Improving ruby extractor (#58)

* Second round - Ruby callbacks and dynamic dispatch

* Another pass on calls outside method bodies

* Addressing true dynamic dispatch

* Chained no-arg calls, .rake indexing

* Default params, predicate/bang calls

* Wrapping it up

Author: dejo1307

internal/config/config.go

@@ -11,6 +11,7 @@ import (
 type Config struct {
 	Repo       string       `yaml:"repo"`
 	Ignore     []string     `yaml:"ignore"`
+	TestGlobs  []string     `yaml:"test_globs"`
 	Extractors []string     `yaml:"extractors"`
 	Explainers []string     `yaml:"explainers"`
 	Renderers  []string     `yaml:"renderers"`
@@ -68,6 +69,12 @@ func Default() *Config {
 			"**/Pods/**",
 			"**/.gradle/**",
 		},
+		// TestGlobs identify test/spec files. They stay ignored for normal indexing
+		// (still listed in Ignore above) — production architecture facts must not
+		// include test symbols — but the engine collects them separately for
+		// reference-only extraction so the dead-code detector can see that a
+		// production symbol is exercised by a test and not mis-report it as dead.
+		TestGlobs:  []string{"**/*_spec.rb", "**/*_test.rb"},
 		Extractors: []string{"cpp", "go", "java", "kotlin", "openapi", "php", "python", "typescript", "swift", "ruby"},
 		Explainers: []string{"cycles", "layers", "crossrepo", "coverage", "unused-routes", "god-class", "hotspots", "dependency-depth", "exported-surface", "complexity-outliers"},
 		Renderers:  []string{"llm_context"},

internal/engine/cache.go

@@ -31,7 +31,16 @@ import (
 // v15: C/C++ salvage function-pointer refs from file-scope ERROR regions (macro-opened structs like MACHINE_START/DT_MACHINE_START ... MACHINE_END).
 // v16: extend that salvage to file-scope assignment_expression/field_expression fragments (machine_desc blocks parse that way when surrounded by other code).
 // v17: full-tree salvage of `.field = fn` macro-struct debris (machine_desc) regardless of where tree-sitter scatters it (skips function bodies).
-const cacheVersion = "v17"
+// v18: Ruby records custom class-body macro names + resolves self/self.class receivers as call edges (dead-code precision). KindTestRef facts index outbound refs from spec/test files.
+// v19: Ruby captures call edges outside method bodies — class/module-body qualified & argument-position calls attach to the class fact; top-level/file-scope calls (fixtures, after_initialize blocks) attach to a new KindFileRef fact folded by the orphan collector.
+// v20: Ruby call capture outside method bodies generalized to a per-scope pass (whole class/module body + top-level program), so assignment RHS and all non-`call` statements — e.g. `x = GlobalSetting.foo`, `CONST = { Proc.new { Group.bar } }` — are covered, not just bare call statements.
+// v21: Ruby records the static prefix of interpolated symbols (`:"report_#{type}"` -> "report_") as a KindFileRef prop, so the dead-code detector treats dynamically dispatched (public_send/send) same-prefix methods as used.
+// v22: Ruby records `super` as a call to the same-named ancestor method, and literal-symbol dispatch args (`obj.try(:foo)`, `send(:bar)`, `respond_to?(:baz)`) as calls to the named method.
+// v23: Ruby captures no-arg calls on a chained receiver (ActiveRecord scope/class-method chains `Model.scope.final`, `assoc.class_method`, `x.class.method`; cheap attribute reads skipped) and indexes .rake/Rakefile files.
+// v24: Ruby walks method default-parameter values for calls (`def f(x = self.class.foo)`) and records single-level predicate/bang calls (`viewer.rich?`, `x.save!`) which — unlike plain attribute reads — are unambiguously method invocations.
+// v25: Ruby folds `delegate :a, :b, ..., to: X` method names as calls, and records calls on a bare-method (non-local identifier) receiver when the method name is scope-like (`some_relation.pluck_job_id`).
+// v26: Ruby records scope-like (underscored) method calls on ANY identifier receiver, including local relation variables (`items = ...; items.preload_relations`).
+const cacheVersion = "v26"
 
 // extractorCache holds per-extractor facts keyed by a content hash of the files
 // the extractor depends on. It is loaded from disk at the start of a snapshot and

internal/engine/engine.go

@@ -23,6 +23,7 @@ import (
 	"github.com/enola-labs/enola/internal/facts"
 	"github.com/enola-labs/enola/internal/linkers/crossrepo"
 	"github.com/enola-labs/enola/internal/renderers"
+	"github.com/enola-labs/enola/pkg/plugin"
 )
 
 // Engine orchestrates the snapshot generation pipeline.
@@ -184,12 +185,12 @@ func (e *Engine) GenerateSnapshot(ctx context.Context, repoPath string, appendMo
 
 	// 1. Walk repository and collect files
 	tStage := time.Now()
-	files, err := e.walkRepo(absRepo)
+	files, testFiles, err := e.walkRepo(absRepo)
 	if err != nil {
 		return nil, fmt.Errorf("walking repo: %w", err)
 	}
 	tWalk = time.Since(tStage)
-	log.Printf("[engine] found %d files in %s", len(files), absRepo)
+	log.Printf("[engine] found %d files (%d test files) in %s", len(files), len(testFiles), absRepo)
 
 	// 2. Compute file hashes (for snapshot metadata)
 	tStage = time.Now()
@@ -218,6 +219,11 @@ func (e *Engine) GenerateSnapshot(ctx context.Context, repoPath string, appendMo
 			}
 		}
 	}
+	// Reference-only extraction over test/spec files. Runs every snapshot (not
+	// cached with the main extractors) and adds only KindTestRef facts, so a
+	// production symbol exercised solely by a test is not mis-reported as dead.
+	e.runTestRefExtractors(ctx, absRepo, testFiles)
+
 	tExtract = time.Since(tStage)
 	newCount := e.store.Count()
 	log.Printf("[engine] extracted %d facts using %d extractors", newCount, len(usedExtractors))
@@ -388,10 +394,12 @@ func (e *Engine) flagUnmatchedRoutes() {
 	}
 }
 
-// walkRepo collects all files in the repo, applying ignore patterns.
-func (e *Engine) walkRepo(repoPath string) ([]string, error) {
-	var files []string
-	err := filepath.WalkDir(repoPath, func(path string, d fs.DirEntry, err error) error {
+// walkRepo collects all files in the repo, applying ignore patterns. It returns
+// the indexable source files plus, separately, the test/spec files matched by
+// TestGlobs — those are excluded from normal indexing but collected for
+// reference-only extraction (see runTestRefExtractors).
+func (e *Engine) walkRepo(repoPath string) (files, testFiles []string, err error) {
+	err = filepath.WalkDir(repoPath, func(path string, d fs.DirEntry, err error) error {
 		if err != nil {
 			return err
 		}
@@ -406,6 +414,12 @@ func (e *Engine) walkRepo(repoPath string) ([]string, error) {
 			if d.IsDir() {
 				return filepath.SkipDir
 			}
+			// An ignored FILE that is a test/spec is not indexed as production
+			// source, but is collected for reference-only extraction so a
+			// production symbol exercised only by a test does not look dead.
+			if e.matchesTestGlob(relPath) {
+				testFiles = append(testFiles, relPath)
+			}
 			return nil
 		}
 
@@ -414,7 +428,49 @@ func (e *Engine) walkRepo(repoPath string) ([]string, error) {
 		}
 		return nil
 	})
-	return files, err
+	return files, testFiles, err
+}
+
+// matchesTestGlob reports whether a repo-relative path matches any TestGlob.
+func (e *Engine) matchesTestGlob(relPath string) bool {
+	return matchAnyGlob(filepath.ToSlash(relPath), e.cfg.TestGlobs)
+}
+
+// matchAnyGlob reports whether a forward-slash path matches any of the patterns,
+// mirroring the "**/<name>/**", trailing-"/**", and "**/<glob>" handling of
+// isIgnored so test-glob matching stays consistent with ignore matching.
+func matchAnyGlob(relPath string, patterns []string) bool {
+	for _, pattern := range patterns {
+		if strings.HasPrefix(pattern, "**/") && strings.HasSuffix(pattern, "/**") {
+			seg := strings.TrimSuffix(strings.TrimPrefix(pattern, "**/"), "/**")
+			if seg != "" && !strings.Contains(seg, "/") {
+				for _, part := range strings.Split(relPath, "/") {
+					if part == seg {
+						return true
+					}
+				}
+			}
+		}
+		if strings.HasSuffix(pattern, "/**") {
+			dirPrefix := strings.TrimSuffix(pattern, "/**")
+			if relPath == dirPrefix || strings.HasPrefix(relPath, dirPrefix+"/") {
+				return true
+			}
+		}
+		if m, err := filepath.Match(pattern, relPath); err == nil && m {
+			return true
+		}
+		if strings.HasPrefix(pattern, "**/") {
+			sub := strings.TrimPrefix(pattern, "**/")
+			if m, err := filepath.Match(sub, filepath.Base(relPath)); err == nil && m {
+				return true
+			}
+			if m, err := filepath.Match(sub, relPath); err == nil && m {
+				return true
+			}
+		}
+	}
+	return false
 }
 
 // isIgnored checks whether a path matches any ignore pattern.
@@ -530,6 +586,47 @@ func (e *Engine) runExtractors(ctx context.Context, repoPath string, files []str
 	return usedNames, nil
 }
 
+// runTestRefExtractors runs reference-only extraction over the test/spec files
+// for every enabled, detected extractor that implements plugin.TestRefExtractor.
+// It scopes each extractor to the test files it owns and adds the resulting
+// KindTestRef facts to the store. Errors are logged, not fatal.
+func (e *Engine) runTestRefExtractors(ctx context.Context, repoPath string, testFiles []string) {
+	if len(testFiles) == 0 {
+		return
+	}
+	for _, ext := range e.extractors.All() {
+		if !e.cfg.IsExtractorEnabled(ext.Name()) {
+			continue
+		}
+		tr, ok := ext.(plugin.TestRefExtractor)
+		if !ok {
+			continue
+		}
+		if detected, err := ext.Detect(repoPath); err != nil || !detected {
+			continue
+		}
+		owned := testFiles
+		if fo, ok := ext.(plugin.FileOwner); ok {
+			owned = owned[:0:0]
+			for _, f := range testFiles {
+				if fo.OwnsFile(f) {
+					owned = append(owned, f)
+				}
+			}
+		}
+		if len(owned) == 0 {
+			continue
+		}
+		refFacts, err := tr.ExtractTestRefs(ctx, repoPath, owned)
+		if err != nil {
+			log.Printf("[engine] extractor %s test-ref error: %v", ext.Name(), err)
+			continue
+		}
+		e.store.Add(refFacts...)
+		log.Printf("[engine] extractor %s: emitted %d test-ref facts from %d files", ext.Name(), len(refFacts), len(owned))
+	}
+}
+
 // runExplainers runs all enabled explainers.
 func (e *Engine) runExplainers(ctx context.Context) ([]facts.Insight, []string, error) {
 	var allInsights []facts.Insight

internal/extractors/rubyextractor/ruby.go

@@ -112,6 +112,35 @@ func (e *RubyExtractor) Extract(ctx context.Context, repoPath string, files []st
 	return allFacts, nil
 }
 
+// ExtractTestRefs implements plugin.TestRefExtractor. It parses test/spec files
+// for the SOLE purpose of capturing their outbound references into production
+// code, emitting one facts.KindTestRef fact per file that carries only RelCalls
+// edges — no symbols. Test methods therefore never become dead-code candidates
+// (which the orphans package explicitly excludes), and no symbol/module/route
+// explainer is affected, while the dead-code detector can still see that a
+// production symbol is exercised by a test and not mis-report it as dead.
+func (e *RubyExtractor) ExtractTestRefs(ctx context.Context, repoPath string, files []string) ([]facts.Fact, error) {
+	var rbFiles []string
+	for _, relFile := range files {
+		if isRubyFile(relFile) {
+			rbFiles = append(rbFiles, relFile)
+		}
+	}
+	perFile := parallel.MapFiles(ctx, rbFiles, func(relFile string) []facts.Fact {
+		src, err := os.ReadFile(filepath.Join(repoPath, relFile))
+		if err != nil {
+			log.Printf("[ruby-extractor] error reading test file %s: %v", relFile, err)
+			return nil
+		}
+		return extractTestRefsAST(src, relFile)
+	})
+	var out []facts.Fact
+	for _, ff := range perFile {
+		out = append(out, ff...)
+	}
+	return out, nil
+}
+
 // --- Rails detection ---
 
 func detectRailsProject(repoPath string) bool {
@@ -127,9 +156,15 @@ func detectRailsProject(repoPath string) bool {
 	return false
 }
 
-// isRubyFile returns true if the file has a .rb extension.
+// isRubyFile returns true if the file is Ruby source: a .rb/.rake file or a
+// Rakefile. Rake tasks are Ruby and call into app code (e.g. from lib/tasks/),
+// so indexing them lets those calls resolve (dead-code precision).
 func isRubyFile(path string) bool {
-	return strings.HasSuffix(strings.ToLower(path), ".rb")
+	lower := strings.ToLower(path)
+	if strings.HasSuffix(lower, ".rb") || strings.HasSuffix(lower, ".rake") {
+		return true
+	}
+	return filepath.Base(path) == "Rakefile"
 }
 
 // OwnsFile implements plugin.FileOwner for incremental caching.