Skip to content

7.0.0

Choose a tag to compare

View all tags
@rohaquinlop rohaquinlop released this 10 Aug 19:51
· 7 commits to main since this release
7.0.0
d51d0ec

Major release: the refactoring suggestions are now a clippy-style lint
system with measured reductions, diff comparison gains a --staged mode
and a [tool.complexipy.diff] TOML section, and the deprecated output
flags and TOML keys are removed. Check the migration
guide before
upgrading.

Features

  • --suggest-refactors is now a clippy-style lint system: stable rule IDs (C001–C005, C007, C011), category and applicability metadata, path:line:col anchors with caret spans, verbatim suggestion rendering, and a documentation link per rule. (#209)
  • Refactor-plan findings are included in JSON, SARIF, and GitLab exports when --suggest-refactors is passed; the SARIF rule catalog is built dynamically from the plans encountered. (#209)
  • compute_diff, has_regressions, DiffEntry, and DiffStatus are now part of the public Python API, so CI tools can consume diff results as objects instead of parsing terminal output. (#210)
  • collect_removable_ignored_locations() and RemovableIgnore are exported from the Python API. Every run now reports ignore comments that are no longer necessary (path:line function=X complexity=N <comment>) when the suppressed function is back under the allowed limit. (#213)
  • --staged flag for git-index comparison — answers "what complexity am I about to commit?" instead of only what changed in the working tree; --staged alone defaults the baseline to HEAD and enforces, while --diff <ref> --staged enforces against the ref. (#218)
  • [tool.complexipy.diff] TOML section so the comparison policy lives in the repository config: branch = "main" makes a plain complexipy . run behave like --diff main (enforcement included), staged = true enables staged comparison by default, and branch = "" opts out. CLI flags take precedence over the section. (#219)
  • Refactor reductions are now measured instead of estimated: for every machine-applicable suggestion (C002, C007) the replacement is spliced into the source, re-parsed, and re-scored, so estimated_reduction and estimated_complexity_after report the literal delta of applying the suggestion — and ranking, overlap resolution, and the noise filter all operate on measured values. The new reduction_is_measured flag separates measured plans from help-only formula estimates, which the CLI renders with a ~ qualifier (Estimated reduction: ~-2) while measured plans render plain (Reduction: -2). Guard suggestions are now faithful splices for loops with statements before or after the if-chain and for multi-line loop headers; measurement failures fall back to the formula estimate — never a panic, never a fabricated number. (#225)

Fixed

  • Snapshot updates now merge with the existing snapshot instead of replacing it: only the files analyzed in a run are touched, so partial runs (e.g. pre-commit hooks analyzing only staged files) no longer erase the baseline for unanalyzed files. (#215)
  • The docs footer now renders its links as styled links instead of raw markdown text, on both the English and Spanish landing pages. (#216)

Changed

  • Refactor-plan reduction estimates are now honest: the reduction math was rewritten and validated against measured before/after complexity, C004 no longer suggests splitting match statements, C006 was deleted because its gate could never fire, and C011 now fires on try → with → try chains. Overlapping plans are deduped against every overlap and capped at 5, reporting dropped ones as "... and N more suggestions". (#209)
  • Condition extraction in the refactor rules now tracks bracket depth, string literals, and walrus := instead of a naive rfind(':'). (#209)
  • file_complexity() now returns cwd-relative paths (matching git diff --name-only), and nested invocations resolve git paths via a git ls-files basename lookup — without this, diffing per-file results silently marked every function as NEW. (#210)
  • Community standards files were added: CODE_OF_CONDUCT.md, CONTRIBUTING.md, SECURITY.md, issue templates, and a pull request template. (#201)

Removed

  • RefactorPlan.steps — replaced by the concrete suggestion / help fields on the plan. (#209)
  • CodeSnippet from the public API; CodeSuggestion is exported in its place. (#209)
  • --output-json / -j, --output-csv / -c, --output-gitlab, and --output-sarif / -sr — use --output-format instead. (#221)
  • --ratchet / -R — --diff enforces the threshold by default. (#221)
  • TOML keys output-json, output-csv, output-gitlab, output-sarif, flat staged, flat ratchet, and the undocumented details = "low" alias. (#221)

PRs

Full Changelog: 6.2.0...7.0.0

Contributors

rohaquinlop
Assets 2
Loading