Issue2000

dist/Lib/test/test_codecs.py as currently delivered is well out of date, but advancing to the Python 2.7 version raises test failures. Some relate to missing dependencies in the CJK codecs (see issue #1066). This note is to remind us to fix the ones that aren't: absence of UTF-32 codec implementation (needed for test_io) and seeming bugs in other UTF codecs.

In a forthcoming changeset I will update the test and add some skips/commenting out citing #1066 and all that.

Fixes for UTF-7 and UTF-16 are now in the trunk, and I'll turn to UTF-32.

Absence of an IDNA codec turns out to be issue #1153. I agree with later contributors to that report that it is not resolved until there's a codec. Pointers to java.net.IDN are helpful to that end.

I've provided UTF-32 codecs and implementations of codecs.encode and codecs.decode, with some modest accommodation in the tests to Jython's
allergy to invalid unicode escapes.

I have chosen not to implement _codecs.readbuffer_encode and _codecs.charbuffer_encode. I don't properly understand the issues, but I gather they were intended as an aid to codec writers in CPython and their retention was controversial. My judgement is they won't be missed.

There's a case for refactoring codecs.java and _codec.java, and for enforcing the bytes vs. unicode distinction in the signatures, but that's another issue.

The other shortcomings (IDNA and CJK) are covered by other issues so I'm closing this.