STINNER Victor <[email protected]> added the comment:
Attached patch is a partial fix: support UTF-16-LE, UTF-16-BE and
UTF-32-LE. Some remarks about my patch:
* UTF-32-BE is not supported because I'm too lazy tonigh
to finish the patch and because such file begins with 0x00 0x00
whereas the parser doesn't like nul bytes
* I disabled the cookie check if the file starts with a BOM (the
cookie is ignored) because the charset name is not normalized
and so if the cookie is not exactly the same as the hardcoded
charset name (eg. "UTF-16LE"), the test will fail.
Eg "utf-16le" != "UTF-16LE" :-(
* compile() would require much more effort to support UTF-16-*
and UTF-32-* because compile() simply rejects any string with
nul byte. It's beause it uses functions like strlen() :-/ That's
why I use subprocess([sys.executable, ...]) in the unit test and
not simply compile()
Support UTF-{16,32}-{LE,BE} would be nice but it requires to hack to
parser (especially compile() builtin function) to support nul bytes...
----------
keywords: +patch
Added file: http://bugs.python.org/file13409/tokenizer_bom.patch
_______________________________________
Python tracker <[email protected]>
<http://bugs.python.org/issue1503789>
_______________________________________
_______________________________________________
Python-bugs-list mailing list
Unsubscribe:
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
