Changelog

Copy Markdown View Source

All notable changes to this project will be documented in this file.

0.8.0

  • New check Rbtz.CredoChecks.Refactor.PreferWithOverCase — flags a two-clause case whose only non-happy-path clause re-returns an error shape unchanged: an {:error, ...} tuple or the bare :error atom (e.g. {:error, reason} -> {:error, reason}). A with expresses this more directly, since its implicit else passes the non-matching value through. Non-error pass-through clauses (nil -> nil, other -> other), guarded clauses, a catch-all happy path paired with the error pass-through, identity cases, and cases with one or three-plus clauses are left alone.

0.7.1

  • AtomHttpStatusCodes: now also flags integer status codes passed to the Phoenix.ConnTest assertion helpers response/2, json_response/2, html_response/2, and redirected_to/2 (e.g. json_response(conn, 200) → json_response(conn, :ok)).

0.7.0

  • New Rbtz.CredoChecks.all/0 — returns every check in this library as a {module, opts} tuple, ready to append to the enabled: list in .credo.exs. The list is auto-discovered, so it stays in sync as checks are added. This is now the recommended way to enable the checks (see the README).
  • New Rbtz.CredoChecks.all_replacing/1 — like all/0, but takes {module, opts} replacements to customise a check's options, or {module, false} to disable it; an unknown module raises ArgumentError.

0.6.0

  • New check Rbtz.CredoChecks.Readability.ModuleAttrCollectionFormatting — flags multi-line collections assigned to module attributes (word sigils, lists, maps/structs, and tuples) that put more than one item on a line. Each item should sit on its own line so the collection is easy to scan, diff, and update. Single-line collections are unaffected, however many items they hold.

0.5.0

  • New check Rbtz.CredoChecks.Readability.PreferBlockFormForMultilineIf — forbids the keyword form if cond, do: x, else: y (and the unless equivalent) when the expression spans more than one line. Switch to the do ... else ... end block form. Single-line keyword form and block form are unaffected.

0.4.0

  • New check Rbtz.CredoChecks.Readability.PreferNilEquality — prefers x == nil / x != nil over is_nil(x) / not is_nil(x) / !is_nil(x) in if / unless / cond / case conditions and assert / refute arguments. Walks through and / or / && / || / not / !. Guards (when ...) and Ecto query DSL are unaffected.

0.3.0

  • AwkwardPipe: new sub-rule (Rule 10) flags single-step pipes wrapped in parentheses just to dot-access their result — e.g. (x |> URI.parse()).host should be URI.parse(x).host.

0.2.0

  • New check Rbtz.CredoChecks.Readability.RedundantClassAttrWrapping — flags HEEx class={...} attributes whose wrapping is unnecessary (class={"foo"} / class={["foo"]} → class="foo"; class={[expr]} → class={expr}).
  • ClassAttrFormatting: the max-line-length rule now scans every source line a class={...} / class="..." attribute spans, not just the opening line. Long string literals buried inside a multi-line list are now flagged, and the issue is reported on the offending inner line.
  • Fix: HeexSource now reports line numbers correctly for ~H""" ... """ heredoc sigils. Previously every issue was off by one because the AST's line: metadata points at the opening ~H""" while content starts on the next line. Affects every HEEx-scanning check.

0.1.0

Initial release — 39 Credo checks (6 Design, 11 Readability, 6 Refactor, 16 Warning) extracted from internal Tiny Robots use. See the README for the full list.