>>> import xml.etree.ElementTree >>> data = '<?xml version="1.0" encoding="iso-8859-1"?>\n<money value="$\xa3\u20ac\U0001017b">$\xa3\u20ac\U0001017b</money>' >>> xml.etree.ElementTree.tostring(xml.etree.ElementTree.fromstring(data), 'unicode') '<money value="$£â\x82¬ð\x90\x85»">$£â\x82¬ð\x90\x85»</money>'

Patch for issue10590 fixes this for Python implementation of ElementTree, but not for C implementation.

Here is a patch for C implementation. Python implementation was fixed in issue17089.

Eli, this issue no longer has open pre-requisites. Issue10590 was replaced by issue17089 which closed now. Issue17089 fixed Python interface to expat parser, but cElementTree uses C interface of expat directly and the proposed pathes fix it.

Here is an updated patch.

LGTM

New changeset 7781ccae7b9a by Serhiy Storchaka in branch '3.3': Issue #16986: ElementTree now correctly parses a string input not only when http://hg.python.org/cpython/rev/7781ccae7b9a New changeset 659c1ce8ed2f by Serhiy Storchaka in branch 'default': Issue #16986: ElementTree now correctly parses a string input not only when http://hg.python.org/cpython/rev/659c1ce8ed2f

Oh, 2.7 still uses old doctests. It's a challenge to backport tests for this issue.

Due to the fact that ElementTree's documentation doesn't promise parsing Unicode string perhaps it shouldn't be backported to 2.7. At least I hadn't backported corresponded pyexpat changes (which affects pure Python ElementTree) to 2.7.

Agreed re 2.7; the problem is not important enough to warrant such a backport, due to the state of maintenance of 2.7 at this point.