Re: [igraph] Thoughts on consistency in functions

[Magnus Torfason]

Gábor Csárdi wrote:
> Magnus,
>
> I think that V(g) and E(g) do almost exactly what you want. Also,
> don't forget that we don't always have V(g)$name, so for graphs
> without the 'name' vertex attribute, we would need to return the
> numeric ids, or give an error message if 'names=TRUE'.

Yes, I would choose to return the numeric IDs, since that is how 
get.edgelist() works currently. It is of course also true that name does 
not always exist, but it is a very special attribute nevertheless, and 
created automatically for many types of graphs.

> But you are right, that the whole system could benefit from a
> redesign. More precisely, a design, as it was never designed properly,
> it only evolved to its current state. E.g. it would be great to have
> named graphs with checking the uniqueness of the names, maybe names
> should be just factors in R, etc. The current interface in clumsy, no
> doubt.

Well, I think it's a great system, and I'm *not* advocating a total 
redesign. The idea was to offer some ideas on low-hanging fruit (meaning 
seemingly trivial to implement, *and* backwards compatible) that would 
make parts of it more consistent.

Thanks for everything, and keep up the good work!

Best,
Magnus