[Lemon-user] remark on graphwriter

[Tapolcai János]

Hi Alpar,

> Could you please check it again if DigraphWriter works correctly?

Yes, it works correctly. In brief, I  have calculated some spanning trees and increment the cost along the edges of these trees. Depending on which edge belongs to which spanning tree I got different edge costs.

> > (1) the GraphReader writing a warning message, when it reads the label
> "@arcs",
> 
>       * First of all, I think it always a bad idea from a library to
>         write any kind warning. Remember, the "user" of the library is
>         the developer and not the end user of the software. Thus a
>         warning like this should be sent to the developer, not to the
>         user. An alternative could be to use the LEMON_DEBUG or
>         LEMON_ASSERT macro to check it. However,

Ok, I got your point. In this case my suggestion is to throw an exception if a GraphReader parses label "@arcs". The user can still hack the input file by replacing it to "@edges".

>       * @arcs and @edged are fully equivalent by design, because test
>         graphs and the output of the well known graph generators are
>         very often used to test both undirected and directed graph
>         algorithms.

I am not getting your point. Do you mean this issue is also confusing in the test graphs and in other well know graph generators? 

>       * Therefore it isn't clear who should emit a warning. I think the
>         mistake was done when you wrote the undirected graph with
>         DigraphWriter and not when you read it. 
>		What follows is that
>         DigraphWriter should warn you when you use it with an undirected
>         graph. BUT, undirected graphs are required to be fully
>         compatible with directed graphs, thus this exception would also
>         be quite odd.
> 
> So I currently can't see any better than warn people to take care of
> this issue.

Another option is to use different file extension for Graphs and Digraphs.

> > (2) a new Reading Rule for GraphReader named simpleGraph(), which
> >  filters the duplicated edges when parsing the lgf file.
> 
> Which 'cost' value should be kept when removing the parallel edges?

You can define the operator as a parameter of the function, and the default operation can be averaging. 

> 
> Once again, the mistake was made at writing, not at reading.

I think, it is easy to make such a mistake, and hard to identify. 

Best,
János