Comments and strings are never read as code (parsing audit 2026-10-10)

eval/go.mod

@@ -14,7 +14,7 @@ require (
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/mattn/go-isatty v0.0.20 // indirect
 	github.com/ncruces/go-strftime v1.0.0 // indirect
-	github.com/provasign/astkit v0.15.4 // indirect
+	github.com/provasign/astkit v0.16.0 // indirect
 	github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
 	github.com/smacker/go-tree-sitter v0.0.0-20240827094217-dd81d9e9be82 // indirect
 	github.com/sourcegraph/beaut v0.0.0-20240611013027-627e4c25335a // indirect

eval/go.sum

@@ -18,8 +18,8 @@ github.com/ncruces/go-strftime v1.0.0 h1:HMFp8mLCTPp341M/ZnA4qaf7ZlsbTc+miZjCLOF
 github.com/ncruces/go-strftime v1.0.0/go.mod h1:Fwc5htZGVVkseilnfgOVb9mKy6w1naJmn9CehxcKcls=
 github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 h1:Jamvg5psRIccs7FGNTlIRMkT8wgtp5eCXdBlqhYGL6U=
 github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/provasign/astkit v0.15.4 h1:Zdo9Dmu6hOcXh8iAijK9nUUIrsbLSdcZjobaprqdX1c=
-github.com/provasign/astkit v0.15.4/go.mod h1:uSQxIxw3dVpK/6KmHvqIE8+DK/nhxDuPklKv7g96Jr4=
+github.com/provasign/astkit v0.16.0 h1:0CGyrZex64wlqK9z6wiPyOLEk26Mhfi+yjWN1xGUaJk=
+github.com/provasign/astkit v0.16.0/go.mod h1:uSQxIxw3dVpK/6KmHvqIE8+DK/nhxDuPklKv7g96Jr4=
 github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec h1:W09IVJc94icq4NjY3clb7Lk8O1qJ8BdBEF8z0ibU0rE=
 github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec/go.mod h1:qqbHyh8v60DhA7CoWK5oRCqLrMHRGoxYCSS9EjAz6Eo=
 github.com/scip-code/scip/bindings/go/scip v0.8.1 h1:ZkyT9RDgkbd6ffvMbi9nYdwh7+gLoz8Z9ftFMxgUiVs=

go.mod

@@ -5,7 +5,7 @@ go 1.26
 require modernc.org/sqlite v1.51.0
 
 require (
-	github.com/provasign/astkit v0.15.4
+	github.com/provasign/astkit v0.16.0
 	github.com/smacker/go-tree-sitter v0.0.0-20240827094217-dd81d9e9be82
 	golang.org/x/sys v0.46.0
 )

go.sum

@@ -14,8 +14,8 @@ github.com/ncruces/go-strftime v1.0.0 h1:HMFp8mLCTPp341M/ZnA4qaf7ZlsbTc+miZjCLOF
 github.com/ncruces/go-strftime v1.0.0/go.mod h1:Fwc5htZGVVkseilnfgOVb9mKy6w1naJmn9CehxcKcls=
 github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 h1:Jamvg5psRIccs7FGNTlIRMkT8wgtp5eCXdBlqhYGL6U=
 github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/provasign/astkit v0.15.4 h1:Zdo9Dmu6hOcXh8iAijK9nUUIrsbLSdcZjobaprqdX1c=
-github.com/provasign/astkit v0.15.4/go.mod h1:uSQxIxw3dVpK/6KmHvqIE8+DK/nhxDuPklKv7g96Jr4=
+github.com/provasign/astkit v0.16.0 h1:0CGyrZex64wlqK9z6wiPyOLEk26Mhfi+yjWN1xGUaJk=
+github.com/provasign/astkit v0.16.0/go.mod h1:uSQxIxw3dVpK/6KmHvqIE8+DK/nhxDuPklKv7g96Jr4=
 github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec h1:W09IVJc94icq4NjY3clb7Lk8O1qJ8BdBEF8z0ibU0rE=
 github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec/go.mod h1:qqbHyh8v60DhA7CoWK5oRCqLrMHRGoxYCSS9EjAz6Eo=
 github.com/smacker/go-tree-sitter v0.0.0-20240827094217-dd81d9e9be82 h1:6C8qej6f1bStuePVkLSFxoU22XBS165D3klxlzRg8F4=

internal/graph/cfamilylocaltypes.go

@@ -50,11 +50,9 @@ func cFamilyLocalTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]str
 					default:
 						continue
 					}
-					body := cls.RawText
-					if i := strings.IndexByte(body, '{'); i >= 0 {
-						body = body[i+1:]
-					}
-					for _, m := range cppLocalDeclRe.FindAllStringSubmatch(body, -1) {
+					// Top-level members only (classBody): an inline method's
+					// local or a commented-out declaration is not a field.
+					for _, m := range cppLocalDeclRe.FindAllStringSubmatch(classBody(idx, cls), -1) {
 						if t := cFamilyBareType(m[1]); t != "" {
 							if _, exists := out[m[2]]; !exists {
 								record(m[2], t)
@@ -69,14 +67,20 @@ func cFamilyLocalTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]str
 		}
 	}
 
-	// Parameters.
-	for name, typ := range cFamilyParamTypes(symbol.Signature, symbol.RawText) {
+	// Parameters. Read on masked text: a symbol whose signature has no
+	// parameter list falls back to its body, and the first `(` there
+	// could be in a comment (`// ghost() { helper(1) }`), which hid the
+	// real parameters.
+	body := ""
+	if symbol.RawText != "" {
+		body = maskCode(symbol.Language, symbol.RawText)
+	}
+	for name, typ := range cFamilyParamTypes(maskCode(symbol.Language, symbol.Signature), body) {
 		record(name, typ)
 	}
 
 	// Body locals (highest precedence): new-expressions.
-	if symbol.RawText != "" {
-		body := stripCommentsAndStrings(symbol.RawText)
+	if body != "" {
 		for _, m := range cppLocalDeclRe.FindAllStringSubmatch(body, -1) {
 			if t := cFamilyBareType(m[1]); t != "" {
 				record(m[2], t)

internal/graph/changeimpact.go

@@ -739,10 +739,7 @@ func declaredSuperNames(s *core.SymbolRecord) []string {
 		if s.Kind != core.KindClass && s.Kind != core.KindInterface {
 			return nil
 		}
-		text := s.Signature
-		if text == "" {
-			text = firstLine(s.RawText)
-		}
+		text := declHeader(s)
 		text = stripAngleBrackets(text)
 		names := matchNameList(extendsRe, text)
 		return append(names, matchNameList(implementsRe, text)...)
@@ -753,29 +750,20 @@ func declaredSuperNames(s *core.SymbolRecord) []string {
 		if s.Kind != core.KindClass && s.Kind != core.KindInterface && s.Kind != core.KindStruct {
 			return nil
 		}
-		text := s.Signature
-		if text == "" {
-			text = firstLine(s.RawText)
-		}
+		text := declHeader(s)
 		return csharpBaseNames(stripAngleBrackets(text))
 	case "swift":
 		if s.Kind != core.KindClass && s.Kind != core.KindStruct &&
 			s.Kind != core.KindEnum && s.Kind != core.KindInterface {
 			return nil
 		}
-		text := s.Signature
-		if text == "" {
-			text = firstLine(s.RawText)
-		}
+		text := declHeader(s)
 		return swiftBaseNames(text)
 	case "kotlin":
 		if s.Kind != core.KindClass && s.Kind != core.KindInterface {
 			return nil
 		}
-		text := s.Signature
-		if text == "" {
-			text = firstLine(s.RawText)
-		}
+		text := declHeader(s)
 		var out []string
 		for _, raw := range kotlinBaseNames(text) {
 			if name, _ := kotlinBaseNameAndCtor(raw); name != "" {
@@ -787,10 +775,7 @@ func declaredSuperNames(s *core.SymbolRecord) []string {
 		if s.Kind != core.KindClass && s.Kind != core.KindInterface {
 			return nil
 		}
-		text := s.Signature
-		if text == "" {
-			text = firstLine(s.RawText)
-		}
+		text := declHeader(s)
 		var out []string
 		if m := objcSuperclassRe.FindStringSubmatch(text); len(m) == 2 {
 			out = append(out, m[1])
@@ -800,17 +785,7 @@ func declaredSuperNames(s *core.SymbolRecord) []string {
 		if s.Kind != core.KindClass {
 			return nil
 		}
-		m := pythonClassBase.FindStringSubmatch(firstLine(s.RawText))
-		if len(m) < 2 {
-			return nil
-		}
-		var out []string
-		for _, base := range splitTrim(m[1], ',') {
-			if base = stripPythonBase(base); base != "" {
-				out = append(out, base)
-			}
-		}
-		return out
+		return pythonClassBases(s)
 	}
 	// Go and Rust have NO declared-supertype clause on the type: Go interface
 	// satisfaction is structural (derived by method-set inclusion, and only
@@ -900,7 +875,7 @@ func (g *CodeGraph) typeDeclaresMember(t *core.SymbolRecord, method string) bool
 	default:
 		return false
 	}
-	body := stripCommentsAndStrings(t.RawText)
+	body := maskCode(t.Language, t.RawText)
 	if i := strings.IndexByte(body, '{'); i >= 0 {
 		body = body[i+1:]
 	}
@@ -1357,7 +1332,7 @@ func paramTypesOf(s *core.SymbolRecord) []string {
 	if !strings.Contains(src, ")") {
 		src = s.RawText
 	}
-	inner := tsDeclParams(src)
+	inner := tsDeclParams(s.Language, src)
 	if s.Language == "go" {
 		gi, ok := goDeclParamsOK(src)
 		if !ok {
@@ -1383,7 +1358,7 @@ func paramTypesOf(s *core.SymbolRecord) []string {
 		return pyParamTypeTokens(inner)
 	}
 	var out []string
-	for _, gr := range splitTopLevel(inner, ',') {
+	for _, gr := range splitParams(s.Language, inner) {
 		token := bareTypeToken(gr)
 		switch s.Language {
 		case "typescript", "tsx", "javascript":
@@ -1405,7 +1380,7 @@ func paramTypesOf(s *core.SymbolRecord) []string {
 // so one unannotated parameter makes the whole list neutral (nil).
 func pyParamTypeTokens(inner string) []string {
 	var out []string
-	for i, gr := range splitTopLevel(inner, ',') {
+	for i, gr := range splitParams("python", inner) {
 		gr = strings.TrimSpace(gr)
 		name, ann, annotated := strings.Cut(gr, ":")
 		name = strings.TrimSpace(name)
@@ -1443,7 +1418,7 @@ func pyParamTypeTokens(inner string) []string {
 // Returns nil (neutral evidence, like paramTypesOf) when a token cannot be
 // recovered.
 func goParamTokens(inner string) []string {
-	groups := splitTopLevel(inner, ',')
+	groups := splitParams("go", inner)
 	n := len(groups)
 	types := make([]string, n)
 	named := false

internal/graph/csharp_attribute_types_test.go

@@ -0,0 +1,36 @@
+package graph
+
+import (
+	"reflect"
+	"testing"
+
+	"github.com/provasign/grove/internal/core"
+)
+
+func TestCSharpAttributeTypeNames(t *testing.T) {
+	idx := newEdgeIndex([]core.SymbolRecord{
+		{ID: "a.cs::JsonPropertyAttribute", FilePath: "a.cs", Language: "csharp", Kind: core.KindClass,
+			Name: "JsonPropertyAttribute", QualifiedName: "JsonPropertyAttribute"},
+	})
+	for _, tc := range []struct {
+		ann  string
+		want []string
+	}{
+		{`[JsonConverter(typeof(StringEnumConverter), typeof(CamelCaseNamingStrategy))]`,
+			[]string{"JsonConverter", "StringEnumConverter", "CamelCaseNamingStrategy"}},
+		{`[JsonProperty(NullValueHandling = NullValueHandling.Ignore, Order = 2)]`,
+			[]string{"JsonPropertyAttribute", "NullValueHandling"}},
+		{`[DataMember(Name = "first_name")]`, []string{"DataMember"}},
+		{`[Obsolete("Use TypeNameAssemblyFormatHandling instead.")]`, []string{"Obsolete"}},
+		{`[global::System.Data.Linq.Mapping.ColumnAttribute(Storage="_Name", CanBeNull=false)]`, []string{"ColumnAttribute"}},
+		{`[CompilationMapping(SourceConstructFlags.UnionCase, 1)]`, []string{"CompilationMapping", "SourceConstructFlags"}},
+		{`[return: NotNull, Flags]`, []string{"NotNull", "Flags"}},
+		{`[JsonConverter(typeof(Outer.Converter), new object[] { Mode.Fast /* Fake.Thing */ })]`,
+			[]string{"JsonConverter", "Converter", "Mode"}},
+	} {
+		s := core.SymbolRecord{Language: "csharp", Annotations: []string{tc.ann}}
+		if got := csharpAttributeTypeNames(idx, &s); !reflect.DeepEqual(got, tc.want) {
+			t.Errorf("%s: got %v, want %v", tc.ann, got, tc.want)
+		}
+	}
+}

internal/graph/csharplocaltypes.go

@@ -44,14 +44,9 @@ func csharpLocalTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]stri
 				}
 				seen[className] = true
 				for _, cls := range csharpTypeFragments(idx, className, dirOf(symbol.FilePath)) {
-					if cls.RawText == "" {
-						continue
-					}
-					body := cls.RawText
-					if i := strings.IndexByte(body, '{'); i >= 0 {
-						body = body[i+1:]
-					}
-					for _, m := range csharpFieldRe.FindAllStringSubmatch(body, -1) {
+					// Top-level members only (classBody): a method's local
+					// `Foo helper = ...` or a commented-out field is not one.
+					for _, m := range csharpFieldRe.FindAllStringSubmatch(classBody(idx, cls), -1) {
 						if t := javaBareType(m[1]); t != "" && !csharpKeyword(t) {
 							if _, exists := out[m[2]]; !exists {
 								out[m[2]] = t
@@ -66,8 +61,8 @@ func csharpLocalTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]stri
 	}
 
 	// Parameters: "Type name" pairs from the declaration's paren group.
-	if params := tsDeclParams(symbol.RawText); params != "" {
-		for _, g := range splitTopLevel(params, ',') {
+	if params := tsDeclParams(symbol.Language, symbol.RawText); params != "" {
+		for _, g := range splitParams(symbol.Language, params) {
 			fields := strings.Fields(strings.TrimSpace(g))
 			// Drop C# parameter modifiers and attributes.
 			for len(fields) > 2 || (len(fields) == 2 && csharpParamModifier(fields[0])) {
@@ -84,7 +79,7 @@ func csharpLocalTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]stri
 
 	// Body declarations (highest precedence).
 	if symbol.RawText != "" {
-		body := stripCommentsAndStrings(symbol.RawText)
+		body := maskCode(symbol.Language, symbol.RawText)
 		if i := strings.IndexByte(body, '{'); i >= 0 {
 			body = body[i+1:]
 		}
@@ -262,11 +257,11 @@ func csDeclSource(s *core.SymbolRecord) string {
 // csParamTypes parses a C# callable's parameter type tokens (normalized)
 // and whether the last one is a `params` array.
 func csParamTypes(s *core.SymbolRecord) (types []string, variadic bool) {
-	params := tsDeclParams(csDeclSource(s))
+	params := tsDeclParams("csharp", csDeclSource(s))
 	if params == "" {
 		return nil, false
 	}
-	for _, g := range splitTopLevel(params, ',') {
+	for _, g := range splitParams("csharp", params) {
 		g = strings.TrimSpace(g)
 		if eq := strings.Index(g, "="); eq >= 0 {
 			g = g[:eq] // default value
@@ -314,8 +309,8 @@ func csharpArgTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]string
 			out[name] = typ
 		}
 	}
-	if params := tsDeclParams(csDeclSource(symbol)); params != "" {
-		for _, g := range splitTopLevel(params, ',') {
+	if params := tsDeclParams("csharp", csDeclSource(symbol)); params != "" {
+		for _, g := range splitParams("csharp", params) {
 			g = strings.TrimSpace(g)
 			if eq := strings.Index(g, "="); eq >= 0 {
 				g = g[:eq]
@@ -330,7 +325,7 @@ func csharpArgTypes(idx *edgeIndex, symbol *core.SymbolRecord) map[string]string
 		}
 	}
 	if symbol.RawText != "" {
-		body := stripCommentsAndStrings(symbol.RawText)
+		body := maskCode(symbol.Language, symbol.RawText)
 		if i := strings.IndexByte(body, '{'); i >= 0 {
 			body = body[i+1:]
 		}
@@ -936,7 +931,8 @@ func csRecoverArgs(idx *edgeIndex, symbol *core.SymbolRecord, cs core.CallSite,
 	if cs.Argc == 0 || cs.Line < symbol.Span.Start || symbol.RawText == "" {
 		return nil
 	}
-	lines := strings.Split(symbol.RawText, "\n")
+	// Masked: `Log("Foo(1, 2)"); Foo(x)` located the call inside the string.
+	lines := maskedSymbolLines(idx, symbol)
 	off := cs.Line - symbol.Span.Start
 	if off < 0 || off >= len(lines) {
 		return nil
@@ -992,16 +988,6 @@ func csRecoverArgs(idx *edgeIndex, symbol *core.SymbolRecord, cs core.CallSite,
 			if depth == 0 {
 				end = k
 			}
-		case '"', '\'':
-			// Skip a literal: overload-relevant literals were already
-			// classified by the extractor, so any call reaching here has
-			// none at top level worth parsing exactly.
-			q := line[k]
-			for k++; k < len(line) && line[k] != q; k++ {
-				if line[k] == '\\' {
-					k++
-				}
-			}
 		}
 		if end >= 0 {
 			break