Strategy for Fortran STOP

[mwtoews]

As mention in #3 (comment) there are potential issues with the Fortran STOP statements, which will exit the Python process.

Find them in src with:

git grep -B1 -i stop

There are approx. 50 of these in src, mostly with gslib's sgsim_code.f90.

There are a few strategies:

• Re-write the Fortran subroutines to functions returning a status, and somehow capture the error message. John could consider this option if he is keen.
• Guard the inputs within Python and raise ValueError, e.g. ensure the power parameters are valid before passing to Fortran.
• Some hackery like https://stackoverflow.com/questions/19596375/intercepting-fortran-stop-from-c

[jtwhite79]

I like first option the most bc it jives with the rest of the code. Can we use the existing functionality for setting the error message? @mwtoews , do you have a feel for how much effort it would be to patch these stops to be more amenable to library calls? Im happy to help on this front if you can work a strategy/approach..or we can divide and conquer?

[mwtoews]

Wrt option 1 to refactor Fortran, I'll take a look at what John as done and see if I can catch "INVALID power variogram" as a start. But also, raise this with John when you get a chance.

[jtwhite79]

So do think most of the stops are sourced to the power variogram? If so, we can just trap that on the python side...

[mwtoews]

I've started some data validation strategies with d470ed5 which operate on either scalars or vectors. E.g., here for install_structured_grid:

pypestutils/pypestutils/core.py

Lines 187 to 194 in d470ed5

	validate_scalar("ncol", ncol, gt=0)
	validate_scalar("nrow", nrow, gt=0)
	validate_scalar("nlay", nlay, gt=0)
	col = ManyArrays(float_any={"delr": delr}, ar_len=ncol)
	row = ManyArrays(float_any={"delc": delc}, ar_len=nrow)
	col.validate("delr", gt=0.0)
	row.validate("delc", gt=0.0)
	validate_scalar("icorner", icorner, isin=[1, 2])

These will be expanded elsewhere later.