283 use BoolVal in stead of True and False

[Wout4]

Had to do a little bit more work than just replacing the return true and false so made it into a pr

[Wout4]

time for test suite:
3154 passed, 1 skipped, 4 warnings in 460.55s (0:07:40)

vs master
3154 passed, 1 skipped, 2 warnings in 512.05s (0:08:32)
On my machine which I expect to not be very consistent

[tias]

In retrospect I am not sure how reliable the testsuite is for time-benchmarking. I guess we would need a repo of exectuable python scripts that return a model, not even the pickles we have in bigtest, if we also want to benchmark model creation time... that is definitely for later ; )

Also you would have to compare the difference of just 1 commit to know the effect of that one change.

Anyway, on master all pass tests in ~3.5 minutes here, but this branch

FAILED tests/test_examples.py::test_examples[./examples/knapsack.py] - TypeError: int() argument must be a string, a bytes-like object or a number, not 'Operator'
FAILED tests/test_examples.py::test_examples[./examples/rcpsp.py] - AssertionError: IntVar upperbound must be integer <class 'cpmpy.expressions.core.Operator'> sum([0, 3, 2, 5, 4, 2, 3, 4, 2, 4, 6, 0])
FAILED tests/test_examples.py::test_examples[./examples/bus_schedule.py] - AssertionError: IntVar upperbound must be integer <class 'cpmpy.expressions.core.Operator'> sum([8, 10, 7, 12, 4, 4])
FAILED tests/test_examples.py::test_examples[./examples/jobshop.py] - AssertionError: IntVar upperbound must be integer <class 'cpmpy.expressions.core.Operator'> sum([3, 2, 2, 2, 1, 4, 0, 4, 3])
======================================= 4 failed, 3250 passed, 3 warnings in 143.88s (0:02:23) ========================================

e.g.

tests/test_examples.py:35: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
<frozen importlib._bootstrap_external>:790: in exec_module
    ???
<frozen importlib._bootstrap>:228: in _call_with_frames_removed
    ???
examples/rcpsp.py:22: in <module>
    start_time = intvar(0, max_duration, shape=nb_jobs)
cpmpy/expressions/variables.py:172: in intvar
    data = np.array([_IntVarImpl(lb,ub, name=_genname(name, idxs)) for idxs in np.ndindex(shape)]) # repeat new instances
cpmpy/expressions/variables.py:172: in <listcomp>
    data = np.array([_IntVarImpl(lb,ub, name=_genname(name, idxs)) for idxs in np.ndindex(shape)]) # repeat new instances
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

self = <[AttributeError("'_IntVarImpl' object has no attribute 'name'") raised in repr()] _IntVarImpl object at 0x7f1ac53c60d0>, lb = 0
ub = sum([0, 3, 2, 5, 4, 2, 3, 4, 2, 4, 6, 0]), name = None

I guess if it passes the tests we should not be too worried by another function call at creation time, at this point.

[Wout4]

ah the examples don't run on my machine and they seem to run but not fail on the automated tests

[tias]

see comments for change on built-in overwrites.

will merge

[tias]

I don't think we can/should in general change the return type of a Python builtin...

'any' can be used a lot in an arbitrary code-base that only includes cpmpy for a small part

maybe if we check that at least one element is a CPMpy expr, but for now I will just undo this

[tias]

this already shows that returning a custom type for 'any' is not a good idea...

Note that I now implemented __bool__() for BoolVal, so the original code would work now (with not), but still, better without for now

[tias]

I think this is what is breaking the tests on my machine
(does this mean that the examples are not run on the github CI? nor on your computer?)