add sort_by.data.table

[rikivillalba]

Proposal for #6662

• Note that sort_by.data.table() will sort using C-locale now, this is incompatible with base::sort_by.data.frame() used on data.tables

[github-actions]

Generated via commit fe8f6f1

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	4 minutes and 32 seconds
Installing different package versions	7 minutes and 52 seconds
Running and plotting the test cases	2 minutes and 26 seconds

[MichaelChirico]

Needs a few more tests -- currently all cases use the formula interface.

[rikivillalba]

Added some tests for other interfaces
https://github.com/Rdatatable/data.table/blob/bc3a413d1912340f93b1b50004a04a3b1ca2b485/inst/tests/tests.Rraw#L20711-L20725

Some questions arose to me

• sort_by.data.frame base method uses .formula2varlist(), which is a public but "internal" base package, that as the manpage says,

most of which are only user-visible because of the special nature of the base namespace.

we can use that in sort_by.data.table right?

• the list columns. I included a test in which the sort column is a list column. As forder treats a list like a dt itself, the result may have the same number of rows as elements each vector of the sorting list. Nevertheless this is the same behaviour as dt[order(x)] where x is a list column.

[MichaelChirico]

we can use that in sort_by.data.table right?

Hmm, thanks for flagging. Actually, I'm not concerned so much about that, but you did make me remember that it's a pretty new {base} function. It won't be available in all the R versions we support --> will cause an R CMD check NOTE there.

And actually, we've crossed this bridge before in #5393:

data.table/R/data.table.R

Lines 2460 to 2465 in d263924

	# same as split.data.frame - handling all exceptions, factor orders etc, in a single stream of processing was a nightmare in factor and drop consistency
	# evaluate formula mirroring split.data.frame #5392. Mimics base::.formula2varlist.
	if (inherits(f, "formula"))
	f <- eval(attr(terms(f), "variables"), x, environment(f))
	# be sure to use x[ind, , drop = FALSE], not x[ind], in case downstream methods don't follow the same subsetting semantics (#5365)
	return(lapply(split(x = seq_len(nrow(x)), f = f, drop = drop, ...), function(ind) x[ind, , drop = FALSE]))

So actually, we should just make a local copy of .formula2varlist() to re-use between split.data.table and sort_by.data.table.

[MichaelChirico]

the list columns. I included a test in which the sort column is a list column. As forder treats a list like a dt itself, the result may have the same number of rows as elements each vector of the sorting list. Nevertheless this is the same behaviour as dt[order(x)] where x is a list column.

As long as the behavior is consistent with [order(...)] I think we can ignore it in this PR.

[MichaelChirico]

Looks good! I think it needs an entry in man/, probably in ?forder would be the most logical place?

[rikivillalba]

Looks good! I think it needs an entry in man/, probably in ?forder would be the most logical place?

Added a reference in setorder.Rd (?forder). PTAL.