Add options for pre-registered DDT names

[gold2718]

Add options for pre-registered DDT names

Add an option for pre-registered DDT names
Add ccpp_constituent_prop_ptr_t as a default (pre-registered) DDT
Add test for 'unknown' DDT variables

User interface changes?: Yes
Added a new option, --ddt-names, to allow users to simply add DDT names that allow generation of a CCPP metadata template from a Fortran source. Because it is an option, it is backwards compatible.

Fixes: #643
ccpp_fortran_to_metadata.py errors out when given a file that has a constituent properties object

Testing:
test removed: None
unit tests: NA
system tests: NA
manual testing: Added a manual test that tests this bug fix

[gold2718]

I wasn't sure how to best add a test for a one-off script such as ccpp_fortran_to_metadata.py so I just plunked a test script and example file in the test directory. Perhaps @mwaxmonsky can provide guidance that is more or less inline with his testing refactoring work .

[climbfuji]

I'll defer to @mwaxmonsky for the best way to test this, otherwise this looks good to me

[mwaxmonsky]

I wasn't sure how to best add a test for a one-off script such as ccpp_fortran_to_metadata.py so I just plunked a test script and example file in the test directory. Perhaps @mwaxmonsky can provide guidance that is more or less inline with his testing refactoring work .

Thanks for adding this in @gold2718!

In terms of setup, this feels in the spirit of the tests in the unit_test directory with a sample file. I'm still testing it out but I tried a pure CMake approach and a pure Python approach and it's kind of a tie in terms of complexity so I'm going ahead with a Python approach at the moment since it would also be useful for refactoring some of the unit_test file IO soon as well.

If I can get it working, would there be any concerns with using something like:

import unittest
import subprocess
import tempfile
import difflib

def test_pre_registered_ddt_type(unittest.TestCase):
    script = "..." #will follow current shell implementation.
    
    d = tempfile.TemporaryDirectory(dir=...) #Will be the same temp directory from script.
    subprocess.run(script)

    expected_data = fopen(...)
    actual_file   = fopen(...)
    deltas = list(difflib.ndiff(expected_data.readlines(), actual_file.readlines()))

    assertEqual([], deltas)

    d.cleanup()

Again, I'm still flushing out a working version and if there is a benefit to using the script approach, I'm happy to add that capability to the python test harness if I can so the script isn't needed.

[gold2718]

@mwaxmonsky, I have no problem with your approach. I slapped the bash script together because it is easy (no subprocess stuff to set up and analyze). To make this work, I recognize a consistent approach is best. Do you want me to use (and fix) your script as a substitute for test_fortran_to_metadata.sh?

[mwaxmonsky]

@mwaxmonsky, I have no problem with your approach. I slapped the bash script together because it is easy (no subprocess stuff to set up and analyze). To make this work, I recognize a consistent approach is best. Do you want me to use (and fix) your script as a substitute for test_fortran_to_metadata.sh?

I would say go with this as is and I can scoop it up into the refactor (or the next one at the current pace) since the test setup isn't a priority and I'm having to work through some other scenarios that will be useful in other places.

The only thing I would suggest is adding this to one of the workflow files as I don't think it is currently run and will be missed in the CI jobs.

[dustinswales]

@gold2718 Thanks for adding this feature. For the SCM I did something much less elegant for the mpi_Comm type

[gold2718]

The only thing I would suggest is adding this to one of the workflow files

Thanks, done!

[peverwhee]

thanks @gold2718 ! This is so helpful

[cacraigucar]

Tested on .F90 file with DDT (which failed previously) and the script now runs properly. Thanks!