Add encoding conversion for CESU-8 from and to UTF-8
As discussed in issue #15931, encoding conversion (transcoding) from/to CESU-8 is missing, so we should add it. When then hopefully can make CESU-8 a dummy encoding.
Updated by duerst (Martin Dürst) 4 months ago
Issue #15931 mentions both https://www.unicode.org/reports/tr26/tr26-4.html and https://docs.oracle.com/javase/8/docs/technotes/guides/jni/spec/types.html#modified_utf_8_strings as definitions of CESU-8, but they are not identical.
The difference is in how they treat U+0000 (NULL) characters: UTR 26 does not treat it in any special way (i.e. it is encoded as "\x00"), but the Java definition treats specially, encoding it as "\xC0\x80". The IANA registration refers to the Unicode definition (see https://www.iana.org/assignments/charset-reg/CESU-8). TR 26 explains that "CESU-8 is useful in 8-bit processing environments where binary collation with UTF-16 is required.". For this to work, U+0000 has to be encoded as "\x00".
Issue #15931 currently implements CESU-8 as defined in UTR 26:
$ ruby -e 'puts "\xC0\x80".force_encoding("cesu-8").valid_encoding?' false $ ruby -e 'puts "\x00".force_encoding("cesu-8").valid_encoding?' true
It is unclear whether this is what the originator of issue #15931 wanted; his use case seems to be Java.