Optimize some string handling in chunk parser

include/lighttpd/chunk_parser.h

@@ -22,7 +22,7 @@ struct liChunkParserCtx {
 
 struct liChunkParserMark {
 	liChunkIter ci;
-	off_t pos;
+	off_t pos, abs_pos;
 };
 
 LI_API void li_chunk_parser_init(liChunkParserCtx *ctx, liChunkQueue *cq);
@@ -31,6 +31,7 @@ LI_API liHandlerResult li_chunk_parser_prepare(liChunkParserCtx *ctx);
 LI_API liHandlerResult li_chunk_parser_next(liVRequest *vr, liChunkParserCtx *ctx, char **p, char **pe);
 LI_API void li_chunk_parser_done(liChunkParserCtx *ctx, goffset len);
 
+/* extract [from..to) */
 LI_API gboolean li_chunk_extract_to(liVRequest *vr, liChunkParserMark from, liChunkParserMark to, GString *dest);
 LI_API GString* li_chunk_extract(liVRequest *vr, liChunkParserMark from, liChunkParserMark to);
 
@@ -44,6 +45,7 @@ INLINE liChunkParserMark li_chunk_parser_getmark(liChunkParserCtx *ctx, const ch
 	liChunkParserMark m;
 	m.ci = ctx->curi;
 	m.pos = ctx->start + fpc - ctx->buf;
+	m.abs_pos = ctx->bytes_in + fpc - ctx->buf;
 	return m;
 }
 

src/main/chunk_parser.c

@@ -55,7 +55,8 @@ void li_chunk_parser_done(liChunkParserCtx *ctx, goffset len) {
 gboolean li_chunk_extract_to(liVRequest *vr, liChunkParserMark from, liChunkParserMark to, GString *dest) {
 	liChunkParserMark i;
 
-	g_string_set_size(dest, 0);
+	g_string_set_size(dest, to.abs_pos - from.abs_pos);
+	li_g_string_clear(dest);
 
 	for ( i = from; i.ci.element != to.ci.element; li_chunkiter_next(&i.ci) ) {
 		goffset len = li_chunkiter_length(i.ci);
@@ -63,7 +64,14 @@ gboolean li_chunk_extract_to(liVRequest *vr, liChunkParserMark from, liChunkPars
 			char *buf;
 			off_t we_have;
 			if (LI_HANDLER_GO_ON != li_chunkiter_read(vr, i.ci, i.pos, len - i.pos, &buf, &we_have)) goto error;
-			g_string_append_len(dest, buf, we_have);
+			if (dest->len + we_have < dest->allocated_len) {
+				/* "fast" append */
+				memcpy(dest->str + dest->len, buf, we_have);
+				dest->len += we_have;
+				dest->str[dest->len] = '\0';
+			} else {
+				g_string_append_len(dest, buf, we_have);
+			}
 			i.pos += we_have;
 		}
 		i.pos = 0;
@@ -72,14 +80,21 @@ gboolean li_chunk_extract_to(liVRequest *vr, liChunkParserMark from, liChunkPars
 		char *buf;
 		off_t we_have;
 		if (LI_HANDLER_GO_ON != li_chunkiter_read(vr, i.ci, i.pos, to.pos - i.pos, &buf, &we_have)) goto error;
-		g_string_append_len(dest, buf, we_have);
+		if (dest->len + we_have < dest->allocated_len) {
+			/* "fast" append */
+			memcpy(dest->str + dest->len, buf, we_have);
+			dest->len += we_have;
+			dest->str[dest->len] = '\0';
+		} else {
+			g_string_append_len(dest, buf, we_have);
+		}
 		i.pos += we_have;
 	}
 
 	return TRUE;
 
 error:
-	g_string_assign(dest, "");
+	li_g_string_clear(dest);
 	return FALSE;
 }
 

src/main/http_request_parser.rl

@@ -24,15 +24,17 @@
 
 	action header_key {
 		getStringTo(fpc, ctx->h_key);
-		g_string_truncate(ctx->h_value, 0);
+		li_g_string_clear(ctx->h_value);
 	}
 	action header_value {
 		getStringTo(fpc, ctx->h_value);
 		/* Remove CRLF */
 		if (ctx->h_value->len > 2) {
-			g_string_truncate(ctx->h_value, ctx->h_value->len - 2);
+			ctx->h_value->len -= 2;
+			ctx->h_value->str[ctx->h_value->len] = '\0';
+			/* g_string_truncate(ctx->h_value, ctx->h_value->len - 2); */
 		} else {
-			g_string_truncate(ctx->h_value, 0);
+			li_g_string_clear(ctx->h_value);
 		}
 	}
 	action header {