That's bloody brilliant.

One question, though. To my eye it looks to be vulnerable to the case where the original list has at least one element which starts with two or more chr(255)characters and at least one element being undef.

Or am I missing something?

I'm hoping deaccent will change chr(255) (y with two dots) to a plain y. It also fails if there strings starting with unicode characters above 255. Handling unicode in full generality is a huge pain, so I punted. You probably need something like Unicode::Collate to do it right.

Use Unicode. Perl is quite good at that

$ perl -MDP -we'my@x=("",undef,"1","123","2","\xff","\x{00ff}");DPeek 
+for map{$_->[1]}sort{$a->[0]cmp$b->[0]}map{[$_//"\x{1ffff}",$_]}@x'
PV(""\0)
PV("1"\0)
PV("123"\0)
PV("2"\0)
PV("\377"\0)
PV("\377"\0)
UNDEF
[download]

And chr(255) is *not* per definition an y with two dots. That is only the case in (encodings supported by perl (cp1252, cp1254, cp1258, hp-roman8, iso-8859-1, iso-8859-9, iso-8859-14, iso-8859-15, iso-8859-16, and UTF-7. If you don't specify the encoding or (lord forbids) *assume* any of the just listed, chr(255):

  7bit-jis                       \xFF
  cp1006                         ﹽ      ARABIC SHADDA MEDIAL FORM
  cp1026                                APPLICATION PROGRAM COMMAND
  cp1047                                APPLICATION PROGRAM COMMAND
  cp1250                         ˙      DOT ABOVE
  cp1251                         я      CYRILLIC SMALL LETTER YA
  cp1252                         ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  cp1254                         ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  cp1256                         ے      ARABIC LETTER YEH BARREE
  cp1257                         ˙      DOT ABOVE
  cp1258                         ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  cp37                                  APPLICATION PROGRAM COMMAND
  cp424                                 APPLICATION PROGRAM COMMAND
  cp437                                 NO-BREAK SPACE
  cp500                                 APPLICATION PROGRAM COMMAND
  cp737                                 NO-BREAK SPACE
  cp775                                 NO-BREAK SPACE
  cp850                                 NO-BREAK SPACE
  cp852                                 NO-BREAK SPACE
  cp855                                 NO-BREAK SPACE
  cp856                                 NO-BREAK SPACE
  cp857                                 NO-BREAK SPACE
  cp858                                 NO-BREAK SPACE
  cp860                                 NO-BREAK SPACE
  cp861                                 NO-BREAK SPACE
  cp862                                 NO-BREAK SPACE
  cp863                                 NO-BREAK SPACE
  cp865                                 NO-BREAK SPACE
  cp866                                 NO-BREAK SPACE
  cp869                                 NO-BREAK SPACE
  cp875                                 APPLICATION PROGRAM COMMAND
  cp932                          
  cp936                          
  cp949                          
  cp950                          
  gsm0338                        ?      QUESTION MARK
  hp-roman8                      ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  iso-2022-jp                    \xFF
  iso-2022-jp-1                  \xFF
  iso-2022-kr                    \xFF
  iso-8859-1                     ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  iso-8859-10                    ĸ      LATIN SMALL LETTER KRA
  iso-8859-13                    ’      RIGHT SINGLE QUOTATION MARK
  iso-8859-14                    ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  iso-8859-15                    ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  iso-8859-16                    ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  iso-8859-2                     ˙      DOT ABOVE
  iso-8859-3                     ˙      DOT ABOVE
  iso-8859-4                     ˙      DOT ABOVE
  iso-8859-5                     џ      CYRILLIC SMALL LETTER DZHE
  iso-8859-9                     ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  koi8-f                         Ъ      CYRILLIC CAPITAL LETTER HARD SIGN
  koi8-r                         Ъ      CYRILLIC CAPITAL LETTER HARD SIGN
  koi8-u                         Ъ      CYRILLIC CAPITAL LETTER HARD SIGN
  MacArabic                      ے      ARABIC LETTER YEH BARREE
  MacCentralEurRoman             ˇ      CARON
  MacChineseSimp                 …      HORIZONTAL ELLIPSIS
  MacChineseTrad                 …      HORIZONTAL ELLIPSIS
  MacCroatian                    ˇ      CARON
  MacCyrillic                    €      EURO SIGN
  MacFarsi                       ے      ARABIC LETTER YEH BARREE
  MacGreek                              SOFT HYPHEN
  MacHebrew                      |      VERTICAL LINE
  MacIcelandic                   ˇ      CARON
  MacJapanese                    …
  MacKorean                      …
  MacRoman                       ˇ      CARON
  MacRomanian                    ˇ      CARON
  MacRumanian                    ˇ      CARON
  MacSami                        ǩ      LATIN SMALL LETTER K WITH CARON
  MacTurkish                     ˇ      CARON
  posix-bc                       ~      TILDE
  UTF-7                          ÿ      LATIN SMALL LETTER Y WITH DIAERESIS
  viscii                         Ữ      LATIN CAPITAL LETTER U WITH HORN AND TILDE


Enjoy, Have FUN! H.Merijn

It would seem that deaccent()would modify the data to a sub-255 value, leaving a single 255 in the Schwartian Transform as a viable sort max key -- as noted above, this should be proven before deployed.

As to your other note, Unicode characters "above 255" are actually multi-byte sequences whose individual bytes still cannot exceed the architectural limitation of chr(255) so I question that perceived vulnerability.