internal: add bounded path registry to prevent cardinality explosion

"cardinality explosion" would make for an epic rock band name...

Signed-off-by: NotAShelf <raf@notashelf.dev>
Change-Id: I53cceb00ab9b17039b1fb1389977bf6b6a6a6964

internal/aggregate/registry.go

@@ -0,0 +1,72 @@
+package aggregate
+
+import (
+	"sync"
+)
+
+// Maintain a bounded set of unique request paths. This prevents metric cardinality explosion by rejecting new paths
+// once the configured limit is reached.
+type PathRegistry struct {
+	mu            sync.RWMutex
+	paths         map[string]struct{}
+	maxPaths      int
+	overflowCount int
+}
+
+// Creates a new PathRegistry with the specified maximum number of unique paths.
+// Once this limit is reached, subsequent Add() calls for new paths will be rejected.
+func NewPathRegistry(maxPaths int) *PathRegistry {
+	return &PathRegistry{
+		paths:    make(map[string]struct{}, maxPaths),
+		maxPaths: maxPaths,
+	}
+}
+
+// Add attempts to add a path to the registry.
+// Returns true if the path was accepted: either already existed or was added,
+// false if rejected due to reaching the limit.
+func (r *PathRegistry) Add(path string) bool {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+
+	// If path already exists, accept it
+	if _, exists := r.paths[path]; exists {
+		return true
+	}
+
+	// If we haven't reached the limit, add the path
+	if len(r.paths) < r.maxPaths {
+		r.paths[path] = struct{}{}
+		return true
+	}
+
+	// Limit reached - reject and increment overflow
+	r.overflowCount++
+	return false
+}
+
+// Contains checks if a path exists in the registry.
+func (r *PathRegistry) Contains(path string) bool {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	_, exists := r.paths[path]
+	return exists
+}
+
+// Count returns the number of unique paths in the registry.
+func (r *PathRegistry) Count() int {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	return len(r.paths)
+}
+
+// Returns the number of paths that were rejected
+// due to the registry being at capacity.
+func (r *PathRegistry) OverflowCount() int {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	return r.overflowCount
+}

internal/aggregate/registry_test.go

@@ -0,0 +1,98 @@
+package aggregate
+
+import (
+	"testing"
+)
+
+func TestPathRegistry_Add(t *testing.T) {
+	registry := NewPathRegistry(3)
+
+	// Add paths within limit
+	if !registry.Add("/api/users") {
+		t.Error("Expected first path to be accepted")
+	}
+	if !registry.Add("/api/posts") {
+		t.Error("Expected second path to be accepted")
+	}
+	if !registry.Add("/api/comments") {
+		t.Error("Expected third path to be accepted")
+	}
+
+	// Add duplicate path - should succeed
+	if !registry.Add("/api/users") {
+		t.Error("Expected duplicate path to be accepted")
+	}
+
+	// Verify count is still 3
+	if count := registry.Count(); count != 3 {
+		t.Errorf("Expected count 3, got %d", count)
+	}
+
+	// Exceed limit
+	if registry.Add("/api/photos") {
+		t.Error("Expected fourth unique path to be rejected")
+	}
+
+	// Verify overflow count
+	if overflow := registry.OverflowCount(); overflow != 1 {
+		t.Errorf("Expected overflow count 1, got %d", overflow)
+	}
+
+	// Add another path beyond limit
+	if registry.Add("/api/videos") {
+		t.Error("Expected fifth unique path to be rejected")
+	}
+
+	// Verify overflow count incremented
+	if overflow := registry.OverflowCount(); overflow != 2 {
+		t.Errorf("Expected overflow count 2, got %d", overflow)
+	}
+
+	// Verify count is still 3
+	if count := registry.Count(); count != 3 {
+		t.Errorf("Expected count 3, got %d", count)
+	}
+}
+
+func TestPathRegistry_Contains(t *testing.T) {
+	registry := NewPathRegistry(3)
+
+	registry.Add("/api/users")
+	registry.Add("/api/posts")
+
+	if !registry.Contains("/api/users") {
+		t.Error("Expected /api/users to be in registry")
+	}
+
+	if !registry.Contains("/api/posts") {
+		t.Error("Expected /api/posts to be in registry")
+	}
+
+	if registry.Contains("/api/comments") {
+		t.Error("Expected /api/comments to NOT be in registry")
+	}
+}
+
+func TestPathRegistry_Count(t *testing.T) {
+	registry := NewPathRegistry(5)
+
+	if count := registry.Count(); count != 0 {
+		t.Errorf("Expected initial count 0, got %d", count)
+	}
+
+	registry.Add("/api/users")
+	if count := registry.Count(); count != 1 {
+		t.Errorf("Expected count 1, got %d", count)
+	}
+
+	registry.Add("/api/posts")
+	if count := registry.Count(); count != 2 {
+		t.Errorf("Expected count 2, got %d", count)
+	}
+
+	// Add duplicate - count should not change
+	registry.Add("/api/users")
+	if count := registry.Count(); count != 2 {
+		t.Errorf("Expected count 2, got %d", count)
+	}
+}

testdata/config.valid.yaml

@@ -19,6 +19,7 @@ site:
 limits:
   max_paths: 1000
   max_events_per_minute: 10000
+  max_sources: 100
 
 server:
   listen_addr: :8080