Re: [igraph] Question on max id node

[Richard Geddes]

G,

I think the problem is in the file format end-of-line character of the file, which originally was made on a windows machine.  I converted the dos format to linux with

tr -d '\r' < dos.pairs > linux.pairs

and results are consistent now.  I've attached the file in question (dos.pairs) and the linux version so you can verify.

I had this problem not too long ago with some scripting code I was writing and should have suspected the same was happening here.

... I ran the update routine:

$0> cd (path to)igraph-dev-0.6
$0> tla update
$0> ./bootstrap.sh
$0> ./configure
$0> make
$0> sudo make install

The 'make check' returned a value of 0 (good)

Yes, I follow about the internal renaming of the nodes...
Is the attribute method the best method to get at the internal igraph node id values... I didn't really get to that part yet.

R


Gabor Csardi wrote:

Ok, i've no idea what is going on, but some clarification won't 
hurt. 'ncol' uses symbolic vertex names, i.e. instead of numbers 
you can have something like 
foo	bar
bat	foobar
etc...
in your file. If you have numbers, those are treated like strings 
as well. So if you have an edge list file with numbers and read it 
with 'ncol', then nothing ensures that vertex named '1' will get 
internal igraph vertex id 1. So having different node id's 
in the two graph is allright. 

Actually, having a slightly different merge matrix can be OK
too, since having different internal vertex ids can result different 
merges if there are two (or more) merges that increase the modularity 
with the same amount. So the output for my file can be still ok, 
although i haven't checked it. You can call igraph_isomorphic to be 
sure that the two graphs (the 'ncol' one and the 'egelist' one) are 
the same indeed.

What bothers me is the number of vertices in the 'ncol' graph. 
That cannot be bigger than the one for the 'edgelist' graph by
definition.

Still, i cannot reproduce it. Please tell me exactly, which version do you
use? The latest patch? igraph--main--0.6--patch-29 ?
Could you reinstall igraph and try it again? 
Does 'make check' run fine?

Could you send the edge list file as an attachment? Just in case 
there are some strange characters in it that look space/tab when 
including it in the email text.

Thanks,
G.

On Mon, Mar 10, 2008 at 05:15:50PM -0400, Richard Geddes wrote:
[...]

  

0       1
0       2
0       10
1       2
1       10
2       3
3       4
3       5
3       6
4       7
4       8
4       9
5       6
5       7
6       7
6       8
6       9
7       8
8       9

0       1
0       2
0       10
1       2
1       10
2       3
3       4
3       5
3       6
4       7
4       8
4       9
5       6
5       7
6       7
6       8
6       9
7       8
8       9