R devel - Feb 2025 - Depends: R (>= 4.1) for packages that use |> and \(...)

>>>>> Ivan Krylov via R-devel writes:
Thanks.  I am already looking handling the 4.2.0 placeholder syntax, but
likely will need to refactor the code I added yesterday.

The "experimental" 4.3.0 extra placeholder feature looks like a lot of
effort: ideally there would be a simpler way.  I'll ask on R Core.

My guess would be that the new syntax is particularly prominently used
in examples: if so, it would be good to also have coverage for this.  

Best
-k
> Many thanks to Henrik for remembering the report in Bugzilla and to
> Kurt for implementing the change and finding out the true number of
> affected packages.
> On Wed, 22 Jan 2025 15:34:41 -0500
> Ian Farm <ian.farm at maine.edu> wrote:
>> Would packages using the underscore placeholder with the native pipe
>> need to also depend on R >= 4.2.0?
> That's a good find! For the R >= 4.2 syntax, we only need to check
for
> getParseData(...)$token %in% 'PLACEHOLDER'. The R >= 4.3 syntax
feature
> is harder to test for:
>>> As an experimental feature the placeholder _ can now also be used
in
>>> the rhs of a forward pipe |> expression as the first argument in
an
>>> extraction call, such as _$coef. More generally, it can be used as
>>> the head of a chain of extractions, such as _$coef[[2]]. 
> I think it might be possible to parse(text = paste('PLACEHOLDER
|>',
> grandparent_expression)) and then look at the top-level function in the
> call, but that feels quite fragile:
> x <- utils::getParseData(parse(f, keep.source = TRUE))
> i <- x$token %in% "PLACEHOLDER"
> pi <- x[i, "parent"]
> ppi <- x[x$id %in% pi, "parent"]
> placeholder_expressions <- utils::getParseText(x, ppi)
> extractor_used <- vapply(placeholder_expressions, function(src) {
>  toplevel <- parse(text = paste("PLACEHOLDER |> ",
src))[[1]][[1]]
>  identical(toplevel, quote(`$`)) ||
>   identical(toplevel, quote(`[`)) ||
>   identical(toplevel, quote(`[[`))
> }, FALSE)
> Alternatively, we may find the first child of the grandparent of the
> placeholder. If it's the placeholder expression, then the pipe must be
> of the form ...|> _..., which is the R >= 4.3 syntax:
> x <- utils::getParseData(parse(f, keep.source = TRUE))
> i <- x$token %in% "PLACEHOLDER"
> vapply(which(i), function(i) {
>  pi <- x[i, "parent"]
>  ppi <- x[x$id %in% pi, "parent"]
>  cppi <- x[x$parent %in% ppi, "id"]
>  min(cppi) == pi
> }, FALSE)
> -- 
> Best regards,
> Ivan
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

>>>>> Kurt Hornik writes:
Thanks again.  All added now with c87633.

Best
-k
>>>>> Ivan Krylov via R-devel writes:
> Thanks.  I am already looking handling the 4.2.0 placeholder syntax, but
> likely will need to refactor the code I added yesterday.
> The "experimental" 4.3.0 extra placeholder feature looks like a
lot of
> effort: ideally there would be a simpler way.  I'll ask on R Core.
> My guess would be that the new syntax is particularly prominently used
> in examples: if so, it would be good to also have coverage for this.  
> Best
> -k
>> Many thanks to Henrik for remembering the report in Bugzilla and to
>> Kurt for implementing the change and finding out the true number of
>> affected packages.
>> On Wed, 22 Jan 2025 15:34:41 -0500
>> Ian Farm <ian.farm at maine.edu> wrote:
>>> Would packages using the underscore placeholder with the native
pipe
>>> need to also depend on R >= 4.2.0?
>> That's a good find! For the R >= 4.2 syntax, we only need to
check for
>> getParseData(...)$token %in% 'PLACEHOLDER'. The R >= 4.3
syntax feature
>> is harder to test for:
>>>> As an experimental feature the placeholder _ can now also be
used in
>>>> the rhs of a forward pipe |> expression as the first
argument in an
>>>> extraction call, such as _$coef. More generally, it can be used
as
>>>> the head of a chain of extractions, such as _$coef[[2]]. 
>> I think it might be possible to parse(text = paste('PLACEHOLDER
|>',
>> grandparent_expression)) and then look at the top-level function in the
>> call, but that feels quite fragile:
>> x <- utils::getParseData(parse(f, keep.source = TRUE))
>> i <- x$token %in% "PLACEHOLDER"
>> pi <- x[i, "parent"]
>> ppi <- x[x$id %in% pi, "parent"]
>> placeholder_expressions <- utils::getParseText(x, ppi)
>> extractor_used <- vapply(placeholder_expressions, function(src) {
>> toplevel <- parse(text = paste("PLACEHOLDER |> ",
src))[[1]][[1]]
>> identical(toplevel, quote(`$`)) ||
>> identical(toplevel, quote(`[`)) ||
>> identical(toplevel, quote(`[[`))
>> }, FALSE)
>> Alternatively, we may find the first child of the grandparent of the
>> placeholder. If it's the placeholder expression, then the pipe must
be
>> of the form ...|> _..., which is the R >= 4.3 syntax:
>> x <- utils::getParseData(parse(f, keep.source = TRUE))
>> i <- x$token %in% "PLACEHOLDER"
>> vapply(which(i), function(i) {
>> pi <- x[i, "parent"]
>> ppi <- x[x$id %in% pi, "parent"]
>> cppi <- x[x$parent %in% ppi, "id"]
>> min(cppi) == pi
>> }, FALSE)
>> -- 
>> Best regards,
>> Ivan
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel

On Thu, 23 Jan 2025 11:16:48 +0100
Kurt Hornik <Kurt.Hornik at wu.ac.at> wrote:
> My guess would be that the new syntax is particularly prominently used
> in examples: if so, it would be good to also have coverage for this.
In today's CRAN snapshot, there turned out to be 198 packages that use
4.1 syntax in examples but not in code, 5 packages that use 4.2 syntax
in examples but 4.1 in the code, and 3 packages that use 4.2 syntax in
examples but not the code. This may be slightly imprecise because I
don't have some of the Rd macro packages installed and run
Rd2ex(stages=NULL) on manually-parsed Rd files without installing the
packages.

Attaching a patch that checks the syntax used in Rd examples at the
same time as the main R code, not necessarily the best way to perform
this check. Is it perhaps worth separating R/* checks from man/*.Rd
checks? Should R CMD check try to reuse the Rd database from the
installed copy of the package?

-- 
Best regards,
Ivan

-------------- next part --------------
A non-text attachment was scrubbed...
Name: check_R_syntax_in_examples.patch
Type: text/x-patch
Size: 4000 bytes
Desc: not available
URL:
<https://stat.ethz.ch/pipermail/r-devel/attachments/20250206/9fd89fa8/attachment.bin>