Python maintenance #4516
Python maintenance #4516
Conversation
jngrad
force-pushed
the
python_maintenance
branch
from
806d3cd to
a84e824
Compare
Fix code samples, spell check and grammar check, cleanup docstrings.
`TabulatedNonBonded.is_active()` now returns False instead of None when the bond is inactive. Let ScriptInterface classes automatically generate interface methods. Always checks for runtime errors when the cell system changes or when a ScriptInterface object is created. Don't return True from methods that can return void (relic from the ES_OK / ES_ERROR signaling strategy that has been replaced by Python exceptions).
The algorithm searched for particles in a cylinder perpendicular to the plane instead of searching for particles in a slab parallel to the plane.
Re-use existing utility functions in tests_common to reduce code duplication. Replace unreachable exceptions by unittest assertions. Fix tolerance of LBGPU test case up to single-precision accuracy.
jngrad
force-pushed
the
python_maintenance
branch
from
a84e824 to
6d91ec5
Compare
| @@ -235,6 +234,13 @@ cdef Vector3d make_Vector3d(a): | |||
| return v | |||
|
|
|||
|
|
|||
| cdef Vector3i make_Vector3i(a): | |||
There was a problem hiding this comment.
should there be an assert that len(a) == 3?
There was a problem hiding this comment.
I don't think it's necessary, the helper functions make_Vector3i() and make_Vector3d() should only be called on a variable a after it passed the sanity check check_type_or_throw_except(a, 3, float, "...").
Handle 0-dimensional arrays in type checks. Improve error messages. Cleanup user guide. Convert uppercase comments to lowercase in the visualizer and remove comments that are redundant with docstrings. Co-authored-by: Patrick Kreissl <[email protected]>
| @@ -92,7 +92,7 @@ class LBGPULinearMomentum(LinearMomentumTest, ut.TestCase): | |||
| """Test for the GPU implementation of the LB.""" | |||
|
|
|||
| lb_class = espressomd.lb.LBFluidGPU | |||
| atol = 1e-6 | |||
| atol = 1e-5 | |||
There was a problem hiding this comment.
what happened here that the tolerance needs to be increased?
There was a problem hiding this comment.
This test is failing at random in CI, e.g. logfile 317102 in the long-range actors refactoring PR. The numerical error is random with a probability distribution that looks Chi-squared. The test fixes all random seeds and the core is supposed to sum the densities in a deterministic order, so I'm not sure what the root cause is. I ran that test in a loop for several thousand times and the largest deviation was 4e-6.
Description of changes:
"bind_centers"vs."bond_centers")tests_common.pymore oftensystem.analysis.nbhood()was removed (fixes Slab neighborhood search is broken #4517)TabulatedNonBonded.is_active()now returnsFalseinstead ofNonewhen the bond is inactiveScriptInterfaceobject is instantiatedTrueto signal they didn't raise an exceptionSystemclass and document why the_Globalsclass (now renamed to_BoxGeometry) has to existScriptInterfaceHelperclasses using automatically generated member methods