set_edge_label is setting labels on intended and unintended graph

31 views
Skip to first unread message

m_p_v_13

unread,
Mar 6, 2010, 2:03:17 AM3/6/10
to sage-support
set_edge_label is setting labels on intended and unintended graph

I generate some graphs and set one's edge labels. When I look at the
next genetated graph, it's edge labels have been changed.
I'm new to Sage/python so perhaps I goofed, but I doubt it. Thanks
for an explanation or workaround.

I am trying to help my daughter investigate an edge coloring graph
theory problem. Not the normal edge coloring constraint, though.

See the <<<<<<<<<<<<<<<< I added to the output section.

for G in graphs(3):
if G.is_connected():
A = G.adjacency_matrix()
print "Adjacency Matrix:"
print A
print "Edges:"
for e in G.edge_iterator( labels=True):
print e
#G.plot( edge_labels=True )
print "Setting edge labels..."
for i in G.edge_iterator( labels=True):
#print i
#print type( i )
G.set_edge_label( i[0], i[1], 'Red' )
print "Adjacency Matrix:"
print A
print "Edges:"
for j in G.edge_iterator( labels=True):
print j
#G.plot( edge_labels=True )
print "----------------------------"

Adjacency Matrix:
[0 1 1]
[1 0 0]
[1 0 0]
Edges:
(0, 1, None)
(0, 2, None)
Setting edge labels...
Adjacency Matrix:
[0 1 1]
[1 0 0]
[1 0 0]
Edges:
(0, 1, 'Red')
(0, 2, 'Red')
----------------------------
Adjacency Matrix:
[0 1 1]
[1 0 1]
[1 1 0]
Edges:
(0, 1,
'Red')
<<<<<<<<<<<<<<<<
(0, 2,
'Red')
<<<<<<<<<<<<<<<<
(1, 2, None)
Setting edge labels...
Adjacency Matrix:
[0 1 1]
[1 0 1]
[1 1 0]
Edges:
(0, 1, 'Red')
(0, 2, 'Red')
(1, 2, 'Red')
----------------------------

Message has been deleted

Pat LeSmithe

unread,
Mar 6, 2010, 3:38:33 AM3/6/10
to sage-s...@googlegroups.com
On 03/05/2010 11:03 PM, m_p_v_13 wrote:
> set_edge_label is setting labels on intended and unintended graph
>
> I generate some graphs and set one's edge labels. When I look at the
> next genetated graph, it's edge labels have been changed.
> I'm new to Sage/python so perhaps I goofed, but I doubt it. Thanks
> for an explanation or workaround.
>
> I am trying to help my daughter investigate an edge coloring graph
> theory problem. Not the normal edge coloring constraint, though.
>

> for G in graphs(3):

Replacing this with

for G in list(graphs(3)):

seems to work. But I think this generates all of the graphs in advance.

m_p_v_13

unread,
Mar 6, 2010, 2:03:06 PM3/6/10
to sage-support
Minh,

Thanks for the quick look at this.

So, what you are saying is that the iterator for graphs() does not
return independent graphs which can be changed without affecting the
others.
That does explain what I am seeing and is consistent with Pat
LeSmith's suggested workaround.

Should this property of the iterators to the generated graphs be
documented?

So, I think I will try making a copy of just the graphs I want to
change or use the list() trick.

Thanks again,
Michael Vogt


On Mar 6, 3:30 am, Minh Nguyen <nguyenmi...@gmail.com> wrote:
> Hi,


>
> On Sat, Mar 6, 2010 at 6:03 PM, m_p_v_13 <michael_p_v...@yahoo.com> wrote:
> > set_edge_label is setting labels on intended and unintended graph
>
> > I generate some graphs and set one's edge labels. When I look at the
> > next genetated graph, it's edge labels have been changed.
>

> Unless I'm mistaken, the graph generator function graphs(3) returns
> three graphs: G_1 with 1 edge, G_2 with 2 edges, and G_3 with three
> edges. Each graph builds upon another one, so G_2 is built using G_1,
> and G_3 is built using G_2. Whatever properties that are in G_1 would
> be preserved in G_2, and similarly for G_3. From your report, I gather
> that you don't want graph properties of G_1 to be carried over to G_2,
> and neither do you want properties of G_2 to be carried over to G_3.
> For completeness, I have copied a Sage session with essentially the
> same code as you posted, but used within Sage 4.3.3:
>
> [mvngu@sage mvngu]$ sage
> ----------------------------------------------------------------------
> | Sage Version 4.3.3, Release Date: 2010-02-21                       |
> | Type notebook() for the GUI, and license() for information.        |
> ----------------------------------------------------------------------
> sage: !more graph.py
> for G in graphs(3, implementation="networkx"):


>     if G.is_connected():
>         A = G.adjacency_matrix()
>         print "Adjacency Matrix:"
>         print A
>         print "Edges:"
>         for e in G.edge_iterator(labels=True):
>             print e

>         print "Setting edge labels..."
>         for e in G.edge_iterator(labels=True):
>             G.set_edge_label(e[0], e[1], "Red")
>         for e in G.edge_iterator(labels=True):
>             print e
>     else:
>         for e in G.edge_iterator(labels=True):
>             print e
>     print "-" * 20
> sage: load("graph.py")
> --------------------
> (0, 1, None)
> --------------------


> Adjacency Matrix:
> [0 1 1]
> [1 0 0]
> [1 0 0]
> Edges:
> (0, 1, None)
> (0, 2, None)
> Setting edge labels...

> (0, 1, 'Red')
> (0, 2, 'Red')
> --------------------

> Adjacency Matrix:
> [0 1 1]
> [1 0 1]
> [1 1 0]
> Edges:
> (0, 1, 'Red')
> (0, 2, 'Red')
> (1, 2, None)
> Setting edge labels...

> (0, 1, 'Red')
> (0, 2, 'Red')
> (1, 2, 'Red')
> --------------------
>
> --

> Regards
> Minh Van Nguyen

m_p_v_13

unread,
Mar 6, 2010, 2:08:46 PM3/6/10
to sage-support
Pat,

Thanks so much for pointing out this workaround.

I think I can specify that I only want connected graphs as parameter
to graphs() to reduce the overhead of list().
Or, I'll just make copies of the ones I need to change and add them to
a collection.

Message has been deleted

Michael Vogt

unread,
Mar 6, 2010, 5:27:38 PM3/6/10
to sage-s...@googlegroups.com
Hello Minh,

Thanks again for the fast response. 
I fixed my code to have the graphs I need make a copy and add to a seperate list

              ...
              for C in graphs(3):
                if C.is_connected():
                    G = copy(C)
                    L.append( G )
                    A = G.adjacency_matrix()
                    ...

This seems to work.

I'm just starting out with Sage and python.

I see you are the author of   sage.graphs.graph_coloring.edge_coloring()

I am about to study that to see how easily it could be modified to do a different edge coloring, detectable. The detectable coloring is such that no vertices have the same number of colors. That is, for three colors, that each vertex has a uniqque vector ( num red, num blue, num yellow ).

I hope you might be able to answer some questions when I run into them.

Thanks,
Michael Vogt












--- On Sat, 3/6/10, Minh Nguyen <nguye...@gmail.com> wrote:

From: Minh Nguyen <nguye...@gmail.com>
Subject: Re: [sage-support] Re: set_edge_label is setting labels on intended and unintended graph
To: sage-s...@googlegroups.com
Date: Saturday, March 6, 2010, 2:19 PM

Hi Michael,

On Sun, Mar 7, 2010 at 6:03 AM, m_p_v_13 <michael...@yahoo.com> wrote:

<SNIP>


> Should this property of the iterators to the generated graphs be
> documented?



--
Regards
Minh Van Nguyen

--
To post to this group, send email to sage-s...@googlegroups.com
To unsubscribe from this group, send email to sage-support+unsub...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-support
URL: http://www.sagemath.org

Jason Grout

unread,
Mar 6, 2010, 4:59:56 PM3/6/10
to sage-s...@googlegroups.com
On 03/06/2010 02:30 AM, Minh Nguyen wrote:
> Hi,
>
> On Sat, Mar 6, 2010 at 6:03 PM, m_p_v_13<michael...@yahoo.com> wrote:
>> set_edge_label is setting labels on intended and unintended graph
>>
>> I generate some graphs and set one's edge labels. When I look at the
>> next genetated graph, it's edge labels have been changed.
>
> Unless I'm mistaken, the graph generator function graphs(3) returns
> three graphs: G_1 with 1 edge, G_2 with 2 edges, and G_3 with three
> edges. Each graph builds upon another one, so G_2 is built using G_1,
> and G_3 is built using G_2. Whatever properties that are in G_1 would
> be preserved in G_2, and similarly for G_3.


Wow, that seems really, really bad. Shouldn't a generator like graphs()
return *copies* of the graphs it is using internally? The extremely
common use-case:

for g in graphs(3):
do something to g, add it to a list, etc.

could potentially really mess up things, then! For example, what if I
delete a vertex? I get a big list index error for the following code:

for g in graphs(3):
g.delete_vertex(0)
print g.am()

Robert, do you agree that this should be changed to return a copy of the
graph, instead of the actual graph used to build other graphs?


Thanks,

Jason

Robert Miller

unread,
Mar 7, 2010, 1:12:58 PM3/7/10
to sage-support
> Robert, do you agree that this should be changed to return a copy of the
> graph, instead of the actual graph used to build other graphs?

See my comments on the ticket.

Reply all
Reply to author
Forward
0 new messages