feat: indexer con validacion en dos pasadas y CLI con output de errores

Reescribe el indexer con estrategia de dos pasadas:
1. Parsea todos los .md y construye sets de IDs conocidos
2. Valida integridad contra IDs conocidos, inserta solo los validos

El CLI ahora muestra INVALID para errores de validacion y ERROR
para errores de insercion, separando claramente ambos.
Añade test de integracion que verifica que entradas invalidas
se rechazan sin afectar a las validas.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

registry/indexer.go

@@ -9,13 +9,16 @@ import (
 
 // IndexResult holds stats from an indexing run.
 type IndexResult struct {
-	Functions int
-	Types     int
-	Errors    []string
+	Functions        int
+	Types            int
+	ValidationErrors []string
+	Errors           []string
 }
 
-// Index walks the registry root, parses all .md files, and populates the database.
-// It purges existing data first to ensure a clean rebuild.
+// Index walks the registry root, parses all .md files, validates integrity,
+// and populates the database. It uses two passes:
+// 1. Parse all entries and collect known IDs
+// 2. Validate references against known IDs, then insert valid entries
 func Index(db *DB, root string) (*IndexResult, error) {
 	if err := db.Purge(); err != nil {
 		return nil, fmt.Errorf("purging database: %w", err)
@@ -23,64 +26,75 @@ func Index(db *DB, root string) (*IndexResult, error) {
 
 	result := &IndexResult{}
 
-	// Index functions
+	// Pass 1: parse everything
+	var functions []*Function
+	var types []*Type
+
 	functionsDir := filepath.Join(root, "functions")
 	if _, err := os.Stat(functionsDir); err == nil {
-		err := filepath.Walk(functionsDir, func(path string, info os.FileInfo, err error) error {
-			if err != nil {
+		filepath.Walk(functionsDir, func(path string, info os.FileInfo, err error) error {
+			if err != nil || info.IsDir() || !strings.HasSuffix(path, ".md") {
 				return nil
 			}
-			if info.IsDir() || !strings.HasSuffix(path, ".md") {
-				return nil
-			}
-
 			f, err := ParseFunctionMD(path)
 			if err != nil {
-				result.Errors = append(result.Errors, fmt.Sprintf("%s: %v", path, err))
+				result.Errors = append(result.Errors, fmt.Sprintf("parse %s: %v", path, err))
 				return nil
 			}
-
-			if err := db.InsertFunction(f); err != nil {
-				result.Errors = append(result.Errors, fmt.Sprintf("insert %s: %v", f.ID, err))
-				return nil
-			}
-
-			result.Functions++
+			functions = append(functions, f)
 			return nil
 		})
-		if err != nil {
-			return nil, fmt.Errorf("walking functions: %w", err)
-		}
 	}
 
-	// Index types
 	typesDir := filepath.Join(root, "types")
 	if _, err := os.Stat(typesDir); err == nil {
-		err := filepath.Walk(typesDir, func(path string, info os.FileInfo, err error) error {
-			if err != nil {
+		filepath.Walk(typesDir, func(path string, info os.FileInfo, err error) error {
+			if err != nil || info.IsDir() || !strings.HasSuffix(path, ".md") {
 				return nil
 			}
-			if info.IsDir() || !strings.HasSuffix(path, ".md") {
-				return nil
-			}
-
 			t, err := ParseTypeMD(path)
 			if err != nil {
-				result.Errors = append(result.Errors, fmt.Sprintf("%s: %v", path, err))
+				result.Errors = append(result.Errors, fmt.Sprintf("parse %s: %v", path, err))
 				return nil
 			}
-
-			if err := db.InsertType(t); err != nil {
-				result.Errors = append(result.Errors, fmt.Sprintf("insert %s: %v", t.ID, err))
-				return nil
-			}
-
-			result.Types++
+			types = append(types, t)
 			return nil
 		})
-		if err != nil {
-			return nil, fmt.Errorf("walking types: %w", err)
+	}
+
+	// Build known ID sets
+	knownFunctions := make(map[string]bool, len(functions))
+	for _, f := range functions {
+		knownFunctions[f.ID] = true
+	}
+	knownTypes := make(map[string]bool, len(types))
+	for _, t := range types {
+		knownTypes[t.ID] = true
+	}
+
+	// Pass 2: validate and insert
+	for _, t := range types {
+		if verr := ValidateType(t, knownTypes); verr != nil {
+			result.ValidationErrors = append(result.ValidationErrors, verr.Error())
+			continue
 		}
+		if err := db.InsertType(t); err != nil {
+			result.Errors = append(result.Errors, fmt.Sprintf("insert %s: %v", t.ID, err))
+			continue
+		}
+		result.Types++
+	}
+
+	for _, f := range functions {
+		if verr := ValidateFunction(f, knownFunctions, knownTypes); verr != nil {
+			result.ValidationErrors = append(result.ValidationErrors, verr.Error())
+			continue
+		}
+		if err := db.InsertFunction(f); err != nil {
+			result.Errors = append(result.Errors, fmt.Sprintf("insert %s: %v", f.ID, err))
+			continue
+		}
+		result.Functions++
 	}
 
 	return result, nil

registry/parser_test.go

@@ -209,6 +209,9 @@ func TestIndexFullCycle(t *testing.T) {
 	if len(result.Errors) != 0 {
 		t.Errorf("unexpected errors: %v", result.Errors)
 	}
+	if len(result.ValidationErrors) != 0 {
+		t.Errorf("unexpected validation errors: %v", result.ValidationErrors)
+	}
 
 	// Verify searchable
 	fns, err := db.SearchFunctions("filter", "", "", "", "")
@@ -236,3 +239,61 @@ func TestIndexFullCycle(t *testing.T) {
 		t.Error("re-index should produce same counts")
 	}
 }
+
+const invalidPipelineMD = `---
+name: bad_pipeline
+kind: pipeline
+lang: go
+domain: core
+version: "1.0.0"
+purity: pure
+description: "Pipeline puro sin uses_functions — debe fallar."
+tags: []
+uses_functions: []
+uses_types: []
+returns: []
+returns_optional: false
+error_type: ""
+imports: []
+tested: false
+tests: []
+test_file_path: ""
+file_path: "functions/pipelines/bad.go"
+---
+`
+
+func TestIndexRejectsInvalid(t *testing.T) {
+	root := t.TempDir()
+
+	// Valid function
+	writeTempFile(t, root, "functions/core/filter_slice.md", functionMD)
+	// Invalid pipeline (pure + empty uses_functions)
+	writeTempFile(t, root, "functions/pipelines/bad.md", invalidPipelineMD)
+	// Valid type
+	writeTempFile(t, root, "types/finance/ohlcv.md", typeMD)
+
+	dbPath := filepath.Join(root, "registry.db")
+	db, err := Open(dbPath)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer db.Close()
+
+	result, err := Index(db, root)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Valid entries should be indexed
+	if result.Functions != 1 {
+		t.Errorf("functions: got %d, want 1 (only the valid one)", result.Functions)
+	}
+	if result.Types != 1 {
+		t.Errorf("types: got %d, want 1", result.Types)
+	}
+
+	// Invalid pipeline should produce validation error
+	if len(result.ValidationErrors) == 0 {
+		t.Error("expected validation errors for invalid pipeline")
+	}
+}