Assy solver fix

[adam-urbanczyk]

Will resolve #767, #790, #809

[codecov]

Codecov Report

Merging #806 (7625432) into master (4c5816d) will decrease coverage by 0.09%.
The diff coverage is 97.77%.

❗ Current head 7625432 differs from pull request most recent head 245e44b. Consider uploading reports for the commit 245e44b to get more accurate results

@@            Coverage Diff             @@
##           master     #806      +/-   ##
==========================================
- Coverage   95.18%   95.09%   -0.10%     
==========================================
  Files          32       32              
  Lines        7413     7436      +23     
  Branches      796      797       +1     
==========================================
+ Hits         7056     7071      +15     
- Misses        220      225       +5     
- Partials      137      140       +3     

Impacted Files	Coverage Δ
cadquery/occ_impl/solver.py	93.61% <97.14%> (-0.09%)	⬇️
cadquery/assembly.py	92.79% <100.00%> (+0.06%)	⬆️
tests/test_assembly.py	100.00% <100.00%> (ø)
cadquery/cqgi.py	79.09% <0.00%> (-2.46%)	⬇️
tests/test_workplanes.py	99.30% <0.00%> (-0.70%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4c5816d...245e44b. Read the comment docs.

[marcus7070]

The docs on the CCSAQ method are an interesting read: https://nlopt.readthedocs.io/en/latest/NLopt_Algorithms/#mma-method-of-moving-asymptotes-and-ccsa

popular for topology optimization...

Sounds good @adam-urbanczyk!

[marcus7070]

Tested on a random assembly with some kind of conflicting constraints. Old scipy.optimize solver:

Warning: Maximum number of iterations has been exceeded.
         Current function value: 0.000764
         Iterations: 1000
         Function evaluations: 1270
         Gradient evaluations: 1270

New nlopt solver:

{'cost': 6.476835679796388e-15, 'iters': 1554, 'status': -inf}

Assuming iterations and values are comparable between the two techniques, that's possibly an improvement for this one case. Certainly not a regression.

[marcus7070]

The appveyor check failed on test_constrain.

[adam-urbanczyk]

Yes, weird

[jmwright]

Looks good, but some comments in cadquery/occ_impl/solver.py would be very welcome.

[adam-urbanczyk]

I did run the assy tests 1000 times locally and cannot reproduce the behavior.

[marcus7070]

@slobberingant has reported this does resolve #767 and that test failure can not be reproduced, so I'm going to merge.