Support for non-default argument follows default argument about f90wrap HOT 9 CLOSED

Thanks for reporting, I'm sure we can come up with a fix. As you can see, Python does not allow optional arguments to come before non-optional ones. One possible solution would be to re-order the arguments, so the Python interface to the function above would become wrap(def_, opt=None). Would that be an acceptable solution?

(By the way, as a quick work around, if you don't need access to the routines that are causing this problem, you could omit them from the wrapper generation with the --skip argument to f90wrap.)

from f90wrap.

Thank you very much for your reply.

One possible solution would be to re-order the arguments, so the Python interface to the function above would become wrap(def_, opt=None). Would that be an acceptable solution?

That was exactly the solution I was thinking so it's definitely an acceptable solution.

(By the way, as a quick work around, if you don't need access to the routines that are causing this problem, you could omit them from the wrapper generation with the --skip argument to f90wrap.)

Unfortunately, it's in some pretty big part of the program so we can't just skip them.

from f90wrap.

@jameskermode Just wondering whether you have any update on this? I'm thinking of writing a script which would reorder the variables during build time myself, if you are currently working on this and are making progress then I guess I'd wait.

from f90wrap.

Hi Nelson, sorry, I didn't manage to work on this yet. I'd hope to be able to do it in the next week or so.

Rather than writing a script to modify your Fortran, if I point you to the relevant part of the f90wrap code would you be willing to try to add this feature directly? You would need to add a new subclass of FortranVisitor to transform.py and call this from transform_to_py_wrapper. The new class should be similar to RenameArgumentsPython but should instead re-order the arguments so any optionals come at the end.

from f90wrap.

Should now be fixed - as is often the case, it was not much work when I got round to it!

I added a version of your test case as a examples/optional_args_issue53, and this is now working correctly. (Note that I had to change from a program to a standalone subroutine to get it to link correctly but this is unrelated to the wrapping issue).

I will leave this issue open for the moment,so you can let me know of any further problems, but if all is well we can close it soon.

from f90wrap.

Sorry for the late reply. I can't actually get f90wrap working again. Following my details in this issue, I will get the error:

ImportError: /home/theory/phukgm/Documents/test/_test.so: undefined symbol: wrap_

For ONETEP, I'm getting the error

ImportError: dynamic module does not define init function (PyInit__onetep)

I feel like this is simply a setup issue with my system, but I can't quite figure out what.

from f90wrap.

I think the first error may be due to how you set up the Fortran code to be wrapped. Can you try to build the example in f90wrap/examples/optional_args_issue53 by running make from that directory?

from f90wrap.

I get ImportError: dynamic module does not define init function (PyInit_sizeof_fortran_t) during the make process.

from f90wrap.

Fixed (in person rather than on Github!). Problem as a mixture of Python 2 and 3, and of intel and GNU compilers.

from f90wrap.