fix(language): a comparison with the literal first is refused as the order it is

[FBumann]

Prompt: "Can we just fix the error message on main with a small pr?" — "Do the pr onto 566"

Note

The following content was generated by AI.

The refusal for two cases that compare a parameter against a literal told the author to compare one parameter against a literal. It now names the order they wrote it in and quotes the same test the other way round. Decidability is unchanged.

What this changes

capacity > 2 resolves to a ParameterComparison, which the exclusivity check reasons about: two cases over it prove apart and the model loads. The same test with the sides swapped resolves to an ArithmeticComparison, which is refused — with advice the author had already taken:

cases 'big' and 'small' cannot be told apart before the data arrives: it compares expressions, whose values only the data decides — compare one parameter against a literal, or precompute the test as a boolean parameter and test that.

Now:

cases 'big' and 'small' cannot be told apart before the data arrives: the literal is on the left, and a comparison is read as arithmetic there — write it as the same test the other way round, capacity > 2.

The number is quoted as the file wrote it — 2, not 2.0, and -2 keeps its sign — and the comparator is mirrored. Only that one shape is named: capacity > soc_initial keeps the general refusal, because two parameters genuinely carry no order before the data arrives.

Why this is on #566 and not on main

It cannot go on main: the message does not exist there. Arithmetic in a where arrives with #566 — origin/main has no ArithmeticComparison or ExpressionComparison in program.py and no such refusal in exclusivity.py. On main, 2 < capacity never reaches the exclusivity check at all; resolution refuses it earlier, accurately for that tree:

a where-comparison tests one name, relation column or position() against a literal or a second column, and a side here is arithmetic…

So #566 is what creates both the capability and the misleading message, and this sits on top of it.

What this does not do

It does not make the comparison decidable. 2 < capacity is still refused; it is only refused in words the author can act on. Teaching the check to read the plain shape out of an expression comparison is #593, which also has to land before ParameterComparison can be folded away — without it that fold would silently stop models loading.

Gates

pixi is refused by this environment's egress proxy, so the gates ran from a uv environment on Python 3.12. That is a departure from the "everything runs in a pixi environment" default.

gate	result
pytest -n auto	1384 passed, 5 skipped (base: 1379 passed)
ruff check, ruff format	clean
pyrefly check	0 errors
typos, reuse lint	clean
compile-tex	not run, for want of a TeX distribution

No generated file moves: the change is one refusal's wording, and no page or golden quotes it.

Mutation table

Guard	Caught by
the plain-shape branch in the refusal	test_a_literal_written_first_is_named_as_the_order_it_is and all three rows of test_the_rewrite_is_the_same_test_with_the_sides_swapped

Both tests were written against the old message first and watched to fail.

Coverage

• test_a_literal_written_first_is_named_as_the_order_it_is — the message names the order, quotes the rewrite, and does not print 2.0 for a 2 the file wrote.
• test_the_rewrite_is_the_same_test_with_the_sides_swapped — a signed literal, a fraction with a mirrored comparator, and !=, which is its own mirror.
• test_a_comparison_of_real_expressions_keeps_the_general_refusal — the branch is narrow: two parameters still get the old sentence.
• A quoted label is not handled, because it cannot reach here: the grammar takes a quoted value only on a comparison's right.

🤖 Generated with Claude Code

https://claude.ai/code/session_01DBU3ocfqkHe99mWmxijw64

---

Generated by Claude Code

[read-the-docs-community]

Documentation build overview

📚 math-spec | 🛠️ Build #34667778 | 📁 Comparing 96fe63a against latest (11a8a2e)

  🔍 Preview build  

10 files changed · ± 10 modified

± Modified

• CHANGELOG/index.html
• reference/notation/index.html
• reference/reading/index.html
• reference/language/expressions/index.html
• reference/math_spec/degree/index.html
• reference/math_spec/errors/index.html
• reference/math_spec/lowering/index.html
• reference/math_spec/program/index.html
• reference/math_spec/resolution/index.html
• reference/math_spec/validation/index.html