Community · E2 · artifact verified

Review pull requests against Five Lines of Code

A single binary that reviews a diff against the ten refactoring rules of Clausen's Five Lines of Code: the countable rules run on a real parser, and Jev answers the judgment rules as typed questions whose probabilities set the bar.

01 · Role in the system

What Jev does here

The split is the design. Six rules - method length, parameter count, and their kin - are decided by a real tree-sitter parse and counting: reproducible, no key, ten languages parsed exactly. The judgment rules - does anything guard nothing, does a condition hide a side effect, is this a framework idiom - go to Jev as typed yes-or-no questions, one per rule per site, whose probabilities are combined in code so the cutoff decides what gets raised and what is only worth a look. Without a key or with --no-jev the mechanical half still runs on its own. The README is explicit about the lens: a method can break every rule here and be correct, and a three-line method can still have a bug - run it next to a real review, never instead of one. Ships as prebuilt binaries for Windows, macOS and Linux, or via cargo.

02 · Control boundary

Where Jev sits

Tree-sitter counts the countable rules offline; Jev answers only the judgment rules as typed yes-or-no questions, and code owns combining probabilities into findings.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

  • A structural-quality lens, not a correctness review, by the README's own statement.
  • Judgment-rule coverage outside the ten exactly-parsed languages falls to Jev over unparsed text.
  • The rule set is one book's ten refactoring rules; teams with their own standards get the framework, not their rules.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.