FORTRAN namelist parsing: consistent org level use

[aidanheerdegen]

This package contains a FORTRAN namelist parser, but most other packages use Marshall Ward's f90nml package.

This may lead to inconsistent behaviour.

Is there any reason not to standardise on one tool?

[micaeljtoliveira]

The main reason for not using Marshall's parser here is that it does not preserve formatting and comments when doing round-trip parsing.

[micaeljtoliveira]

Forgot to say: there's still one feature missing in this collection of parsers that is the addition of keys. That means that it might not be possible yet for some of the other packages to use the Fortran parser defined here.

[aidanheerdegen]

Forgot to say: there's still one feature missing in this collection of parsers that is the addition of keys.

I don't understand what this means.

That means that it might not be possible yet for some of the other packages to use the Fortran parser defined here.

This seems unsatisfactory. How is a user to know if they can rely on this parser, or under what circumstances they should use one or the other? Is there an intention to make this parser feature complete?

Marshall has a very extensive list of test examples:

https://github.com/marshallward/f90nml/blob/main/tests/

Could we test against this to provide an indication of feature completeness?

[micaeljtoliveira]

Forgot to say: there's still one feature missing in this collection of parsers that is the addition of keys.

I don't understand what this means.

It means that after parsing a file, if you try to add a key to the dictionary holding the parsed data, it will return an error saying it's not implemented.

That means that it might not be possible yet for some of the other packages to use the Fortran parser defined here.

This seems unsatisfactory. How is a user to know if they can rely on this parser, or under what circumstances they should use one or the other? Is there an intention to make this parser feature complete?

Marshall has a very extensive list of test examples:

https://github.com/marshallward/f90nml/blob/main/tests/

Could we test against this to provide an indication of feature completeness?

Yes, we will get there in due time. The key addition for round-trip parsing is currently the top priority, as it's blocking other projects.