Fixes for add_variables in CVXOPT, PPL, GLPK MIP backends and add_linear_constraints in CVXOPT

[mkoeppe]

The add_variables method had defects for the CVXOPT, PPL, GLPK.
Also, add_linear_constraints had a defect for CVXOPT.
See new tests added in this ticket.

I first noticed it with the CVXOPT backend. Then, by copying the same doctests to the other backends, I found more errors in the PPL, GLPK backends.
(See #20302 for a discussion regarding doctests for the backends -- there should be a better way than relying on copy-paste discipline.)

Note, I have copied the doctests to the CPLEX and Gurobi backends but am currently not able to run them.

CC: @dimpase @nathanncohen @ingolfured @videlec @jdemeyer

Component: numerical

Keywords: CVXOPT, PPL, GLPK

Author: Matthias Koeppe

Branch/Commit: e47b608

Reviewer: Vincent Delecroix, Dima Pasechnik

Issue created by migration from https://trac.sagemath.org/ticket/20303

[mkoeppe]

Branch: u/mkoeppe/fixes_for_the_cvxopt_mip_backend

[mkoeppe]

New commits:

39473c9

Fix and FIXME

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1 +1,4 @@
+Is anyone using the CVXOPT backend? (See #18572) (Not me.)
 
+While working on other backends, just by looking at the code, I noticed some issues, see branch. Clearly more doctests need to be added.
+(Also see #20302 for a discussion regarding doctests for the backends.) 

[mkoeppe]

Commit: 39473c9

[mkoeppe]

Author: Matthias Koeppe

[dimpase]

comment:4

this backend was partially just a warm-up for the CVXOPT SDP interface.

Besides, if you wanted to solve your LPs with an interior-point method, CVXOPT was the only way. Is it still the case?

[mkoeppe]

comment:5

Replying to @dimpase:

Besides, if you wanted to solve your LPs with an interior-point method, CVXOPT was the only way. Is it still the case?

I think the answer is yes; neither the GLPK backend nor the CPLEX backend seem to have support for these solvers' interior point algorithms.

[videlec]

comment:6

Hello,

You should explain in the ticket description what is not working... and also provide doctests to show that it works now!

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

bceca77	CVXOPTBackend: Use 'is None' instead of '== None'
728750f	CVXOPTBackend.add_variables: Pass arguments to add_variable, correct default for lower_bound
b4a8ed7	Copy new CVXOPTBackend.add_variables tests to other backends
b0d89e4	GLPKBackend.add_variables: Set column names correctly
183ce25	PPLBackend.add_variable, add_variables: Don't silently ignore binary=True, integer=True

[sagetrac-git]

Changed commit from 39473c9 to 183ce25

[mkoeppe]

Changed keywords from CVXOPT to CVXOPT, PPL, GLPK

[mkoeppe]

Description changed:

--- 
+++ 
@@ -1,4 +1,8 @@
-Is anyone using the CVXOPT backend? (See #18572) (Not me.)
+The `add_variables` method had defects for the CVXOPT, PPL, GLPK. 
+See new tests.
 
-While working on other backends, just by looking at the code, I noticed some issues, see branch. Clearly more doctests need to be added.
-(Also see #20302 for a discussion regarding doctests for the backends.) 
+I first noticed it with the CVXOPT backend. Then, by copying the same doctests to the other doctests, I found more errors in PPL, GLPK backends.
+(See #20302 for a discussion regarding doctests for the backends -- there should be a better way than relying on copy-paste discipline.) 
+
+Note, I have copied the doctests to the CPLEX or Gurobi backends but am currently not able to run them.
+

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

e47b608

CVXOPTBackend.add_linear_constraints: Add doctest, simplify code