Closed szhorvat closed 11 months ago
Yep will do. Thanks for the heads-up @szhorvat
Would it be better to just use sapply
so as not to assume any specific type?
I prefer vapply()
to sapply()
because of exactly that -- catch some unexpected type, which is unlikely in this context. It is not a "problem" because integer is numeric, that is is.numeric(integer(1))
is TRUE
.
In the upcoming release, 1.6.0, the return type of vcount()
is still integer
. Originally, it was going to change only once we update to 0.10. Since this type change caused failures with several packages, should we just change it to numeric
right now, in 1.6.0, @krlmlr? See the discussion and commit above.
Network with 2 * 10^9 nodes or edges is a big one... :)
Perhaps @szhorvat it'd be better to continue in the igraph repository?
Hi @mbojan,
There's code in the vignette of this package that assumes that
igraph::vcount
returns aninteger
andigraph::ecount
returns anumeric
.https://github.com/mbojan/rgraph6/blob/master/vignettes/rgraph6.Rmd#L165
Generally, we do not guarantee the type, and in an upcoming version
vcount
will also return anumeric
. The reason is to be able to support graphs with more vertices than.Machine$integer.max
(similar to long vectors).Could you please fix the vignette so that it doesn't assume this?
There's no new igraph version to test yet, but work is ongoing here.