[R] Is simplify2array working for dimension > 2?

[Jean-Claude Arbaut]

Re your mention in your other mail (which didn't go to r-help) of this part
of the doc:

"The default value, TRUE, returns a vector or matrix if appropriate,
whereas if simplify = "array" the result may be an array of “rank”
(==length(dim(.))) **one higher than the result of FUN(X[[i]]).**"

It's not I understood this, but it makes some sense:

sapply(1:2, function(i) diag(3)) is a matrix, but
sapply(1:2, function(i) diag(3), simplify = "array") is a rank-3 array.

So this is the "appropriate case", and recursively simplifying nested list
to high rank arrays is not something that's supposed to be done.

And indeed, simplify2array(list(diag(2), diag(2))) does yield a rank-3
array as well. And it works for a list of rank-3 arrays, converted to
rank-4, etc.

So list of array is ok, list of list is not, except for rank 2. The
behavior for rank 2 led me to think it applied as well for higher rank, and
the doc for the 'higher' argument seemed to confirm this, but I was a bit
optimistic.

Thanks for your answer. I believe the doc may be improved a little bit, but
the intent looks clearer now.



Le jeu. 8 févr. 2024 à 10:32, Bert Gunter <bgunter.4567 using gmail.com> a écrit :

> Jean-Claude:
>
> Well, here's my "explanation". Caveat emptor!
>
> Note that:
> "simplify2array() is the utility called from sapply() when simplify is
> not false"
>
> and
>
> > sapply(a, I, simplify = "array")
>      [,1]   [,2]
> [1,] list,2 list,2
> [2,] list,2 list,2
>
> So it seems that simplify2array() is not intended to operate in the
> way that you expected, i.e. that recursive simplification is done.
> And, indeed, if you check the code for the function, you will see that
> that is the case. Perhaps the key phrase in the docs is in the
> sapply() part that says:
>
> "sapply is a user-friendly version and wrapper of lapply by default
> returning a vector, matrix or, if simplify = "array", an array ***if
> appropriate***, by applying simplify2array(). "   In other words,
> recursive simplification is considered not "appropriate".
>
> FWIW I also find this somewhat confusing and think that explicitly
> saying that recursive simplification is not done might make it less
> so. But writing docs that  address all our possible misconceptions is
> pretty difficult (or impossible!), and maybe adding that explicit
> caveat would confuse others even more... :-(
>
> Cheers,
> Bert
>
>
>
>
>
>
>
>
>
> On Thu, Feb 8, 2024 at 12:12 AM Jean-Claude Arbaut <arbautjc using gmail.com>
> wrote:
> >
> > Reading the doc for ?simplify2array, I got the impression that with the
> > 'higher = T' argument the function returns an array of dimension greater
> > than 2 when it makes sense (the doc says "when appropriate", which is
> > rather vague). I would expect
> >
> > a <- list(
> >   list(list(1, 2), list(3, 4)),
> >   list(list(5, 6), list(7, 8))
> > )
> > simplify2array(a, higher = T)
> >
> > to return the same (possibly up to a dimension permutation) as
> > array(1:8, dim = c(2, 2, 2))
> >
> > However, in this case simplify2array returns a matrix (i.e. 2 dimensional
> > array), whose elements are lists.
> > It's the same as
> > structure(list(list(1, 2), list(3, 4), list(5, 6), list(7, 8)), dim =
> c(2,
> > 2))
> >
> > I get the same behavior with
> > a <- list(
> >   list(c(1, 2), c(3, 4)),
> >   list(c(5, 6), c(7, 8))
> > )
> > but then the matrix elements are numeric vectors instead of lists.
> >
> > Did I miss something to get the result I expected with this function? Or
> is
> > it a bug? Or maybe the function is not supposed to return a higher
> > dimensional array, and I didn't understand the documentation correctly?
> >
> > There is a workaround, one can do for instance
> > array(unlist(a), dim = c(2, 2, 2))
> > and there may be better options (checking dimensions?).
> >
> > In case it's important: running R 4.3.2 on Debian 12.4.
> >
> > Best regards,
> >
> > Jean-Claude Arbaut
> >
> >         [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> > and provide commented, minimal, self-contained, reproducible code.
>

	[[alternative HTML version deleted]]