f2ddaffc31
Remove CoarseTime
...
It is not clear why @valyala introduced this coarse time. Benchmarks on
different systems show that the speedup is no where enough to justify
the added code complexity and bugs it seems to have introduced.
Mac:
BenchmarkCoarseTimeNow-8 2000000000 2.49 ns/op
0 B/op 0 allocs/op
BenchmarkTimeNow-8 500000000 3.14 ns/op
0 B/op 0 allocs/op
Ubuntu:
BenchmarkCoarseTimeNow-4 300000000 6.74 ns/op
0 B/op 0 allocs/op
BenchmarkTimeNow-4 100000000 15.9 ns/op
0 B/op 0 allocs/op
This reverts commit https://github.com/erikdubbelboer/fasthttp/commit/6309f42188ecb28ccf8ac58442739cdb43d75d9e
and https://github.com/erikdubbelboer/fasthttp/commit/32c72cde80f0c591604f825586d6a4bbbb39d9c5 .
See: https://github.com/valyala/fasthttp/issues/271 ,
https://github.com/valyala/fasthttp/pull/269 and
https://github.com/valyala/fasthttp/issues/261 .
2018-08-17 17:15:41 +08:00
f24d00fcc6
A lot of typo fixes
2017-10-08 13:30:35 +01:00
bcdf16d04f
Removed superflouos slashes from documentation
2017-09-12 13:15:04 +03:00
9ffce8c687
client: properly extract tls ServerName from address without port
2017-04-26 17:57:12 +03:00
32c72cde80
Export CoarseTimeNow and clarify that RequestCtx.Time() and RequestCtx.ConnTime() return time truncated to a second
2017-02-10 15:45:14 +02:00
6309f42188
optimization: substitute time.Now() calls with coarse-grained time in hot paths
2017-02-09 23:59:06 +02:00
1c39678a4d
client: retry non-idempotent requests if the server closes idle keep-alive connection before receiving the next request
...
This case usually appears when working with Apache or nginx servers.
See http://httpd.apache.org/docs/2.4/mod/core.html#keepalivetimeout
and http://nginx.org/en/docs/http/ngx_http_core_module.html#keepalive_timeout
2016-11-28 11:50:36 +02:00
a965b01816
Added an example for LBClient
2016-10-21 14:53:44 +03:00
d18f43c48c
Issue #191 : Respect ServerName set in user-provided tls.Config
2016-10-20 15:26:25 +03:00
4ca4d0ea09
Client, HostClient, PipelineClient: verify TLS server name
2016-10-12 19:49:23 +03:00
07f692d02d
Revert "Redirection support for Do(*Request, *Response)"
...
This breaks Get*() with redirects. DoFollowRedirects must be implemented
on top of Do() instead of hacking on internal doRequestFollowRedirects.
This reverts commit 6438022109
2016-10-05 12:44:51 +03:00
0509ba9f93
Revert "removing legacy doRequestFollowRedirects()"
...
This reverts commit 97e154e4cd
2016-10-05 12:44:45 +03:00
97e154e4cd
removing legacy doRequestFollowRedirects()
2016-10-05 12:37:18 +03:00
6438022109
Redirection support for Do(*Request, *Response)
2016-10-05 12:37:18 +03:00
aa4742b5ef
HostClient: make sure only a single connsCleaner is running. Based on the PR #160 by @iizotop
2016-10-05 12:20:17 +03:00
a52a42acd6
Added HostClient.PendingRequests(), which may be used for balancing load among multiple HostClient instances
2016-09-21 23:03:30 +03:00
4f66eb3fbb
Do not busy wait on free connection in DoDeadline and DoTimeout, since this may result in a lot of goroutines stalled in the busy wait for free connections
2016-09-12 14:19:13 +03:00
3e1cb31dcd
Added SwapBody to Request and Response for zero-copy body pass between Request and Response objects
2016-09-12 13:42:59 +03:00
9f7999be97
fix DoDeadline side effect when no free conn ( #145 )
2016-08-17 20:24:33 +03:00
881ac52b00
Added support for multiple concurrent connections to server in PipelineClient
2016-07-16 17:36:32 +03:00
f4ebbdd48b
Removed incorrect statement from PipelineClient.Do docs
2016-07-15 14:41:07 +03:00
d1dd56c016
Issue #131 : document redirects' following for client functions
2016-07-12 10:40:04 +03:00
1fc1f4cbf5
Reduce memory usage under high load by pooling request/response bodies.
...
- Use request/response body pools.
- Defer request/response body allocation until it is really required.
- Return request/response bodies to the pool as soon as they become unused.
This minimizes the total amount of memory occupied by active request/response
bodies under high load.
2016-06-03 16:54:22 +03:00
f36b47782a
removed memory allocations from tryDial. This should improve performance for non-keepalive connections
2016-05-16 17:17:18 +03:00
281be30c26
Client optimization: reduce the number of SetReadDeadline/SetWriteDeadline calls for keep-alive connections, since they don't scale properly. See https://github.com/golang/go/issues/15133 for details.
2016-04-15 23:33:41 +03:00
e9544cbb6f
client: get rid of slow math/rand for sleep jitter generation
2016-04-04 15:02:33 +03:00
3f0e1bb92f
PipelineClient: take into account queued requests waiting to be sent to the server in PendingRequests()
2016-04-03 11:19:52 +03:00
9c292dcd3e
PipelineClient: do not send timed out requests to the server
2016-04-03 11:11:27 +03:00
316adc4782
PipelineClient.Do: try substituting the oldest request in the pending queue by the current one on queue overflow
2016-04-03 10:58:55 +03:00
8ae2d3e53c
PipelineClinet: added MaxBatchDelay option
2016-04-02 23:18:34 +03:00
ca21b21eba
Return ErrPipelineOverflow from PipelineClient.Do if the pending requests' queue is overflown. This should prevent from caller's goroutines leak on stalled pipeline client
2016-04-01 18:21:35 +03:00
22c9594090
Added PipelineClient for issuing pipelined requests to the server
2016-04-01 18:06:59 +03:00
3298603593
client: added MaxIdleConnDuration for limiting the duration of idle keep-alive connections established by clients
2016-03-30 19:23:17 +03:00
124195fdd8
create new default tls.Config instead of re-using existing config, since it may not be copied due to sync.Once member
2016-03-06 19:11:34 +02:00
8280b7a162
Moved empty noCopy struct to the top of container structs. See @stemar94 's comment at https://github.com/golang/go/issues/12884 for details
2016-03-06 00:17:08 +02:00
9f43aa1601
Do not expose noCopy.Lock
2016-03-04 21:20:52 +02:00
9fa69c74af
Embed noCopy struct into structs, which mustn't be copied
...
This should help `go vet` detecting invalid structs' copyings.
See https://github.com/golang/go/issues/8005#issuecomment-190753527 for details.
2016-03-04 16:57:24 +02:00
3546c31c2b
Issue #60 : skip body copying in DoTimeout. This should improve DoTimeout performance when dealing with big request and/or response bodies
2016-03-01 01:13:50 +02:00
e252d37e21
Issue #57 : Added DisableHeaderNamesNormalizing config option to Client and HostClient to be consistent with the corresponding option in Server
2016-02-29 13:11:32 +02:00
7583b9b2ac
client: use exponential backoff algorithm for sleeping between unsuccessful attempts to obtain free connection to the server. This should reduce CPU usage
2016-02-22 19:05:46 +02:00
0ab0d45a50
Eliminated two memory allocations when client follows redirect
2016-02-19 17:27:47 +02:00
9c9b5abffa
Documentation update according to golint
2016-02-12 17:47:57 +02:00
df85922f86
Added (Get|Do)Deadline helper functions to Client and HostClient
2016-02-10 13:35:48 +02:00
a2e6d65e9d
Issue #15 : return more clear error if the server forgets sending 'Connection: close' response header before closing the connection
2016-02-05 13:45:46 +02:00
df213349e2
client: use TLS session cache by default
2016-02-04 20:24:00 +02:00
c450104658
Client: made more clear timeout handling
2016-01-29 13:38:56 +02:00
e3369ec00b
Dial all the resolved TCP addresses in round-robin manner until the connection is established
2016-01-15 19:41:09 +02:00
576ba8868b
Return immediately on ErrDialTimeout when dialing HostClient hosts
2016-01-15 19:12:09 +02:00
a5f4f2e007
Dial all available addresses in HostClient.Addrs before giving up
2016-01-15 19:08:40 +02:00
c78eb83578
Client: re-send idempotent requests after more unsuccessful cases
2016-01-15 18:30:13 +02:00
Erik Dubbelboer
xPushkin
Aliaksandr Valialkin
Sick Yoon
Nicolas Bazire