Constraint programming extensions

[Wikunia]

Using ConstraintProgrammingExtensions.jl for

• [x] AllDifferent,
• [x] DifferentFrom
• [x] AllEqual
• [x] Strictly still uses my bridge as it doesn't seem to be implemented in CPE
• [ ] TableSet is a matrix for me but a set of vectors in CPE. I think a matrix is easier for the user.
• [ ] BoolSets are fundamentally different in my solver so I'll stick with my implementation for now. See also https://github.com/dourouc05/ConstraintProgrammingExtensions.jl/issues/20

Ping @dourouc05 as you might be interested :wink:

[codecov[bot]]

Codecov Report

Merging #270 (32e2dcb) into master (63af545) will decrease coverage by 0.05%. The diff coverage is 93.75%.

@@            Coverage Diff             @@
##           master     #270      +/-   ##
==========================================
- Coverage   97.60%   97.55%   -0.06%     
==========================================
  Files          53       53              
  Lines        4469     4458      -11     
==========================================
- Hits         4362     4349      -13     
- Misses        107      109       +2     

Impacted Files	Coverage Δ
src/MOI_wrapper/Bridges/util.jl	100.00% <ø> (ø)
src/constraints/all_equal.jl	95.87% <ø> (ø)
src/constraints/not_equal.jl	100.00% <ø> (ø)
src/MOI_wrapper/Bridges/strictly_greater_than.jl	75.00% <60.00%> (ø)
src/ConstraintSolver.jl	98.57% <100.00%> (-0.57%)	:arrow_down:
src/MOI_wrapper/constraints.jl	99.59% <100.00%> (-0.01%)	:arrow_down:
src/constraints/all_different.jl	100.00% <100.00%> (ø)
src/constraints/linear_constraints.jl	97.32% <100.00%> (ø)
src/lp_model.jl	90.62% <100.00%> (ø)
src/simplify.jl	98.78% <100.00%> (ø)
... and 3 more

---

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 63af545...32e2dcb. Read the comment docs.

[Wikunia]

Haha I remember I copied Strictly from you. I probably just misread that. Yeah looks the same to me 😆

[dourouc05]

I thought that this kind of commit would reduce the amount of code in the solver, but you're adding 50 lines with it :/. Do you think anything should be done better in CPE? (While remaining at MOI level, without dependency on JuMP, that is.)

[Wikunia]

The extra lines are probably from the manifest file. Otherwise I only removed lines and added your package

[odow]

Great to see this coming along.

[dourouc05]

For users, maybe it would be best if you provided @deprecate replacements for building sets? This would print a warning to change the calls, which could make updating their version of your solver easier (and you could remove these lines when tagging a v0.8, for instance). Normally, the syntax is: @deprecate old new false, without parentheses (I just hope that it works across packages :/).

[Wikunia]

I'm actually wondering whether I should export ConstraintSolverExtensions.AllDifferent and others, such that the user can use AllDifferent instead of needing to have using ConstraintSolverExtensions as well. What do you think in general about how solvers should handle this @dourouc05 ?

Oh I forgot the user will use my things anyway as yours need the dim and I have the layer around it.

[dourouc05]

Based on what MOI solvers have, they do not export anything: rather, every set is exported by JuMP if need be (and the symbol MOI is exported by JuMP too, so that you have direct access for sets that do not have a shortcut in JuMP). In my opinion, it's the job of JuCP (or whatever it ends up being called) to do that; in the meantime, it's probably OK that solvers do it too.

[dourouc05]

Indeed, I don't have yet bridges to reverse inequalities. Do you want to make a PR with https://github.com/Wikunia/ConstraintSolver.jl/blob/master/src/MOI_wrapper/Bridges/strictly_greater_than.jl or can I copy that code?

[Wikunia]

Feel free to copy the code :wink:

[dourouc05]

Yay :D! I just added a link to this solver from CPE's README.

[dourouc05]

If you want, you can replace your bridge by the new one: https://github.com/dourouc05/ConstraintProgrammingExtensions.jl/commit/660a0d6225d5a33f2a977e57ff97e8232cbd0637. I don't think your other bridges make sense in CPE for the moment, as long as we haven't settled on how to represent Boolean constraints.