cognitive_complexity

cognitive_complexity 0.4.0

cognitive_complexity 0.3.0

• Auto-discover default targets when no positional paths are given: Pub
  workspace members (<member>/lib), single-package lib/, or packages/*/lib
  / pkgs/*/lib monorepo layouts. The GitHub Action targets input now
  defaults to auto-discovery, so workspace and monorepo callers no longer need a
  hardcoded target list (#137).
• Order --git-diff delta tables by review significance in every format (step
  summary, PR comment, text, json): violations first, then new score
  descending, then delta descending, then location.
  DeltaAnalyzer.computeDeltas gains failThreshold, maxFunctionLines, and
  failOnIncrease parameters and the comparator is exposed as
  compareBySignificance (#53).

• Omit zero-complexity (delta == 0, score == 0) added and removed
  declarations from --git-diff Markdown, CLI, and JSON delta tables (unless
  they trigger --max-function-lines), add DeltaSummary.countRemoved (🗑️),
  and format added/removed Score cells as _new_ -> **N** and
  N -> _deleted_.
• Add shallow CLI (dart run cognitive_complexity:shallow) and
  ShallowAnalyzer (ShallowReport, ShallowFinding, ShallowClassification)
  to detect single-caller shallow helpers (HIGH_ARITY, MICRO_HELPER,
  SIG_HEAVY, CROSS_FILE_SINGLE_CALLER) and simulate exact nesting-aware
  caller Cognitive Complexity after re-inlining (SAFE_INLINE,
  FLATTEN_AND_INLINE, LOAD_BEARING).
• Enhance DataFlowAnalyzer and DataFlowResult
  (cognitive_complexity:data_flow) with nesting-aware complexity impact
  metrics (enclosingScore, sliceScoreInPlace, sliceScoreAtRoot,
  estimatedEnclosingScoreAfter) and extractionWarnings when a proposed slice
  requires >= 5 input parameters or provides low complexity payoff.
• Add opt-in --max-file-lines and --max-function-lines CLI flags and GitHub
  Action inputs (max-file-lines, max-function-lines) with --git-diff
  pragmatic ratchet support (FileLineMetric, FileLineDelta). Disabled
  (null) by default to preserve full backward compatibility.
• Support // cognitive_complexity:ignore (per-declaration) and
  // cognitive_complexity:ignore_for_file (file-wide) suppression comment
  directives via CommentDirectiveParser.
• Add file_split CLI (dart run cognitive_complexity:file_split) and
  FileSplitAnalyzer (analyzeFile / analyzeFiles) to compute acyclic file
  decomposition cuts targeting coarse <= 800-line subsystems using Tarjan's
  SCC condensation, immediate dominator cone absorption, dynamic shared-tail
  diamond re-evaluation, one-shot sufficiency & surplus small-cut re-absorption,
  zero-crossing sibling cone coalescing, static method promotion and embedded
  string/asset literal (>75%) detection for oversized classes, multi-file and
  directory batch execution, --[no-]use-parts support, and zero-churn
  export ... show barrel generation.
• Skip writing the --comment-output file entirely when a run has zero
  violations and zero complexity increases, so the GitHub Action stays quiet on
  the PR thread instead of posting an all-zeroes summary comment. The full
  report still lands in $GITHUB_STEP_SUMMARY. A diff that only improves
  complexity now also stays quiet.
• Add DeltaSummary.isClean(), the predicate behind that decision.
• GitHub Action: when a previously reported PR gets clean, the existing sticky
  comment is updated in place to a resolved status (never deleted), and only
  a successful run is treated as clean, so a crashed scan can no longer mark
  stale findings resolved. --comment-output is only requested when the scanner
  runs in github format against a diff base.
• Support --exclude and --[no-]ignore-generated CLI options to configure
  file exclusion and generated code filtering via PathFilter.
• Add pathFilter parameter to ComplexityAnalyzer and DeltaAnalyzer.
• Remove noisy GitHub workflow ::warning annotations on non-violating
  complexity increases, keeping inline annotations reserved for violations
  (::error) while full deltas remain tracked in step summaries and PR comment
  tables.
• Add ensure_cli_readme_test.dart to verify CLI --help documentation in
  README.md.
• Update GitHub Action documentation in README.md to reference modular
  subdirectory path (packages/cognitive_complexity) and add rendered Action
  Inputs Reference table.

• Fix action.yml monorepo auto-detection to evaluate default lib vs
  packages existence against the caller's $GITHUB_WORKSPACE instead of
  $ACTION_PATH (#76).
• Fix action.yml to run the scanner from the caller's $GITHUB_WORKSPACE
  rather than $ACTION_PATH, so relative targets and --git-diff resolve
  against the repository under audit when the action is used from another
  repository (#76).
• Add --comment-output and --max-comment-rows CLI options: with
  --format=github, write a second report capped to the most significant rows
  (violations, then increases, then additions) for posting as a PR comment,
  while the step summary keeps the full table (#49).
• Add commentFile and maxCommentRows parameters to GitHubReporter.
• Add max-comment-rows input to the GitHub Action (default 0 = unlimited)
  to keep sticky comments under GitHub's 65536-character body limit.
• Anchor GitHub annotations to the declaration line only, so they render
  under the function signature instead of after the closing brace (#55).
• Move package into pub workspace monorepo layout under packages/cognitive_complexity.
• Restore root action.yml and skills/ structure for monorepo workspace.

• Fix data_flow crashing with PathNotFoundException when run as a
  standalone AOT executable (dart run cognitive_complexity:data_flow@ /
  dart install), where SDK auto-discovery resolved to the app bundle
  (#21):
  • Fail fast with an actionable error (pass --sdk-path, set DART_SDK,
    or put an SDK on PATH) instead of handing the analyzer an unresolved
    SDK location.
  • Teach PATH-based discovery the Flutter checkout layout: a
    flutter/bin/dart wrapper script now resolves to
    flutter/bin/cache/dart-sdk.
  • Add a FLUTTER_ROOT discovery fallback.
  • Add a --sdk-path option to the data_flow CLI (plumbed to the
    existing DataFlowAnalyzer.sdkPath parameter); invalid explicit paths
    are rejected with a clear error.
• Fix --fail-on-increase ignoring --fail-threshold: when both are set,
  an increase now only fails the run if the new score exceeds the
  threshold. Sub-threshold increases are still reported (delta table, PR
  comment) but no longer block healthy changes such as added error
  handling. Without --fail-threshold, the strict any-increase ratchet is
  unchanged.
• Update the dart-cognitive-complexity agent skill:
  • Add explicit rule under Tier 1/2 prescribing that extracted private helper
    routines that do not read or mutate class instance state (this) must be
    declared as private top-level functions (or static methods) rather than
    instance methods.

• Add an example.
• Remove accidental export of internal CognitiveComplexityVisitor from
  package:cognitive_complexity/cognitive_complexity.dart.

• Streamline README.md into a focused, scannable TL;DR.
• Extract deep-dive reference documentation into modular guides under doc/:
  • doc/scoring.md: Scoring matrix, nesting multipliers, and Dart 3 AST rules.
  • doc/cli.md: Command-line options, git diff delta evaluation, and CI ratcheting.
  • doc/data_flow.md: Statement data-flow analysis, method extraction theory, and programmatic API.
  • doc/github_actions.md: Complete GitHub Action workflow setup, parameters, and fork security permissions.

• Fix column wrapping and alignment formatting in markdown report tables.

• Update the dart-cognitive-complexity agent skill: anti-goodhart 3-tier decomposition rubric with deterministic tier selection via the data_flow CLI, and a gated Tier 3 method-object reference (references/method-object.md) absorbed from kevmoo/kevmoo_skills.
• No library or CLI code changes.

• Refine the skill's 3-Tier Decomposition Rubric per empirical feedback
  (#19):
  • "Slice at natural seams" rule: enlarge slices to whole loops/state
    machines before classifying; solves coupled 2-variable state
    (queue + visited) without weakening the Tier 3 gate (stays at >= 3
    with recorded evidence).
  • Named records for private/file-local slices at any output count;
    dedicated Result dataclasses reserved for public API boundaries
    (aligns the rubric with the data_flow signature synthesizer).
  • Pattern E: loop-body extraction with explicit signal returns
    (enum/sealed) instead of boolean control-flow flags; ordered
    dispositions for control-flow escapes.
  • Domain-modeling exit: promoting coupled state to a real named, tested
    domain class is out of rubric scope — the gate bans _Runner facades,
    not real abstractions.
• No library or CLI code changes.

• Add support for GitHub Actions step summary and comment outputs.

• Add Data-Flow analysis and helper extraction engine.

• Initial release of reachability and dead/unused declaration analyzer.
• Support for CLI execution, package entrypoint harvesting, and Flutter/test
  framework adapters.
• Deterministic analysis modes: open-world (libraries) and closed-app
  (standalone/executables).
• Custom suppression directives: // undead:ignore and
  // undead:ignore_for_file.