git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[RFC PATCH v2 16/16] Smart HTTP fetch: gzip requests

From
Shawn O. Pearce <spearce@spearce.org>
Date
Oct 13, 2009, 02:25 UTC
Message-ID
<1255400715-10508-17-git-send-email-spearce@spearce.org>
In-Reply-To
<1255400715-10508-1-git-send-email-spearce@spearce.org>

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>
---
 remote-curl.c |   52 ++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 files changed, 50 insertions(+), 2 deletions(-)
diff --git a/remote-curl.c b/remote-curl.c
index 31d1d34..d53215d 100644
--- a/remote-curl.c
+++ b/remote-curl.c
@@ -261,11 +261,12 @@ static size_t rpc_in(const void *ptr, size_t eltsize,
 	return size;
 }
 
-static int post_rpc(struct rpc_state *state)
+static int post_rpc(struct rpc_state *state, int use_gzip)
 {
 	struct active_request_slot *slot;
 	struct slot_results results;
 	struct curl_slist *headers = NULL;
+	unsigned char *gzip_body = NULL;
 	int err = 0, large_request = 0;
 
 	/* Try to load the entire request, if we can fit it into the
@@ -279,6 +280,7 @@ static int post_rpc(struct rpc_state *state)
 
 		if (left < LARGE_PACKET_MAX) {
 			large_request = 1;
+			use_gzip = 0;
 			break;
 		}
 
@@ -311,6 +313,48 @@ static int post_rpc(struct rpc_state *state)
 			fflush(stderr);
 		}
 
+	} else if (use_gzip && 1024 < state->len) {
+		/* The client backend isn't giving us compressed data so
+		 * we can try to deflate it ourselves, this may save on.
+		 * the transfer time.
+		 */
+		size_t size;
+		z_stream stream;
+		int ret;
+
+		memset(&stream, 0, sizeof(stream));
+		ret = deflateInit2(&stream, Z_DEFAULT_COMPRESSION,
+				Z_DEFLATED, (15 + 16),
+				8, Z_DEFAULT_STRATEGY);
+		if (ret != Z_OK)
+			die("cannot deflate request; zlib init error %d", ret);
+		size = deflateBound(&stream, state->len);
+		gzip_body = xmalloc(size);
+
+		stream.next_in = (unsigned char *)state->buf;
+		stream.avail_in = state->len;
+		stream.next_out = gzip_body;
+		stream.avail_out = size;
+
+		ret = deflate(&stream, Z_FINISH);
+		if (ret != Z_STREAM_END)
+			die("cannot deflate request; zlib deflate error %d", ret);
+
+		ret = deflateEnd(&stream);
+		if (ret != Z_OK)
+			die("cannot deflate request; zlib end error %d", ret);
+
+		size = stream.total_out;
+
+		headers = curl_slist_append(headers, "Content-Encoding: gzip");
+		curl_easy_setopt(slot->curl, CURLOPT_POSTFIELDS, gzip_body);
+		curl_easy_setopt(slot->curl, CURLOPT_POSTFIELDSIZE, size);
+
+		if (state->verbose) {
+			fprintf(stderr, "POST %s (gzip %lu to %lu bytes)\n",
+				state->service_name, state->len, size);
+			fflush(stderr);
+		}
 	} else {
 		/* We know the complete request size in advance, use the
 		 * more normal Content-Length approach.
@@ -336,11 +380,13 @@ static int post_rpc(struct rpc_state *state)
 		}
 	}
 	curl_slist_free_all(headers);
+	free(gzip_body);
 	return err;
 }
 
 static int one_shot_rpc_service(const char *service,
 	int verbose,
+	int use_gzip,
 	const char **client_argv,
 	struct discovery *heads,
 	struct strbuf *result)
@@ -384,7 +430,7 @@ static int one_shot_rpc_service(const char *service,
 			break;
 		state.pos = 0;
 		state.len = n;
-		err |= post_rpc(&state);
+		err |= post_rpc(&state, use_gzip);
 	}
 	if (result)
 		strbuf_read(result, client.out, 0);
@@ -474,6 +520,7 @@ static int fetch_git(struct fetch_args *args,
 
 	err = one_shot_rpc_service("git-upload-pack",
 		args->verbose,
+		1 /* gzip request */,
 		argv, heads, &res);
 	if (res.len)
 		safe_write(1, res.buf, res.len);
@@ -611,6 +658,7 @@ static int push_git(struct discovery *heads,
 
 	err = one_shot_rpc_service("git-receive-pack",
 		args->verbose,
+		0 /* no gzip */,
 		argv, heads, &res);
 	if (res.len)
 		safe_write(1, res.buf, res.len);
-- 
1.6.5.52.g0ff2e
Previous: Shawn O. PearceNext: Junio C Hamano
Message 35 of 38 in “Return of smart HTTP”
  1. 00/16 Return of smart HTTPShawn O. Pearce, Oct 13, 2009
  2. 01/16 pkt-line: Add strbuf based functionsShawn O. Pearce, Oct 13, 2009
  3. Johannes SixtOct 13, 2009
  4. Shawn O. PearceOct 13, 2009
  5. 02/16 pkt-line: Make packet_read_line easier to debugShawn O. Pearce, Oct 13, 2009
  6. 03/16 fetch-pack: Use a strbuf to compose the want listShawn O. Pearce, Oct 13, 2009
  7. 04/16 Move "get_ack()" back to fetch-packShawn O. Pearce, Oct 13, 2009
  8. 05/16 Add multi_ack_2 capability to fetch-pack/upload-packShawn O. Pearce, Oct 13, 2009
  9. Jakub NarebskiOct 13, 2009
  10. Shawn O. PearceOct 13, 2009
  11. 06/16 remote-curl: Refactor walker initializationShawn O. Pearce, Oct 13, 2009
  12. 07/16 remote-helpers: Fetch more than one ref in a batchShawn O. Pearce, Oct 13, 2009
  13. Daniel BarkalowOct 13, 2009
  14. Shawn O. PearceOct 13, 2009
  15. 08/16 remote-helpers: Support custom transport optionsShawn O. Pearce, Oct 13, 2009
  16. Daniel BarkalowOct 13, 2009
  17. Shawn O. PearceOct 13, 2009
  18. Daniel BarkalowOct 13, 2009
  19. Shawn O. PearceOct 13, 2009
  20. Daniel BarkalowOct 13, 2009
  21. Shawn O. PearceOct 13, 2009
  22. 09/16 Move WebDAV HTTP push under remote-curlShawn O. Pearce, Oct 13, 2009
  23. Mike HommeyOct 13, 2009
  24. Johannes SchindelinOct 13, 2009
  25. 10/16 Git-aware CGI to provide dumb HTTP transportShawn O. Pearce, Oct 13, 2009
  26. Johannes SixtOct 13, 2009
  27. 11/16 Add one shot RPC options to upload-pack, receive-packShawn O. Pearce, Oct 13, 2009
  28. 12/16 Smart fetch and push over HTTP: server sideShawn O. Pearce, Oct 13, 2009
  29. Johannes SixtOct 13, 2009
  30. Shawn O. PearceOct 13, 2009
  31. 13/16 Discover refs via smart HTTP server when availableShawn O. Pearce, Oct 13, 2009
  32. 14/16 Smart push over HTTP: client sideShawn O. Pearce, Oct 13, 2009
  33. Felipe ContrerasOct 13, 2009
  34. 15/16 Smart fetch over HTTP: client sideShawn O. Pearce, Oct 13, 2009
  35. 16/16 Smart HTTP fetch: gzip requestsShawn O. Pearce, Oct 13, 2009
  36. Junio C HamanoOct 13, 2009
  37. eduard stefanOct 13, 2009
  38. Junio C HamanoOct 13, 2009

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.