[kvlonge]

Why

SELECT 1 FROM orders, the first thing many people type, was refused as SQ0003, an operand type problem. Its help suggested a cast and its span covered the whole statement. The real cause is the unaliased output name, which SQLQuack has not verified for that shape.

Changes

• New code SQ0009 unaliased_output_name, appended to the code registry and its lock. Title: "this column has no alias, and SQLQuack has not verified the name Snowflake gives it". Help: "add an alias, e.g. AS one". sqlquack explain SQ0009 documents it.
• The refusal is placed on the select item. Known gap: a bare literal still spans the statement, because the parser records no position for literals.

Verification

• A CLI end-to-end test asserts the text and JSON diagnostic (code, span, label, help).
• Core, rules, semantics and CLI tests pass; clippy and fensu check pass.
• Merged onto feat: measure translation results by catalogue form #69, the translation checker holds its baseline, with unchanged figures.
• D94 gate result follows as a comment.

[kvlonge]

D94 gate passed for head 73e55ab against main 292aca7 (clean checkout, RECORDER=0):

• make check-ci: exit 0
• make test: exit 0
• verify-recorded coverage, head and base: both exit 0, and the coverage reports are byte-identical (8,459 cases, 0 failed). Coverage didn't change.

Landing by fast-forward.