refactor: use 'better' logic for updating vertex attributes

[maelle]

Fix #1327

[krlmlr]

Thanks, looks good.

@clpippel: what do you think?

[maelle]

@clpippel friendly reminder if you have time and want to have a look 😸

[maelle]

@clpippel friendly reminder 😸

[clpippel]

I am not sure this is OK. I tried:

library(igraph)
graph <- make_ring(1)
index <- V(graph)
index <- igraph:::as_igraph_vs(graph, index)
(length(index) == vcount(graph) && identical(index, seq_len(vcount(graph))))
# [1] FALSE

The problem is that the class of index is `numeric' and the class of seq_len is 'integer'.
This corrects the problem:

(length(index) == vcount(graph) && identical(index, as.numeric(seq_len(vcount(graph)))))

Assuming vcount(graph) is not double precision.

This was also a surprise to me:

index <- V(g)
identical(index, V(g))
[1] FALSE

[clpippel]

@maelle, @krlmlr,
Note that E(g) is not identical with itself. Function identical(E(g), E(g)) returns "FALSE". Object E(g) contains a "weak reference", which changes after every call. So it is not a good idea to use identical() together with E(g) or V(g).

It seems consistent to apply the same changes to vertex.attributes() (line 512).

[maelle]

Thank you @clpippel!!

Yes I have opened #1332 to not forget about the other "confusing" loop. Thanks for the reminder!

[krlmlr]

Thanks both for looking into it!