SCIP.jl with exact-rational branch about scip.jl

[matbesancon]

Hi, are you sure SCIP.jl is correctly linked to your SCIP installation? You can test that with:

@eval SCIP.libscip

which points to the path where the SCIP shared library is

[leoneifler]

Hi, sorry that is a typo in the documentation. The parameter ist 'exact/enabled'. Will fix this.

[melkael]

Thank you for your help, it now works with "exact/enabled". However, it seems like Julia initializes the problem creation when SCIP.Optimizer() is called, which prevents me from setting the exact mode. I get the following error:

[set.c:767] ERROR: Exact solving mode can only be enabled/disabled before reading/creating a problem.

[leoneifler]

Hmm, that is a bit of an issue. I don´t know what can be done on the Julia side of things. The reason this error is necessary is that reading a problem in inexact mode already might not give you exactly the problem you want to solve (inaccuracies when converting to binary, setting small values to 0, etc.)

As a hacky fix, you could just set the parameters default value to true in set.c (SCIP_DEFAULT_EXACT_ENABLED)

[melkael]

Sounds like a good solution, I will try that, thank you!