[David-the-happy-piglet]

XPath shorthand currently formats values with Python repr(). XPath 1.0 does not interpret those backslash escapes: text containing both quote types can raise XPathError, while backslashes and newlines can cause matching nodes to be missed. Regex shorthand such as ^A\d+$ also fails to match A123.

Encode values as XPath string literals, using concat(...) when both quote types are present. Apply the same encoding to @resource-id shorthand. Regex patterns retain their original characters and semantics.

Add 84 cases using real lxml/PageSource evaluation, covering exact, substring, prefix, suffix, resource-ID and regex selectors, matching and nonmatching nodes, and literal edge cases.

Validation on macOS 26.5.2 arm64, CPython 3.12.15 and lxml 6.1.3, based on master bad641e7:

• Before the fix, the new tests produce 39 failures and 45 passes.
• After the fix, all 99 XPath tests pass (84 new and 15 existing).
• Full unit suite: 190 passed, one existing failure in test_missing_command_shows_help. Unchanged upstream has the same failure with 106 passes; that CLI behavior is covered separately by fix: show successful help for an empty u2cli invocation #1198.
• Coverage run produces a 44% report and exits nonzero for that same baseline failure.
• Changed-file isort, diff checks, CLI help/version and imports pass.
• Source distribution and wheel build pass using the project's declared Poetry build backend through uv build.
• Independent local review found no actionable issues and reproduced the final tests' failures on upstream and success with the fix.

Android device tests, Windows/Linux and other Python versions were not run.

[Copilot]

🔵 Needs a closer look

Device-side compatibility remains unverified because validation covers only lxml/PageSource rather than the Android XPath service, particularly for concat() and literal handling.

No open issues 🎉

What changed in this PR

This pull request fixes XPath shorthand selectors by encoding values as valid XPath 1.0 string literals, preserving quotes, backslashes, newlines, Unicode, and regex escapes.

Changes:

• Added XPath-safe literal encoding with concat(...) support for values containing both quote types.
• Applied encoding to text, content-description, resource-ID, and regex shorthand selectors.
• Added 84 lxml-backed regression tests covering matching behavior and literal edge cases.

File	Description
uiautomator2/​xpath.py	Implements XPath 1.0-compatible string literal encoding for shorthand selectors.
tests/​test_xpath_literals.py	Adds comprehensive coverage for exact, substring, prefix, suffix, resource-ID, regex, and literal edge cases.

_{Review effort: Lite}