Smart HTTP fetch: gzip requests
The upload-pack requests are mostly plain text and they compress rather well. Deflating them with Content-Encoding: gzip can easily drop the size of the request by 50%, reducing the amount of data to transfer as we negotiate the common commits. Signed-off-by: Shawn O. Pearce <spearce@spearce.org> CC: Daniel Barkalow <barkalow@iabervon.org> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Shawn O. Pearce committed
Oct 30, 2009 at 17:47 UTC
b8538603a34b13ccb1a3a096f461b8a35d7ea878
1 file changed
+50
remote-curl.c
+50
@@ -289,6 +289,7 @@ struct rpc_state {
289
int in;
290
int out;
291
struct strbuf result;
292
+ unsigned gzip_request : 1;
293
};
294
295
static size_t rpc_out(void *ptr, size_t eltsize,
@@ -327,6 +328,8 @@ static int post_rpc(struct rpc_state *rpc)
328
struct active_request_slot *slot;
329
struct slot_results results;
330
struct curl_slist *headers = NULL;
331
+ int use_gzip = rpc->gzip_request;
332
+ char *gzip_body = NULL;
333
int err = 0, large_request = 0;
334
335
/* Try to load the entire request, if we can fit it into the
@@ -340,6 +343,7 @@ static int post_rpc(struct rpc_state *rpc)
343
344
if (left < LARGE_PACKET_MAX) {
345
large_request = 1;
346
+ use_gzip = 0;
347
break;
348
}
349
@@ -355,6 +359,7 @@ static int post_rpc(struct rpc_state *rpc)
359
curl_easy_setopt(slot->curl, CURLOPT_POST, 1);
360
curl_easy_setopt(slot->curl, CURLOPT_NOBODY, 0);
361
curl_easy_setopt(slot->curl, CURLOPT_URL, rpc->service_url);
362
+ curl_easy_setopt(slot->curl, CURLOPT_ENCODING, "");
363
364
headers = curl_slist_append(headers, rpc->hdr_content_type);
365
headers = curl_slist_append(headers, rpc->hdr_accept);
@@ -372,6 +377,49 @@ static int post_rpc(struct rpc_state *rpc)
377
fflush(stderr);
378
}
379
380
+ } else if (use_gzip && 1024 < rpc->len) {
381
+ /* The client backend isn't giving us compressed data so
382
+ * we can try to deflate it ourselves, this may save on.
383
+ * the transfer time.
384
+ */
385
+ size_t size;
386
+ z_stream stream;
387
+ int ret;
388
+
389
+ memset(&stream, 0, sizeof(stream));
390
+ ret = deflateInit2(&stream, Z_BEST_COMPRESSION,
391
+ Z_DEFLATED, (15 + 16),
392
+ 8, Z_DEFAULT_STRATEGY);
393
+ if (ret != Z_OK)
394
+ die("cannot deflate request; zlib init error %d", ret);
395
+ size = deflateBound(&stream, rpc->len);
396
+ gzip_body = xmalloc(size);
397
+
398
+ stream.next_in = (unsigned char *)rpc->buf;
399
+ stream.avail_in = rpc->len;
400
+ stream.next_out = (unsigned char *)gzip_body;
401
+ stream.avail_out = size;
402
+
403
+ ret = deflate(&stream, Z_FINISH);
404
+ if (ret != Z_STREAM_END)
405
+ die("cannot deflate request; zlib deflate error %d", ret);
406
+
407
+ ret = deflateEnd(&stream);
408
+ if (ret != Z_OK)
409
+ die("cannot deflate request; zlib end error %d", ret);
410
+
411
+ size = stream.total_out;
412
+
413
+ headers = curl_slist_append(headers, "Content-Encoding: gzip");
414
+ curl_easy_setopt(slot->curl, CURLOPT_POSTFIELDS, gzip_body);
415
+ curl_easy_setopt(slot->curl, CURLOPT_POSTFIELDSIZE, size);
416
+
417
+ if (options.verbosity > 1) {
418
+ fprintf(stderr, "POST %s (gzip %lu to %lu bytes)\n",
419
+ rpc->service_name,
420
+ (unsigned long)rpc->len, (unsigned long)size);
421
+ fflush(stderr);
422
+ }
423
} else {
424
/* We know the complete request size in advance, use the
425
* more normal Content-Length approach.
@@ -398,6 +446,7 @@ static int post_rpc(struct rpc_state *rpc)
446
}
447
448
curl_slist_free_all(headers);
449
+ free(gzip_body);
450
return err;
451
}
452
@@ -523,6 +572,7 @@ static int fetch_git(struct discovery *heads,
572
memset(&rpc, 0, sizeof(rpc));
573
rpc.service_name = "git-upload-pack",
574
rpc.argv = argv;
575
+ rpc.gzip_request = 1;
576
577
err = rpc_service(&rpc, heads);
578
if (rpc.result.len)