implements getsolvetime

[fserra]

fix #17

[mlubin]

👍
If you've tested it locally let's merge after travis passes.

[fserra]

If this is a valid test:

m = Model(solver=SCIPSolver())
@variable(m, x)
solve(m)
getsolvetime(m)

Then it is good to merge

[mlubin]

Yes, it's a valid test (wouldn't hurt to add something to the unit tests).

[fserra]

how do you call getsolvetime from mathprogbase??

[mlubin]

MathProgBase.getsolvetime

[fserra]

Ok, this is a mess.

@leethargo we are freeing the transformed problem after the SCIPsolve in CSIPsolve and so the time obtained by SCIPgetSolvingTime is 0.

We have to do something about this, like a function that needs to be called before modifying a problem.
Otherwise, we would need to store also the solving time data in CSIP 😧

How would such a function be integrated? Would we need to call it basically in every method with a ! in its name?

[rschwarz]

Was problem modification the only issue with the missing freeTransform after the solve?

[fserra]

I think yes

On Sep 14, 2016 15:09, "Robert Schwarz" notifications@github.com wrote:

Was problem modification the only issue with the missing freeTransform
after the solve?

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#18 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AFNG9rA2GcawinYURrF9XUgw6ANCcWnsks5qp_IbgaJpZM4J8oHC
.

[rschwarz]

I don't like it, but as with PySCIPOpt, it may be unavoidable if we don't want to restrict the feature set.

[rschwarz]

@fserra : can you add your tests?

[fserra]

done