Hypothesis test failure - test_slgof.py::test_slgof_hypo

[mferrera]

A test case generate by hypothesis failed.

FAILED tests/test_slgof.py::test_slgof_hypo - AssertionError
Falsifying example: test_slgof_hypo(
    swl=0.0,
    sorg=0.15625,
    sgcr=0.15504997195913137,
    h=1e-07,
)

You can reproduce this example by temporarily adding @reproduce_failure('6.104.1',b'AXicY2BABgcUwBQTiKnvdC8/O3APkHmwUD/46pR/WwFrKwmp') as a decorator on your test case

[berland]

This is a real bug, the table that pyscal wants to output in this case starts with:

SLGOF
-- 
-- pyscal: 0.13.dev11+g4162b12
-- swirr=0, sgcr=0.15505, swl=0, sorg=0.15625, sgro=0, krgendanchor=sorg
-- Corey krg, ng=2, krgend=1, krgmax=1
-- Corey krog, nog=2, kroend=1
-- krg = krog @ sg=0.46456
-- SL     KRG       KROG      PC        
0.1562500 1.0000000 0.0000000 0.0000000
0.1562500 0.9999999 0.0000000 0.0000000
0.1563500 0.9997095 0.0000000 0.0000000
0.1564500 0.9994192 0.0000001 0.0000000
...

where the error is that the SL values are identical for the first two rows.

[berland]

The problem seems related to these floating point details:

>>> value_1 = 0.8437499719590555
>>> round(value_1 * 10000)  # this is what Pyscal uses to drop duplicates
8437
>>> print(f"{value_1:.5f}")  # This is how values are printed
0.84375
>>> value2 = 0.84375
>>> round(value2 * 10000)
8438  # different integer, so it will not be regarded as similar to value_1
>>> print(f"{value2:.5f}")
0.84375  # But formatted equivalent to value_1

This happens in GasOil.__init__()

[berland]

Still flaky after #458

FAILED tests/test_slgof.py::test_slgof_hypo - AssertionError
Falsifying example: test_slgof_hypo(
    swl=0.02432496616483698,
    sorg=1e-07,
    sgcr=0.02432496616483698,
    h=1e-07,
)

You can reproduce this example by temporarily adding @reproduce_failure('6.112.2', b'AXicY2BgOBDw9/fETYGFDAwMBwv1g69O+beVAasoAKBtE7s=') as a decorator on your test case

[alifbe]

I think the issue is with the PR #458. Since SWINTEGERS is multiplied by factor of 10. h_min is now even smaller and causing the issue (probably even more severe). I propose we do what you did and change this line h_min = 10.0 / float(SWINTEGERS)

pyscal/src/pyscal/gasoil.py

Line 92 in d6f0bf7

h_min = 1.0 / float(SWINTEGERS)

[berland]

Agree, h_min should have been adjusted in #458