ffba55a·Author: Peter Hill·Closed by: Peter Hill·2020-05-20
Description
Fix some minor bugs in the CHEASE equilibrium ceq and read_chease modules:
Bugfix: chease diameter and rcenter used nt/2 instead of nt/2 + 1
Error should be small given typical grid sizes of ntheta > 100
Finish read_chease module when done with ceq
ceqin calls read_chease::read_infile which allocates arrays; they need cleaning up when we're done with ceq
Guard deallocates in read_chease::finish
Fix regression in generate_read_chease.rb: missing closing quote
Also: - fix missing public attribute on finish
Because of the fix bugfix above, I needed to loosen the tolerances in unit_tests/theta_grid because the results changed a little bit. As it is really a "smoke test", that is it only tells us if something has gone majorly wrong, this is probably ok. I’ve also reimplemented this test in pfunit_tests/geo/test_theta_grid, where Miller, CHEASE and EFIT have their own golden answers, and tighter tolerances because of this. There’s also no interpolation there, so it’s stricter in that sense too.
I’ve also added (pfunit) tests and docstrings for 99% of ceq. There’s a few bits that I can’t test right now because they call mp_abort
Fix some minor bugs in the CHEASE equilibrium
ceq
andread_chease
modules:Bugfix: chease diameter and rcenter used nt/2 instead of nt/2 + 1
Error should be small given typical grid sizes of ntheta > 100
Finish read_chease module when done with ceq
ceqin calls read_chease::read_infile which allocates arrays; they need cleaning up when we're done with ceq
Guard deallocates in read_chease::finish
Fix regression in generate_read_chease.rb: missing closing quote
Also: - fix missing
public
attribute onfinish
Because of the fix bugfix above, I needed to loosen the tolerances in
unit_tests/theta_grid
because the results changed a little bit. As it is really a "smoke test", that is it only tells us if something has gone majorly wrong, this is probably ok. I’ve also reimplemented this test inpfunit_tests/geo/test_theta_grid
, where Miller, CHEASE and EFIT have their own golden answers, and tighter tolerances because of this. There’s also no interpolation there, so it’s stricter in that sense too.I’ve also added (pfunit) tests and docstrings for 99% of
ceq
. There’s a few bits that I can’t test right now because they callmp_abort