[R] So-called non-API entry points

[jonkeane]

Describe the enhancement requested

There has been some churn on the so-called non-api entry points that CRAN is checking. Some of these come from cpp11 directly and those are ok (as of now, at least, it seems — but that could change arbitrarily!). But others just popped up recently and are from our code.

The ones from cpp11 that we can (mostly) ignore (for now): SETLENGTH, SET_GROWABLE_BIT, SET_TRUELENGTH

The new ones are: LEVELS and Rf_findVarInFrame3

arrow/r/src/arrow_cpp11.h

Lines 47 to 48 in e635cc2

	#define IS_ASCII(x) (LEVELS(x) & ASCII_MASK)
	#define IS_UTF8(x) (LEVELS(x) & UTF8_MASK)

arrow/r/src/arrow_cpp11.h

Line 397 in e635cc2

if (Rf_findVarInFrame3(arrow::r::ns::arrow, r6_class, FALSE) == R_UnboundValue) {

The second one has a similar call Rf_findVarInFrame that might be useable (and doesn't seem to complain) in that same file. Not sure about the LEVELS.

Component(s)

R

[nealrichardson]

LEVELS looks like something Romain pulled in when we switched to cpp11 (55defbf#diff-ef920fdae5124d20b341b9710adf8a33367cd2d24e08510913419ead9c8bb40fR86). I wonder if there's a proper cpp11 way to handle some of this string typing now.

[jonkeane]

Ok, I've got Rf_findVarInFrame3 fixed in https://github.com/apache/arrow/compare/main...jonkeane:non_api?expand=1

I haven't found a clean cpp11 replacement for LEVELS yet

[jonkeane]

Also ROFL, "added in R 4.5.0", from WRE

IS_ASCII
Use charIsASCII, added in R 4.5.0.

IS_UTF8
Use charIsUTF8, added in R 4.5.0, or avoid completely.

Though https://cran.rstudio.com/doc/manuals/r-devel/R-exts.html#Querying-CHARSXP-encoding-1 looks promising

[jonkeane]

Issue resolved by pull request 43173
#43173