Details

Record a contributing technology. Idempotent; keeps the list sorted so the emitted order does not depend on which duplicate the optimizer saw first.

Every Endpoint#details= goes through here, and the optimizer's dedup calls it once per merged duplicate, so it copies in bulk: technologies in one sort instead of re-sorting after every add_technology. The result is what those calls produced — empty names dropped, duplicates collapsed, sorted — whatever state the source list was left in.

The project (nearest build manifest: pyproject.toml, package.json, go.mod, ...) the endpoint's first code path belongs to, relative to the scan base. It tells apart the same route served by two apps of a monorepo, which the optimizer already keeps as two endpoints.

Empty when there is no project to name: no manifest above the source, the manifest sits at the scan base itself (a single-project scan), or the endpoint is a declaration (spec, collection). Serialized only when non-empty, so reports for single-project scans stay byte-identical, and a consumer reads a missing key as "no project".

The project (nearest build manifest: pyproject.toml, package.json, go.mod, ...) the endpoint's first code path belongs to, relative to the scan base. It tells apart the same route served by two apps of a monorepo, which the optimizer already keeps as two endpoints.

Empty when there is no project to name: no manifest above the source, the manifest sits at the scan base itself (a single-project scan), or the endpoint is a declaration (spec, collection). Serialized only when non-empty, so reports for single-project scans stay byte-identical, and a consumer reads a missing key as "no project".

Every technology that produced this endpoint, sorted, deduplicated.

The optimizer collapses endpoints on (method, url, scope), and the winning duplicate keeps technology while the loser's is discarded — code_paths records that a second file also declared the route, but nothing records which analyzer read that file. Scanning casdoor whole therefore attributed 9 endpoints to oas2 when its swagger document alone yields 235: the fact was destroyed at the merge, in proportion to how well the sources agreed. This list keeps it. technology is unchanged (still the winner) and is always a member of this list.

Serialized only when two or more technologies contributed. A single entry would repeat technology on every endpoint of every report, so suppressing it keeps the JSON/YAML/TOML documents byte-identical for the unmerged majority; a consumer reads technologies when present and falls back to [technology] when absent — the same fallback it needs for output from a Noir that predates the field.

Every technology that produced this endpoint, sorted, deduplicated.

The optimizer collapses endpoints on (method, url, scope), and the winning duplicate keeps technology while the loser's is discarded — code_paths records that a second file also declared the route, but nothing records which analyzer read that file. Scanning casdoor whole therefore attributed 9 endpoints to oas2 when its swagger document alone yields 235: the fact was destroyed at the merge, in proportion to how well the sources agreed. This list keeps it. technology is unchanged (still the winner) and is always a member of this list.

Serialized only when two or more technologies contributed. A single entry would repeat technology on every endpoint of every report, so suppressing it keeps the JSON/YAML/TOML documents byte-identical for the unmerged majority; a consumer reads technologies when present and falls back to [technology] when absent — the same fallback it needs for output from a Noir that predates the field.