feat(rewrite): address assertion rewriting blind spots with systematic tests

[RonnyPfannschmidt]

Summary

Note: This is an AI-assisted experiment (Cursor + Claude Opus 4) that executed a plan after assessing initial details about assertion rewriting blind spots. More work is needed to make this clean and production-ready — this PR has known weaknesses and is meant as a starting point for discussion, not a final implementation.

This PR builds on top of the walrus operator double-evaluation fix (#14445) and addresses additional blind spots in the assertion rewriter:

• Add a systematic test infrastructure for assertion rewriting coverage (testing/test_assertrewrite_coverage.py)
• Implement visit_Subscript for container[key] introspection (shows container and key separately in failure messages)
• Implement visit_IfExp for ternary expression introspection (shows condition value while preserving short-circuit semantics)
• Flatten method call display to show where result = obj.method(args) instead of nesting the bound-method intermediate

Before/After

Subscript (new)

# Before: assert 1 == 99
# After:  assert 1 == 99
#          +  where 1 = {'a': 1, 'b': 2}['a']

IfExp (new)

# Before: assert 0 == 99
# After:  assert 0 == 99
#          +  where 0 = (... if True else ...)

Method calls (improved)

# Before: assert 42 == 100
#          +  where 42 = compute()
#          +    where compute = Obj().compute
# After:  assert 42 == 100
#          +  where 42 = Obj().compute()

Known weaknesses / areas for improvement

• The visit_IfExp only introspects the condition, not the branch values (to preserve short-circuit) — a more sophisticated implementation could emit an if/else block to get branch values too
• Container literal introspection (list, dict, set literals) is still unaddressed (xfail)
• Callable variables still show <function ...> repr instead of variable name (xfail)
• The method call flattening may interact unexpectedly with deeply chained attribute access patterns not yet covered
• No changelog fragment yet — this needs proper evaluation before merging
• Includes the Walrus expression duplicate evaluation failures with rewrite #14445 walrus fix commits (those should be merged first or this PR rebased after)

Test plan

• 75 tests pass in testing/test_assertrewrite_coverage.py (61 pass + 14 edge cases)
• 3 remaining xfail for lower-priority blind spots (container literals, callable variable repr)
• All 132 existing testing/test_assertrewrite.py tests pass (no regressions)
• All 162 testing/test_assertion.py tests pass (no regressions)
• Single-evaluation guarantees verified for all new visitors
• Short-circuit semantics preserved for IfExp
• Pre-commit hooks pass (ruff, mypy, codespell)

Made with Cursor

[Copilot]

Pull request overview

This PR extends pytest’s assertion rewriting to reduce “blind spots” and regressions (notably around walrus operator single-evaluation), while introducing a new systematic test suite to track introspection/semantics coverage across expression types.

Changes:

• Add targeted regression tests for #14445 (walrus operator double-evaluation) in testing/test_assertrewrite.py.
• Introduce a new systematic assertion-rewriting coverage test module (testing/test_assertrewrite_coverage.py) with helpers for introspection, semantic equivalence, and single-evaluation checks.
• Update the assertion rewriter to improve introspection for subscripts (container[key]), if-expressions (a if cond else b), and to flatten method-call explanations.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
testing/test_assertrewrite.py	Updates expectations for walrus/boolop output and adds #14445 regression tests.
testing/test_assertrewrite_coverage.py	New structured coverage suite for assertion rewriting behaviors (introspection/semantics/single-eval).
src/_pytest/assertion/rewrite.py	Implements new visitors (Subscript/IfExp), flattens method call formatting, and refines walrus-related evaluation behavior.
changelog/14445.bugfix.rst	Adds changelog entry documenting the walrus double-evaluation fix.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.