Skip to content

85_coverage-to-95-percent.md

Latest commit

 

History

History
139 lines (122 loc) · 5.33 KB

85_coverage-to-95-percent.md

File metadata and controls

139 lines (122 loc) · 5.33 KB
Raw
id 85
title Increase test coverage to 95% by extracting shared rule helpers
status ✅
summary Refactor duplicated private helpers into shared packages, then test the shared code once.

Increase test coverage to 95% by extracting shared rule helpers

Goal

Reach 95%+ combined test coverage (unit + e2e) by eliminating duplicated private helper functions across rule packages. Instead of testing 17 copies of the same toInt/headingLine pattern, extract them into shared packages and test once.

Background

Current combined coverage is 89.4%. Of the remaining gap, 40% comes from ~31 duplicated private helper functions spread across 15+ rule packages. Each copy sits at 40--75% coverage because fixture tests only exercise the int branch of toInt (missing float64, int64) and the ATX branch of headingLine (missing setext, fallback).

Tasks

Phase 1 --- type conversion helpers:

  • Create internal/rules/settings/settings.go with exported ToInt(v any) (int, bool), ToFloat(v any) (float64, bool), and ToStringSlice(v any) ([]string, bool)
  • Write table-driven tests in internal/rules/settings/settings_test.go covering all type branches (int, float64, int64, string, bool, nil)
  • Replace private toInt in 10 rule packages (tablereadability, tableformat, tokenbudget, linelength, paragraphreadability, paragraphstructure, concisenessscoring, nomultipleblanks, maxfilelength, firstlineheading) with settings.ToInt
  • Replace private toFloat in 4 rule packages (tablereadability, tokenbudget, paragraphreadability, concisenessscoring) with settings.ToFloat
  • Handle the emptysectionbody variant (has extra int64 validation) --- decide whether to fold that branch into the shared version or keep it local (decision: kept local because it rejects non-whole floats, which the shared ToInt accepts and truncates; same rationale for maxsectionlength)
  • Delete the now-unused private functions and their per-package coverage tests

Phase 2 --- AST heading and paragraph helpers:

  • Create internal/rules/astutil/astutil.go with HeadingLine(h, f), ParagraphLine(p, f), IsTable(p, f), HeadingText(h, src), and ExtractText(n, src, buf)
  • Write tests in internal/rules/astutil/astutil_test.go covering setext headings, ATX headings, the fallback-to-1 path, table detection, and nested-emphasis extraction
  • Replace 4 identical headingLine copies (blanklinearoundheadings, noduplicateheadings, headingincrement, notrailingpunctuation) with astutil.HeadingLine; keep the extended variants in headingstyle, firstlineheading, and emptysectionbody local
  • Replace 4 identical paragraphLine copies and 3 identical isTable copies with astutil.ParagraphLine and astutil.IsTable (noemphasisasheading also replaced)
  • Replace headingText + extractText in noduplicateheadings and notrailingpunctuation; keep the headingstyle extended version local
  • Delete unused private functions and their per-package coverage tests

Phase 3 --- error-path tests for remaining gaps:

  • internal/fix: test max-passes boundary (10 iterations without convergence)
  • internal/lint: test resolveGlob with invalid patterns, NewGitignoreMatcher with malformed gitignore files
  • internal/rules/include: test readFSFile with nonexistent and unreadable files
  • cmd/mdsmith: test formatDiagnostics write error via the error-writer pattern
  • internal/rules/requiredstructure: test cueExprForValue with []any and map[string]any inputs; test extractYAML with unclosed front matter

Phase 4 --- deep edge cases (pursue only if 95% not met after phases 1--3):

  • requiredstructure: CUE value types ([]any, map[string]any), extractYAML unclosed front matter, writeNodeText CodeSpan branch, advanceToMatch no-match path, extractPIFileParam multi-line PI
  • crossfilereferenceintegrity: anchor-only refs, encoded URLs, DefaultSettings, configDiag via invalid glob, toStringSlice mixed types
  • concisenessscoring/classifier: validateArtifact field validation (empty model_id, version, threshold, weights), compileLexicon per-list errors
  • catalog: Category, scanIncludesForTarget fallback paths (max depth, read error, no includes, direct match, cycle skip), resolveGitignore param variations

Run linter and tests after every phase:

  • go test ./... passes
  • go tool golangci-lint run reports no issues

Acceptance Criteria

  • Combined coverage (unit + e2e) reaches 95%
  • No private toInt/toFloat copies remain in the 10 + 4 packages listed above (replaced by settings.ToInt/settings.ToFloat); intentional exceptions: emptysectionbody and maxsectionlength kept local variants because they reject non-whole floats (see Phase 1 task note)
  • No private headingLine/paragraphLine copies remain for the 4-copy and 3-copy groups (replaced by astutil.HeadingLine/astutil.ParagraphLine)
  • Shared packages have 100% statement coverage
  • All tests pass: go test ./...
  • go tool golangci-lint run reports no issues