[itecz26]

an element name that is_safe_name rejects, a non-ascii one like título or an escaped one like di[v, is translated to a name() comparison. name() is the qualified name as written in the document, so a bare x-ä matches <x-ä> inside xmlns="urn:d" while the node test emitted for an ascii name only matches no-namespace elements, which is what the docs promise for both, and :not(x-ä) wrongly excludes those elements. with a prefix, ns|x-ä compares the document prefix as a string instead of resolving ns through the xpath prefix mapping like the ns:f node test does, so it misses <p:x-ä> bound to the same uri and matches ns:x-ä bound to another one; [ns|data-ä] in xpath_attrib has the same prefix problem. test local-name() and the namespace instead, mirroring the |* translation, and keep name() only for a prefix that cannot be a node test itself. the xpath expectations in test_translation change accordingly, and test_select_with_default_namespace and test_select_with_namespace gain selects that fail on master.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (0a0c197) to head (f2ac28d).

Additional details and impacted files

@@            Coverage Diff            @@
##            master      #218   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         3           
  Lines         1019      1031   +12     
  Branches       170       174    +4     
=========================================
+ Hits          1019      1031   +12     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[AdrianAtZyte]

Looks good to me, thanks.

• I think unprefixed attribute names should get the same treatment: [p\:ö] still compares name(), so it matches a namespaced p:ö attribute, which is the same issue this fixes for elements. attribute::*[local-name() = … and namespace-uri() = ''] would make both consistent.
• Please, flatten the xpath_attrib() branches into a single if/elif chain (no namespace / unsafe namespace / safe name / unsafe name), the nested else makes it hard to follow.
• The new comment in xpath_element() could be a single sentence on what the code does.

[itecz26]

good catch on the unprefixed case. unprefixed names unusable in a name test now go through attribute::*[local-name() = ... and namespace-uri() = ''] as well, with a translation check for [p:ref] and a select check that [n:ö] doesn't match a namespaced n:ö attribute. xpath_attrib() is a flat if/elif chain now (wildcard / no namespace / unsafe prefix / safe name / unsafe name), and the xpath_element() comment is down to one sentence. also added a translation assert for the unsafe-prefix attribute branch since it lost its coverage once split out.