Repository navigation
ElementTree.tostring(..., default_namespace=...) incorrectly strips namespace from attributes #113581
Description
Bug report
Bug description:
import xml.etree.ElementTree as ET
foo = ET.Element('{http://example.com}foo', {'{http://example.com}baz': '1234'})
# Expected output: <foo xmlns="http://example.com" xmlns:ns0="http://example.com" ns0:baz="1234" />
# Prints: <foo xmlns="http://example.com" baz="1234" />
print(ET.tostring(foo, default_namespace='http://example.com', encoding="unicode"))A careful reading of the "Namespaces in XML" specification suggests that the two xml documents are not the same.
https://www.w3.org/TR/xml-names/#defaulting
The namespace name for an unprefixed attribute name always has no value.
In the expected output the namespace name of the attribute is http://example.com, but in the actual output the namespace name of the attribute has no value.
Note that our parser does correctly implement this part of the "Namespaces in XML" specification
# Prints {'{http://example.com}baz': '1234'}
print(ET.XML('''<foo xmlns="http://example.com" xmlns:ns0="http://example.com" ns0:baz="1234" />''').attrib)
# Prints {'baz': '1234'}
print(ET.XML('''<foo xmlns="http://example.com" baz="1234" />''').attrib)Related:
Note 61290 is about tostring() raising an exception when it should not. My bug on the other hand is about tostring() producing incorrect output.
CPython versions tested on:
3.12
Operating systems tested on:
Linux
Activity
[-]ElementTree: default_namespace incorrectly strips napespace from attributes[/-][+]ElementTree.tostring(..., default_namespace=...) incorrectly strips napespace from attributes[/+][-]ElementTree.tostring(..., default_namespace=...) incorrectly strips napespace from attributes[/-][+]ElementTree.tostring(..., default_namespace=...) incorrectly strips namespace from attributes[/+]serhiy-storchaka commentedon Aug 31, 2026
PR #156747 fixes this, together with gh-61290.
A default namespace declaration does not apply to attribute names, so a qualified attribute name always needs a prefix, and its namespace ends up declared twice:
This is the output you expected, except for the generated prefix. lxml serializes it the same way.
gh-61290: Fix serializing attributes with the default_namespace option (
pythongh-61290: Fix serializing attributes with the default_namespace…