Fix esolver by changing sol path in esolver command

[zakandrewking]

This small fix was necessary for me to solve with esolver in cobrapy 0.6.2.

Using esolver version: QSopt_ex 2.5.10.3.

[hredestig]

Cheers for the fix, I guess you are aware but esolver solver interface is deprecated (doesn't say we can't merge this for now of course), do you need it? In that case perhaps you can comment on opencobra/optlang#113 ?

[hredestig]

The build plans use this version of esolver https://opencobra.github.io/pypi_cobrapy_travis/esolver.gz

perhaps try with that one?

[zakandrewking]

I'm on a Mac at the moment. Do you know what version that is? (thanks!)

[hredestig]

Using QSopt_ex (SVN-version 2.5.10:, built Jan 8 2015-13:20:35)
Using EGlib (SVN-version 2.6.21:, built Jan 5 2015-17:59:18)

[zakandrewking]

Same as me: 2.5.10

Do you know why some of the travis checks passed and some failed on this esolver change? thank!

[hredestig]

Yes, we don't install esolver on mac so those aren't affected

[Midnighter]

Hey @zakandrewking, I'm trying to move along the open PRs. I'm in two minds about this one. On one hand, I'd like to cut out the old solvers from cobrapy as soon as possible. That means that there should be a proper optlang interface for the esolver. I'm not sure who would work on that but we can figure it out. On the other hand, this is a minute change and if it's still wanted I don't mind merging. What's your take on this now?

[zakandrewking]

It's important to have at least one working exact solver, so we might as well keep esolver support here until optlang adopts it.

[Midnighter]

That's fair. This PR suggests that esolver is currently broken in cobrapy, though?

[zakandrewking]

Yes, to get it working with this version of esolver & COBRApy, I had to make this change.

[Midnighter]

Can you please update your PR by rebasing onto the current devel branch? Let's see what the tests say then.

[zakandrewking]

OK. Tests running.

[Midnighter]

I cancelled the Mac builds since Linux didn't pass anyway.

[zakandrewking]

ah, yeah, what i did must be somehow mac specific. i don't have time to figure out why it's doing that, so we can close the PR

[Midnighter]

So we will soon have an interface to the GLPK exact solver in optlang (opencobra/optlang#145). That's, of course, where it should live. I'm closing this then and will remove the esolver in the near future. If you need the esolver back instead of GLPK exact, please raise it on the optlang side.