Convert input.f90 to python

[timothy-nunn]

Converts PROCESS input file parsing, validation, and actioning (mostly setting variables) to Python

Limitations

• Cannot parse string arrays (might be possible but I do not think its done in the code)

Regression failures

It seems there is a slight rounding difference that is causing the regression tests to fail. At the moment, some of the optimising runs fail because they slightly drift off their usual course. This can observed by amending the test_parse_real on main and adding:

[
        ("0.546816593988753", 0.546816593988753),
        ("0.6", 0.6),
]

The test fails with

FAILED tests/unit/test_input.py::test_parse_real[0.546816593988753-0.546816593988753] - assert 0.5468165939887529 == 0.546816593988753
FAILED tests/unit/test_input.py::test_parse_real[0.6-0.6] - assert 0.6000000000000001 == 0.6

It should be noted that this failure only happens when the difference is big enough, for example neither of the following fail despite asserting that the same parsed value is exactly equal to two slightly different values:

[
        ("0.008", 0.0080000000000000002),
        ("0.008", 0.008),
]

None of these tests fail on this branch

Notes

• Did not include factor, ftol as an input anymore, they are HYBRD variables and should be removed.
• Did not include nvar as it should be obsolete
• I did not copy across the warnings for dr_blkt_inboard and dr_blkt_outboard + i_blanket_type == 3 because this just printed a message. It should be checked that the docs make this very clear.
  

  PROCESS/source/fortran/input.f90

  Lines 1281 to 1307 in 6fc1dea

  	case ('dr_blkt_inboard')
  	if (i_blanket_type == 3) then
  	!CCFE HCPB model with Tritium Breeding Ratio calculation
  	write(outfile,*) '**********'
  	write(outfile,*) 'ERROR. BLNKITH input is not required for CCFE HCPB model with Tritium Breeding Ratio calculation -'
  	write(outfile,*) 'please remove it from the input file'
  	write(outfile,*) '**********'
  	else
  	call parse_real_variable('dr_blkt_inboard', dr_blkt_inboard, 0.0D0, 10.0D0, &
  	'Inboard blanket thickness (m)')
  	! Inboard blanket does not exist if the thickness is below a certain limit.
  	if(dr_blkt_inboard>=0.0D00.and.dr_blkt_inboard<=1.0D-3) then
  	dr_blkt_inboard = 0.0D00 ! Inboard blanket thickness is zero
  	i_blkt_inboard = 0 ! Inboard blanket does not exist
  	end if
  	end if
  	case ('dr_blkt_outboard')
  	if (i_blanket_type == 3) then
  	!CCFE HCPB model with Tritium Breeding Ratio calculation
  	write(outfile,*) '**********'
  	write(outfile,*) 'ERROR. dr_blkt_outboard input is not required for CCFE HCPB model with Tritium Breeding Ratio calculation -'
  	write(outfile,*) 'please remove it from the input file'
  	write(outfile,*) '**********'
  	else
  	call parse_real_variable('dr_blkt_outboard', dr_blkt_outboard, 0.0D0, 10.0D0, &
  	'Outboard blanket thickness (m)')
  	end if

• Did not include the following warning because it is likely no longer relevant
  

  PROCESS/source/fortran/input.f90

  Lines 1636 to 1643 in 6fc1dea

  	if (i_tf_sc_mat == 2) then
  	write(outfile,*) ' '
  	write(outfile,*) '**********'
  	write(outfile,*) 'Warning if you are using an old input file:'
  	write(outfile,*) 'i_tf_sc_mat=2 usage has changed -'
  	write(outfile,*) 'please check validity!'
  	write(outfile,*) '**********'
  	write(outfile,*) ' '

[codecov-commenter]

Codecov Report

❌ Patch coverage is 81.90476% with 38 lines in your changes missing coverage. Please review.
✅ Project coverage is 31.59%. Comparing base (cfab6dc) to head (25f4001).
⚠️ Report is 463 commits behind head on main.

Files with missing lines	Patch %	Lines
process/init.py	74.07%	21 Missing ⚠️
process/input.py	86.82%	17 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3552      +/-   ##
==========================================
+ Coverage   31.20%   31.59%   +0.39%     
==========================================
  Files          83       84       +1     
  Lines       19945    20112     +167     
==========================================
+ Hits         6223     6355     +132     
- Misses      13722    13757      +35     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[timothy-nunn]

Differences seen in the once-through input files at 0% tolerance come from a slight difference parsing in Python vs Fortran.

A difference is seen in the ST once through when parsing the following variables in Python instead of Fortran

beta, coreradius, coreradiationfraction, casths_fraction, d_vv_bot, dr_bore, dr_shld_outboard, dr_vv_outboard, etahtp, f_nd_alpha_electron, f_sync_reflect, fbeta_max, fipir, fne0, fradpwr, fradwall, fstrcase, fstrcond, ind_plasma_internal_norm, rhopedn, shldlth, thkcas, thshield_vb, vgap_xpoint_divertor

where the error accumulates more so due to the naive implementation (sum the digits multiplied by their place value).

I have taken some of these variables and written a test into test_input.py. These tests assert that the float is parsed and returned accurately when accessed. When running these tests with the input.f90 parser, they all failed:

tests/unit/test_input.py::test_parse_real_didnt_work_in_f90[0.546816593988753] FAILED                                                      [ 70%]
tests/unit/test_input.py::test_parse_real_didnt_work_in_f90[0.13134204235647895] FAILED                                                    [ 72%]
tests/unit/test_input.py::test_parse_real_didnt_work_in_f90[0.75] FAILED                                                                   [ 75%]
tests/unit/test_input.py::test_parse_real_didnt_work_in_f90[0.7] FAILED                                                                    [ 77%]
tests/unit/test_input.py::test_parse_real_didnt_work_in_f90[0.3] FAILED                                                                    [ 79%]
tests/unit/test_input.py::test_parse_real_didnt_work_in_f90[0.1293140904093427] FAILED 

But all passed when using my new Python parser. Note that Fortran even had problem parsing simple numbers with one decimal place (ie they had an error that was large enough that Python did not consider them equal)

[jonmaddock]

This is great Tim, SO much better. Just some clarifying comments required please. Did you check out cerberus for this?

[jonmaddock]

Can this comment be clarified?

[jonmaddock]

Great, I'm happy with this. Did you look into cerberus?

[timothy-nunn]

Great, I'm happy with this. Did you look into cerberus?

@jonmaddock I didn't end up using cerberus because the validation we implement here is very minimal and existing at the time of investigation. In future, it will be worth looking at this as we convert our data structure and IO formats.