LQOI

[joaquimg]

Changes might be to aggressive.

Is enough for MathOptInterfaceCPLEX, with LinQuadOptInterface

How do we proceed?

[joaquimg]

I removed stuff from MOI which is lives now in MathOptInterfaceCPLEX.

This is needed to have MathOptInterfaceCPLEX working. It is passing CPLEX.jl tests.
Should I test anything else?

Is the current organization ok?

cc @mlubin @odow @blegat

[odow]

Callbacks are still unresolved. I'm not in a rush to have this merged. It almost seems like there should be a CPLEXBase.jl with this version of the C wrapper to go with MathOptInterfaceCPLEX.jl. Then CPLEX.jl can stay with the MPB version and callbacks.

[joaquimg]

I am not very found of having both CPLEXBase.jl and CPLEX.jl unless the second is completely discontinued and turned into legacy.

[odow]

This PR, and #139 are both pretty unwieldy.

I propose closing both, and starting again from the latest master.

That new PR would

• add the code from MathOptInterfaceCPLEX.jl
• add all of the new cpx_xxx.jl files
• retain all of the old code, except, where the arguments are the same, deprecate in favour of the new named functions. e.g., read_model becomes cpx_readcopyprob.
  An alternative to the last point:
• the various add_constrs! add_var! methods leave a high surface area for bugs. Leave all of the function definitions, but replace the bodies with an appropriate error.

[joaquimg]

I am ok with that

[odow]

Closing in favor of #181