switch default Sage graphs over to c_graph, and split up graph.py and graph_generators.py

[Robert Miller]

This is currently under discussion here:

​http://groups.google.com/group/sage-devel/browse_thread/thread/8edd29e9bddc67e5

I realized that it's probably actually time to switch over, since there are a few other developers working on Sage graphs besides just me now. That way if anything slows down, we are likely to find it out pretty quickly, and get it fixed. And, with the new defaults, things already feel more speedy:

BEFORE:

sage -t  "devel/sage-main/sage/graphs/graph.py"             
 [113.1 s]

AFTER:

sage -t  "devel/sage-main/sage/graphs/graph.py"             
 [78.5 s]

[Nathann Cohen]

Hmmm... When applying your patch here is what I get :

~/sage/sage-A/sage/graphs$ sage -t graph.py 
sage -t  "devel/sage-A/sage/graphs/graph.py"                
*** *** Error: TIMED OUT! PROCESS KILLED! *** ***
*** *** Error: TIMED OUT! *** ***
A mysterious error (perhaps a memory error?) occurred, which may have crashed doctest.
         [360.5 s]
exit code: 768
 
----------------------------------------------------------------------
The following tests failed:


        sage -t  "devel/sage-A/sage/graphs/graph.py"
Total time for all tests: 360.5 seconds

I know that this could not give you much information, but I do not know of any way to make -t more verbose... any idea ?

Nathann

[Nathann Cohen]

Hmmm, it seems to come from the docstring

G = graphs.CubeGraph(8)
H = G.distance_graph([1,3,5,7])

Which is *very* long !

[Alex Ghitza]

Replying to ncohen:

---

The following tests failed:

sage -t "devel/sage-A/sage/graphs/graph.py"

Total time for all tests: 360.5 seconds }}} I know that this could not give you much information, but I do not know of any way to make -t more verbose... any idea ?

I would try this :)

sage -t -verbose graph.py

[Nathann Cohen]

Yes, I found it... So stupid of not having checked :-)

By the way, the problem indeed comes from this distance_graph doctest !

Nathann

[Robert Miller]

This is because distance_graph ultimately relies on shortest_path, which uses the networkx graphs. In other words, for each pair of vertices distance_graphs calls shortest_path, which creates a copy of the graph, calls a NetworkX distance function on it, and discards it. I have created a separate ticket for this issue at #7640.

[Robert Miller]

I guess it should be somewhere on the ticket that #7651 should be taken care of first, also.

[Nathann Cohen]

With the patch from #7640, it gives :

BEFORE:

G = graphs.CubeGraph(8)
sage: time a=G.distance_graph([1,3,5,7])
CPU times: user 8.15 s, sys: 0.03 s, total: 8.17 s
Wall time: 8.18 s

AFTER:

G = graphs.CubeGraph(8)
sage: time a=G.distance_graph([1,3,5,7])
CPU times: user 6.51 s, sys: 0.03 s, total: 6.55 s
Wall time: 6.56 s

[Robert Miller]

Also need to be closed first: #7671 and #7672

[Robert Miller]

Add #7673 to the list.

[Robert Miller]

sorry, jason, i accidentally removed you from the cc block (not sure what happened there)

[Robert Miller]

The current patch does not split up graph.py, but the final version will...

[Robert Miller]

depends on 4.3.1.alpha1

[Robert Miller]

depends on trac_7634-switchover.patch

[Jason Grout]

Oh boy. I spent some time yesterday and today adding a few functions and cleaning up some documentation to graph.py. Well, you beat me to being ready, so I guess I'll rebase my patch on top of this one, hoping it gets into 4.3.1 so my patch on top of it can get in too.

[Nathann Cohen]

Well, this one does its jobs, and *finally* splits the graph.py file. c_graphs are now the default ones in Sage, and this patch renames what has to be.

Positive review/Good work/Excellent news !

Nathann

[Robert Miller]

based on trac_7634-refactor.patch