Improve route harvesting (open-telemetry#635)

Author: grcevski

pkg/components/transform/route/harvest/java.go

@@ -5,6 +5,8 @@ package harvest
 
 import (
 	"bufio"
+	"bytes"
+	"io"
 	"log/slog"
 	"net/url"
 	"regexp"
@@ -116,23 +118,44 @@ func (h *JavaRoutes) addRouteIfValid(line string, routes []string) []string {
 	return routes
 }
 
+func (h *JavaRoutes) processSymbolLine(lineBytes []byte, routes []string) []string {
+	if len(lineBytes) > 0 {
+		// Remove newline characters
+		lineBytes = bytes.TrimRight(lineBytes, "\r\n")
+		// Validate the line and sanitize
+		line, ok := h.validLine(string(lineBytes))
+		if ok {
+			routes = h.addRouteIfValid(line, routes)
+		}
+	}
+
+	return routes
+}
+
 func (h *JavaRoutes) ExtractRoutes(pid int32) (*RouteHarvesterResult, error) {
 	routes := []string{}
 	out, err := jvmAttachFunc(int(pid), []string{"jcmd", "VM.symboltable -verbose"}, h.log)
 	if err != nil {
 		return nil, err
 	}
 
-	scanner := bufio.NewScanner(out)
-	for scanner.Scan() {
-		line := scanner.Text()
-		line, ok := h.validLine(line)
+	defer out.Close()
 
-		if !ok {
-			continue
+	reader := bufio.NewReader(out)
+	for {
+		// Read line by line, handling arbitrarily long lines
+		line, err := reader.ReadBytes('\n')
+		if err != nil {
+			if err == io.EOF {
+				// Process the last line if it doesn't end with newline
+				routes = h.processSymbolLine(line, routes)
+				break
+			}
+			h.log.Error("error reading line", "error", err)
+			return nil, err
 		}
 
-		routes = h.addRouteIfValid(line, routes)
+		routes = h.processSymbolLine(line, routes)
 	}
 
 	routes = h.sortRoutes(routes)

pkg/components/transform/route/part_matcher.go

@@ -43,27 +43,37 @@ func (rm *PartialRouteMatcher) Find(path string) string {
 	}
 
 	tokens := tokenize(path)
-	return rm.findCombined(tokens, 0, []string{})
+	return rm.findCombined(tokens, 0, make([]string, len(tokens)), 0)
 }
 
 // findCombined tries to match the full path using combinations of partial matches from different roots
-func (rm *PartialRouteMatcher) findCombined(tokens []string, startIdx int, matchedParts []string) string {
+func (rm *PartialRouteMatcher) findCombined(tokens []string, startIdx int, matchedParts []string, matchedLen int) string {
 	// If we've consumed all tokens, we found a complete match
 	if startIdx >= len(tokens) {
-		if len(matchedParts) > 0 {
-			return strings.Join(matchedParts, "")
+		if matchedLen > 0 {
+			return strings.Join(matchedParts[:matchedLen], "")
 		}
 		return ""
 	}
 
+	newMatchedParts := matchedParts
+
+	// This check is here for ensuring that we don't run into unexpected condition where
+	// the original tokens slice is shorter than the parts slice. In practice, this should
+	// never happen, the tokens will always be >= to the number of parts. However, if we
+	// encounter unexpected pattern and we don't expand the array we'll hit a panic.
+	if matchedLen == len(matchedParts) {
+		newMatchedParts = make([]string, len(matchedParts)+1)
+		copy(newMatchedParts, matchedParts)
+	}
+
 	// Try each root tree for partial matching from current position
 	for _, root := range rm.roots {
 		if partialMatch, consumed := rm.findPartial(tokens[startIdx:], root); partialMatch != "" && consumed > 0 {
 			// Found a partial match, try to match the rest
-			newMatchedParts := make([]string, len(matchedParts)+1)
-			copy(newMatchedParts, matchedParts)
-			newMatchedParts[len(matchedParts)] = partialMatch
-			if result := rm.findCombined(tokens, startIdx+consumed, newMatchedParts); result != "" {
+			newMatchedParts[matchedLen] = partialMatch
+
+			if result := rm.findCombined(tokens, startIdx+consumed, newMatchedParts, matchedLen+1); result != "" {
 				return result
 			}
 		}

pkg/components/transform/route/part_matcher_test.go

@@ -391,3 +391,32 @@ func BenchmarkPartialRouteMatcherComplex(b *testing.B) {
 		}
 	}
 }
+
+// TestMatchedPartsSliceExpansion tests the specific condition where the matchedParts slice
+// needs to be expanded when matchedLen == len(matchedParts). This happens when we have
+// enough partial route matches to fill the initial slice and need to add more.
+func TestMatchedPartsSliceExpansion(t *testing.T) {
+	// Create routes that can be combined into a long path requiring slice expansion
+	routes := []string{
+		"/api",
+		"/v1",
+		"/users",
+		"/{id}",
+		"/profile",
+		"/settings",
+		"/preferences",
+		"/notifications",
+	}
+
+	m := NewPartialRouteMatcher(routes)
+
+	// Create a path that will match multiple partial routes sequentially,
+	// forcing the matchedParts slice to expand beyond its initial size
+	testPath := "/api/v1/users/123/profile/settings/preferences/notifications"
+
+	// Purposefully call with empty parts slice to force the iteration to hit the copy
+	tokens := tokenize(testPath)
+	result := m.findCombined(tokens, 0, make([]string, 0), 0)
+
+	assert.Equal(t, "/api/v1/users/{id}/profile/settings/preferences/notifications", result)
+}