explore_mappings cuts twice and reports neither.
The schema text stops at MAX_SCHEMA_CHARS (12 000) — and the break there leaves only the current source's column loop, so with several sources each one overruns separately and the constant is not the bound it reads as. Proposals stop at take(12), a fixed number whether the source has three tables or three hundred.
Nothing downstream says how much was covered. The only signal is the alert raised when accepted == 0. Eleven proposals against an eighty-column wide table and eleven against a complete small schema render identically on the page.
What to add
- A bound that scales with the source —
min(tables × 3, 60) or similar — instead of a fixed twelve, and a cap on the schema text that actually holds across sources.
- Record the run: tables scanned, columns scanned, proposals accepted, which tables got at least one. Show it on the Data Mappings page.
This is the only signal that would reveal a gap once confirmation stops passing through a person, and it is what #501 needs to report covered against a real base rather than only against the bench corpora.
explore_mappingscuts twice and reports neither.The schema text stops at
MAX_SCHEMA_CHARS(12 000) — and thebreakthere leaves only the current source's column loop, so with several sources each one overruns separately and the constant is not the bound it reads as. Proposals stop attake(12), a fixed number whether the source has three tables or three hundred.Nothing downstream says how much was covered. The only signal is the alert raised when
accepted == 0. Eleven proposals against an eighty-column wide table and eleven against a complete small schema render identically on the page.What to add
min(tables × 3, 60)or similar — instead of a fixed twelve, and a cap on the schema text that actually holds across sources.This is the only signal that would reveal a gap once confirmation stops passing through a person, and it is what #501 needs to report
coveredagainst a real base rather than only against the bench corpora.