String#casecmp compares uppercase characters instead of lowercase
The current implementation of
String#casecmp converts characters to uppercase before comparing them. However, all references I've found for
strcasecmp (the C function on which
String#casecmp is based) indicate characters should be converted to lowercase before being compared.
For example, this man page says:
The POSIX.1-2008 standard says ... shall behave as if the strings had been converted to lowercase and then a byte comparison performed.
The difference in behavior is apparent when comparing / sorting strings containing
` (the characters that occur between
a). Converting to lowercase sorts these punctuation characters before
z along with most of the other punctuation in ASCII, but converting to uppercase sorts these characters after
Updated by jeremyevans0 (Jeremy Evans) 6 months ago
The documentation of
String#casecmp does not specify how it is is implemented, so it seems fair to consider switching. However, this change is likely to cause backwards compatibility issues. While it seems unlikely there are many applications relying on the current behavior, I would guess there are at least a few.
String#casecmp? uses lowercase and not uppercase, I think making such a change is reasonable, but we may want to delay making this change until Ruby 3.
Attached is a patch if we want to make this change.
Updated by mame (Yusuke Endoh) 6 months ago
Until ruby 1.8.7, it seemed to use downcase. It was changed at r14227 to support encoding. I think the behavior change was not intended, so this is merely a bug?
# ./bin/ruby-1.8.7-p374 -e 'p "a".casecmp("[")' 1 # ./bin/ruby-1.9.0-0 -e 'p "a".casecmp("[")' -1
Updated by nobu (Nobuyoshi Nakada) 6 months ago
rb_enc_upper is used at https://github.com/ruby/ruby/commit/269bd16b28e86d1333969389b7b402f2915e336f#diff-7a2f2c7dfe0bf61d38272aeaf68ac768R1431, while previous
rb_memcicmp maps to the lowercase.