From: Jarkko Hietaniemi Date: Sun, 3 Mar 2002 16:54:55 +0000 (+0000) Subject: EBCDIC: special cases of special cases. X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=commitdiff_plain;h=0c23d0e05cbaf49b7d1af4bdcdf8c1bcda2eaa0b;p=p5sagit%2Fp5-mst-13.2.git EBCDIC: special cases of special cases. p4raw-id: //depot/perl@14964 --- diff --git a/t/uni/case.pl b/t/uni/case.pl index f982b1d..17f7814 100644 --- a/t/uni/case.pl +++ b/t/uni/case.pl @@ -67,7 +67,48 @@ sub casetest { my $c = pack "U0U", hex $i; my $d = $func->($c); my $e = unidump($d); - print $d eq $spec->{$i} ? + if (ord "A" == 193) { # EBCDIC + # We need to a little bit of remapping. + # + # For example, in titlecase (ucfirst) mapping + # of U+0149 the Unicode mapping is U+02BC U+004E. + # The 4E is N, which in EBCDIC is 2B-- + # and the ucfirst() does that right. + # The problem is that our reference + # data is in Unicode code points. + # + # The Right Way here would be to use, say, + # Encode, to remap the less-than 0x100 code points, + # but let's try to be Encode-independent here. + # + # These are the titlecase exceptions: + # + # Unicode Unicode+EBCDIC + # + # 0149 -> 02BC 004E (02BC 002B) + # 01F0 -> 004A 030C (00A2 030C) + # 1E96 -> 0048 0331 (00E7 0331) + # 1E97 -> 0054 0308 (00E8 0308) + # 1E98 -> 0057 030A (00EF 030A) + # 1E99 -> 0059 030A (00DF 030A) + # 1E9A -> 0041 02BE (00A0 02BE) + # + # The uppercase exceptions are identical. + # + if ($i =~ /^(0149|01F0|1E96|1E97|1E98|1E99|1E9A)$/) { + $e =~ s/004E/002B/; # N + $e =~ s/004A/00A2/; # J + $e =~ s/0048/00E7/; # H + $e =~ s/0054/00E8/; # T + $e =~ s/0057/00EF/; # W + $e =~ s/0059/00DF/; # Y + $e =~ s/0041/00A0/; # A + } + # We have to map the output, not the input, because + # pack/unpack U has been EBCDICified, too, it would + # just undo our remapping. + } + print $w eq $e ? "ok $test # $i -> $w\n" : "not ok $test # $i -> $e ($w)\n"; $test++; }