Re: [igraph] Suggestion?

[Herb Roseman]

Hi:

There is an ambiguity.  You get the same warning  if the two nodes are the same.  In general, it is legitimate for a node in the mathematical object defined as a network to have an edge that goes from itself to itself. (See Trudeau.) For example, a network modeling the translation of DNA to proteins would have this type of edge.  In my particular problem (which I admit has little interest to most of the world) Spinoza legitimately develops corollaries by a proposition referring to itself.  Most of the users of igraph seem to be in the social sciences and my guess is that communications between social units may include such a node.  

Since only the user can know the logic of its network, it is her responsibility to process self-referring nodes.  In my opinion (although I can see the other side) the function should be agnostic. :) and return an empty list whenever there isn't a set of edges that connect one node to another.  There's an argument for throwing an error here, but for occasional python users an empty list is more convenient.  In the self-referential case the function would return [node node].  That would eliminate the ambiguity and conform igraph to mathematical practice.

From your response, I can see that there are other design considerations which are connected to the C layer. 

I'm sorry that this explanation is so long, but philosophers like to write.

Very best.

Herb     

On Sat Jan 03 2015 at 2:25:16 PM Tamas Nepusz <address@hidden> wrote:

> For the get_shortest_paths function it would be convenient if the function
> returned an empty list if the path was not possible.  This would provide a
> simpler way of handling this situation than the runtime warning message.
> Suppressing this message seems complex to the new user of python.
The warning comes from the C layer:

https://github.com/igraph/igraph/blob/master/src/structural_properties.c#L740

The Python interface simply turns every warning coming from the C layer into
a proper Python warning and although I could make a special exception for this
warning in the Python layer, I'm a bit reluctant to do that -- others might
dislike some other warning coming from the C layer and then I would have to do
the same for those warnings as well :) Is there any particular reason why this
warning bothers you so much that you would like to silence it at all costs?

An alternative solution would be to add a Boolean flag to the Python module
that the user can use to disable the translation of warnings coming from the
C layer into Python warnings; you could then say something like
"igraph.warnings = False" from Python to turn these warnings off.

All the best,
T.

_______________________________________________
igraph-help mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/igraph-help