[Rd] why does [A-Z] include 'T' in an Estonian locale?

[peter dalgaard]

Just for amusement: Similar messups occur with Danish and its three extra letters:

> Sys.setlocale("LC_ALL", "da_DK")
[1] "da_DK/da_DK/da_DK/C/da_DK/en_US.UTF-8"
> sort(c(LETTERS,"Æ","Ø","Å"))
 [1] "A" "B" "C" "D" "E" "F" "G" "H" "I" "J" "K" "L" "M" "N" "O" "P" "Q" "R" "S"
[20] "T" "U" "V" "W" "X" "Y" "Z" "Æ" "Ø" "Å"

> grepl("[A-Å]", "Ø")
[1] FALSE
> grepl("[A-Å]", "Æ")
[1] FALSE
> grepl("[A-Æ]", "Å")
[1] TRUE
> grepl("[A-Æ]", "Ø")
[1] FALSE
> grepl("[A-Ø]", "Å")
[1] TRUE
> grepl("[A-Ø]", "Æ")
[1] TRUE

So for character ranges, the order is Å,Æ,Ø (which is how they'd collate in Swedish, except that Swedish uses diacriticals rather than Æ and Ø).

> Sys.setlocale("LC_ALL", "sv_SE")
[1] "sv_SE/sv_SE/sv_SE/C/sv_SE/en_US.UTF-8"
> sort(c(LETTERS,"Æ","Ø","Å"))
 [1] "A" "B" "C" "D" "E" "F" "G" "H" "I" "J" "K" "L" "M" "N" "O" "P" "Q" "R" "S"
[20] "T" "U" "V" "W" "X" "Y" "Z" "Å" "Æ" "Ø"
> sort(c(LETTERS,"Ä","Ö","Å"))
 [1] "A" "B" "C" "D" "E" "F" "G" "H" "I" "J" "K" "L" "M" "N" "O" "P" "Q" "R" "S"
[20] "T" "U" "V" "W" "X" "Y" "Z" "Å" "Ä" "Ö"



> On 30 May 2023, at 17:45 , Ben Bolker <bbolker using gmail.com> wrote:
> 
>  Inspired by this old Stack Overflow question
> 
> https://stackoverflow.com/questions/19765610/when-does-locale-affect-rs-regular-expressions
> 
> I was wondering why this is TRUE:
> 
> Sys.setlocale("LC_ALL", "et_EE")
> grepl("[A-Z]", "T")
> 
> TRE's documentation at <https://laurikari.net/tre/documentation/regex-syntax/> says that a range "is shorthand for the full range of characters between those two [endpoints] (inclusive) in the collating sequence".
> 
> Yet, T is *not* between A and Z in the Estonian collating sequence:
> 
> sort(LETTERS)
> [1] "A" "B" "C" "D" "E" "F" "G" "H" "I" "J" "K" "L" "M" "N" "O" "P" "Q" "R" "S"
> [20] "Z" "T" "U" "V" "W" "X" "Y"
> 
>  I realize that this may be a question about TRE rather than about R *per se* (FWIW the grepl() result is also TRUE with `perl = TRUE`, so the question also applies to PCRE), but I'm wondering if anyone has any insights ...  (and yes, I know that the correct answer is "use [:alpha:] and don't worry about it")
> 
> (In contrast, the ICU engine underlying stringi/stringr says "[t]he characters to include are determined by Unicode code point ordering" - see
> 
> https://stackoverflow.com/questions/76365426/does-stringrs-regex-engine-translate-a-z-into-abcdefghijklmnopqrstuvwyz/76366163#76366163
> 
> for links)
> 
> ______________________________________________
> R-devel using r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes using cbs.dk  Priv: PDalgd using gmail.com