[Rd] An update on the vctrs package

Abs Spurdle @purdle@@ @ending from gm@il@com
Thu Nov 8 21:38:56 CET 2018


Kia Ora

> vec_size(data.frame(x = 1:10))
> #> [1] 10
> vec_size(array(dim = c(10, 4, 1)))
> #> [1] 10

I think that "length" (in the R way) or "size" should incorporate the
number of columns (or other dimension information).

How about:
> vec_npoints (x)
> vec_nrecords (x)
> vec_nobs (x)

Depending on whether you want to promote geometry, databases or statistics.
I like npoints because I like to be Geometry-Friendly :).

This next issue isn't probably the feedback you're looking for.
However, given that you brought it up...

> vec_size(mean)
> #> Error: `x` is a not a vector

This is very interesting.
I'm writing a function where I'm planning to do the complete opposite.

I could return the number of lines of a function or the number of it's
arguments.
However, it would be nice to do something like.

> object.size (mean)
f: Nn -> N1

(Note that I'm using "N" for numeric rather than the more obvious "R" for real).

It's difficult for an R function to determine the input types of
another R function and even more difficult to determine the output
types.

It may be possible to incorporate a comment as a variable (inside a
function) rather than a comment as a comment.

> mymean = function (x)
+ {   #info
+     .map = "f: Nn -> N1"
+
+     mean (x)
+ }


kind regards
Abs



More information about the R-devel mailing list