would have prevented such regressions. Such regressions need to be fixed and tests for all major f2py features need to be implemented.
• Documentation: it requires a revision as some parts of it are obsolete and not all f2py features are properly exposed.

This list of topics is likely incomplete but other issues ought to be minor compared to the topics listed here.

[melissawm]

At this point, all issues in the description are closed. We should probably close this issue, but could follow up with either a new issue for larger projects or the f2py projects on github. @pearu @HaoZeke thoughts?

[HaoZeke]

SGTM, going ahead with the closure then :)
The only thing left is the F90 derived types and I think we have a PR there and some issues so we're ok.