Strange lxml behavior

Question

I create xml manually and then try to validate it with xsd scheme. Validation doesn't pass at first, but if I convert xml to a string and back - then new xml passes validation.

from lxml import etree

xsd = etree.fromstring("""
<schema xmlns="http://www.w3.org/2001/XMLSchema" targetNamespace="some_namespace">
    <element name="el"></element>
</schema>""")
schema = etree.XMLSchema(xsd)

xml1 = etree.Element('el', nsmap={None: "some_namespace"})
xml2 = etree.fromstring(etree.tostring(xml1))

schema.assertValid(xml2)  # this passes
schema.assertValid(xml1)  # this fails

I see that xml1 and xml2 have different tags:

print xml1.tag  # --> el
print xml2.tag  # --> {some_namespace}el

But why xml1 and xml2 have such a difference? Looks like they should be the same.

Answer 1

Here you create an el element (no namespace):

xml1 = etree.Element('el', nsmap={None: "some_namespace"})

Using a nsmap parameter does not bind the element to a namespace; it just provides a mapping for serialization.

When etree.tostring(xml1) is executed, the serialization behaviour "kicks in". When the serialized result has been parsed, xml2 is an {some_namespace}el element instead of el.

To make it work, change the line to:

xml1 = etree.Element('{some_namespace}el', nsmap={None: "some_namespace"})