[with patch, positive review] make openopt an optional spkg

[William Stein]

It's here:

​http://sage.math.washington.edu/home/wstein/patches/openopt-0.24.spkg

[David Joyner]

This installs fine in 4.0.2.rc0 on ubuntu 9.04 amd64. Running tests now.

[David Joyner]

Passes sage -testall and also installs fine on a 10.4 macbook.

[Minh Van Nguyen]

The optional package openopt-0.24.spkg is now in the optional packages repository at

​http://www.sagemath.org/packages/optional/openopt-0.24.spkg

[Minh Van Nguyen]

After uncompressing

​http://sage.math.washington.edu/home/wstein/patches/openopt-0.24.spkg

and doing hg status, I see lots of changes haven't been checked in. So I'm reopening this ticket. All changes need to be checked in via Mercurial. But note that the openopt project uses SVN for source code management.

[William Stein]

Replying to mvngu:

After uncompressing

​http://sage.math.washington.edu/home/wstein/patches/openopt-0.24.spkg

and doing hg status, I see lots of changes haven't been checked in.

No, that's wrong. Everything was checked in. The problem is that there was no hgignore, so all possible files that could get added to the repo (i.e. the stuff in src) got listed. I've added an .hgignore and rebuilt the spkg then posted it again in the optional package repo.

[Harald Schilly]

installs fine on kbuntu 9.04/32bit /w sage 4.1. I'm able to run an arbitrary example from the openopt website as a test problem. It uses the "ralg" solver provided by openopt.

preparser(False)
from numpy import *
from openopt import NLP
n = 10
an = arange(n) # array [0, 1, 2, ..., n-1]
x0 = n+15*(1+cos(an))
f = lambda x: (x**2).sum() + sqrt(x**3-arange(n)**3).sum()
df = lambda x: 2*x + 0.5*3*x**2/sqrt(x**3-arange(n)**3)
c = lambda x: an**3 - x**3
dc = lambda x: diag(-3 * x**2)
lb = arange(n)
p = NLP(f, x0, df=df, lb=lb, c=c, dc=dc, iprint = 100, maxIter = 10000, maxFunEvals = 1e8)
r = p.solve('ralg')
# expected r.xf = [0, 1, 2, ..., n-1]

sage: r = p.solve('ralg')
-----------------------------------------------------
solver: ralg   problem: unnamed   goal: minimum
 iter    objFunVal    log10(maxResidual)
    0  9.129e+03            -100.00
  100  4.104e+03            -100.00
  169  2.878e+02            -100.00
istop:  3 (|| X[k] - X[k-1] || < xtol)
Solver:   Time Elapsed = 2.41   CPU Time Elapsed = 1.88
objFunValue: 287.75368 (feasible, max constraint =  0)

sage: # expected r.xf = [0, 1, 2, ..., n-1]
sage: r.xf

array([ 0.5964556 ,  1.00355187,  2.00415294,  3.00156818,  4.0012493 ,
        5.00080644,  6.00036981,  7.00052146,  8.00016061,  9.00015341])

[David Joyner]

What other tests need to be run before this can be given a positive review?

[Harald Schilly]

Well, I've never reviewd a spkg before. So far as I can tell everything works as expected and therefore green light from me. Everything is only Python, and I think we can assume that it works on all platforms and there are no interferences with other parts of Sage, too.

[David Joyner]

Okay, me too. I'll change it to positive review then.

[Minh Van Nguyen]

Merged in the optional spkg repository on the Sage website.