[ajianaz]

What

Part 2 of #1344. The hand-written match (method, route_path) in handlers/mod.rs becomes a data table, HANDLERS: &[(&str, &str, Handler)], plus handler_for(method, path). The 78 rows are the former arms, same handler functions, same section comments; route() now does handler_for(...) and calls the handler, with the same 404 fallback.

The source-scanning test (dispatch_match_routes_equal_the_registry_rows, which read mod.rs as text) is replaced by handler_table_equals_the_registry, which compares HANDLERS and api_registry::ENDPOINTS as data and rejects duplicates. A second test pins handler_for to the exact method and path. every_registry_row_has_a_handler (renamed from ..._arm) still proves every row reaches a handler at run time. The "keep in sync with the dispatch match" note in api_registry.rs now points at the table and the test.

No behaviour change: docs/api-reference.md and docs/core-contract.json regenerate byte-identical, so uteke-cloud's route_parity gate is unaffected (uteke-cloud#117 conditions kept).

Why

ENDPOINTS is compiled into both the lib (docgen) and the bin, so it cannot hold the handler function; the table lives bin-side and a test keeps the two row sets identical. That removes the last place where a route list was verified by grepping source text. Typed requests per operation and the request_type string stay open for #1439 phase 2.

Testing

• cargo test -p uteke-server: 37, 126, 3 and 7 tests pass; full workspace passes on a clean run (one earlier run showed a single failure that did not reproduce in two further runs; no test was changed besides the dispatch tests).
• cargo clippy --workspace --all-targets -D warnings and cargo fmt are clean.
• cargo run -p docgen leaves docs/ untouched.
• Most of the diff is the arms moving into table rows.

[github-actions]

🔍 Cora AI Code Review

✅ No issues found. Code looks good!

---

Review powered by cora-code · BYOK · MIT