#12833: Crashes and doctests problems with Gurobi
--------------------------------------+-------------------------------------
Reporter: ncohen | Owner: ncohen
Type: defect | Status: needs_work
Priority: major | Milestone: sage-5.0
Component: linear programming | Resolution:
Keywords: | Work issues:
Report Upstream: N/A | Reviewers: John Perry
Authors: Nathann Cohen | Merged in:
Dependencies: | Stopgaps:
--------------------------------------+-------------------------------------
Changes (by john_perry):
* status: needs_review => needs_work
* reviewer: => John Perry
Comment:
You're gonna hate me.
1. The docs for `MixedIntegerLinearProgram` state,
> * "solver" -- 3 solvers should be available through this class:
Only 3? `;-)`
2. The docs for `MixedIntegerLinearProgram` advise the user to get the
Gurobi solver using `solver="GUROBI"`. The ticket description, on the
other hand, states
> I also noticed that the doctests of gurobi_backend were actually *all
wrong* because of a solver="GUROBI" instead of solver="Gurobi".
So here's what I tried:
{{{
sage: p = MixedIntegerLinearProgram(solver='GUROBI')
---------------------------------------------------------------------------
ValueError Traceback (most recent call
last)
...
ValueError: 'solver' should be set to 'GLPK', 'Coin', 'CPLEX', 'Gurobi' or
None (in which case the default one is used).
}}}
The documentation should change.
2. The tests for `mip.pyx` fail with Gurobi installed, '''with or
without this patch.''' I've attached a file that summarizes the output,
but here are some representative samples:
{{{
File "/atlas/sage-5.0.beta9/devel/sage-main/sage/numerical/mip.pyx", line
564:
sage: p.show()
Expected:
Maximization:
x_0 +x_1
Constraints:
-3.0 x_0 +2.0 x_1 <= 2.0
Variables:
x_0 is a continuous variable (min=0.0, max=+oo)
x_1 is a continuous variable (min=0.0, max=+oo)
Got:
Maximization:
x_0 +x_1
Constraints:
R0: -3.0 x_0 +2.0 x_1 <= 2.0
Variables:
x_0 is a continuous variable (min=0.0, max=+oo)
x_1 is a continuous variable (min=0.0, max=+oo)
}}}
(Notice the `R0:`.)
{{{
File "/atlas/sage-5.0.beta9/devel/sage-main/sage/numerical/mip.pyx", line
837:
sage: p.solve()
Expected:
0.0
Got:
-0.0
...
File "/atlas/sage-5.0.beta9/devel/sage-main/sage/numerical/mip.pyx", line
500:
sage: sorted(map(reorder_constraint,p.constraints()))
Expected:
[(1.0, ([0, 1], [1.0, -1.0]), 4.0), (1.0, ([0, 2], [1.0, -2.0]),
None)]
Got:
[(1.0, ([0, 3], [1.0, -2.0]), None), (4.0, ([0, 1, 2], [1.0, -1.0,
1.0]), 4.0)]
}}}
If I remove the links to Gurobi files, the same tests pass. If I put the
links to Gurobi back in, the tests fail again.
I guess this is because Gurobi becomes the default solver once it is
installed: the `default_mip_solver()` method has this code:
{{{
for s in ["CPLEX", "GUROBI", "Coin", "GLPK"]:
try:
default_mip_solver(s)
return s
except ValueError:
pass
}}}
I don't have CPLEX installed, so MIP goes with GUROBI. (And it is
supposed to be GUROBI or Gurobi? LOL) This may reflect some bugs (I'll let
you decide that) but perhaps the best way to fix this is for MIP to ask
for GLPK as the default backend in its doctests. After all, this isn't an
MIP bug; at worst, it's a bug in our Gurobi backend.
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/12833#comment:3>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
