[aarjang]

Description

Fixes #5071.

The workflow expression evaluator split a comparison at the first operator in _COMPARISON_OPERATORS table order and evaluated the whole remainder as the right operand, so a chained comparison such as {{ 0 <= inputs.n <= 10 }} became 0 <= (inputs.n <= 10) — a number compared against a boolean. Validation accepted the condition and nothing was raised at run time, so an if, while or do-while step silently took the wrong branch (with n = 50 the condition read true and the then branch ran).

This PR makes the evaluator scan the expression left to right for every top-level comparison operator (longest operator first at each position, with the same quote/bracket skipping as _find_top_level) and evaluate the chain pairwise with an implicit and, the way Python and Jinja2 chain comparisons: 0 <= n <= 10 is 0 <= n and n <= 10. This matches the module's documented contract ("a safe Jinja2 subset").

• Single comparisons keep their previous semantics; operators inside quoted or bracketed operands are still ignored.
• Mixed chains work the way Python defines them, membership included: a < b == c, a < b in items.
• A chain binds tighter than and/or, as before.
• Every operand is evaluated before any pair is compared, so validation probes (_collect_leaves) visit the same leaves regardless of which pair fails first.

Testing

• New regression test TestExpressions::test_chained_comparisons_evaluate_like_python fails on main exactly as the issue describes ({{ 0 <= inputs.n <= 10 }} is True for n = 50) and passes with this change. It also covers mixed-operator chains, chain-vs-and/or precedence, and operator text inside quoted operands.

• The reproduction script from [Bug]: Chained comparisons such as 0 <= x <= 10 in workflow expressions silently return the wrong result #5071 now prints the expected values, validate_workflow reports no errors, and the if step runs the outside branch for n = 50.

• tests/test_workflows.py + tests/workflows: all pass locally (.venv/bin/python -m pytest per CONTRIBUTING).

• Full suite: 13,649 passed; 11 failures in the script-parity tests (test_*_python_parity.py) are a pre-existing local flake — the identical 11 fail on a clean main checkout on the same machine, and each passes in isolation on both branches.

• Tested locally with uv run specify --help

• Ran existing tests with uv sync && uv run pytest

• Tested with a sample project (if applicable)

AI Disclosure

• I did not use AI assistance for this contribution
• I did use AI assistance (fill in the disclosure below)

AI disclosure: Written with Claude Code (model: Claude Fable 5, default reasoning effort, human-supervised). The agent located the defect, wrote the fix and the regression test, and ran the verification (failing-on-main test, issue repro script, full suite comparison against clean main); I reviewed the diff and the evidence before submitting. The commit carries a matching Assisted-by: trailer.

🤖 Generated with Claude Code