Use 32-bit offsets in the string pool search tree to reduce its size

CHANGELOG.md

@@ -1,3 +1,7 @@
+## 1.24.1
+
+* Limit the size and depth of the string pool for better performance
+
 ## 1.24.0
 
 * Add feature filters using the Mapbox GL Style Specification filter syntax

memfile.hpp

@@ -6,7 +6,7 @@ struct memfile {
 	char *map;
 	long long len;
 	long long off;
-	long long tree;
+	unsigned long tree;
 };
 
 struct memfile *memfile_open(int fd);

pool.cpp

@@ -35,7 +35,7 @@ int swizzlecmp(const char *a, const char *b) {
 }
 
 long long addpool(struct memfile *poolfile, struct memfile *treefile, const char *s, char type) {
-	long long *sp = &treefile->tree;
+	unsigned long *sp = &treefile->tree;
 	size_t depth = 0;
 
 	// In typical data, traversal depth generally stays under 2.5x
@@ -95,6 +95,16 @@ long long addpool(struct memfile *poolfile, struct memfile *treefile, const char
 		exit(EXIT_FAILURE);
 	}
 
+	if (off >= LONG_MAX || treefile->off >= LONG_MAX) {
+		// Tree or pool is bigger than 2GB
+		static bool warned = false;
+		if (!warned) {
+			fprintf(stderr, "Warning: string pool is very large.\n");
+			warned = true;
+		}
+		return off;
+	}
+
 	struct stringpool tsp;
 	tsp.left = 0;
 	tsp.right = 0;

pool.hpp

@@ -2,9 +2,9 @@
 #define POOL_HPP
 
 struct stringpool {
-	long long left;
+	unsigned long left;
-	long long right;
+	unsigned long right;
-	long long off;
+	unsigned long off;
 };
 
 long long addpool(struct memfile *poolfile, struct memfile *treefile, const char *s, char type);

version.hpp

@@ -1,6 +1,6 @@
 #ifndef VERSION_HPP
 #define VERSION_HPP
 
-#define VERSION "tippecanoe v1.24.0\n"
+#define VERSION "tippecanoe v1.24.1\n"
 
 #endif