README
¶
High Performance Memcached Client
Why this Library?
A Simple Memcached Client Library.
Similar to: https://github.com/bradfitz/gomemcache
But mostly focused on the new meta commands: https://github.com/memcached/memcached/wiki/MetaCommands
It implemented using Batching & Pipelining to reduce syscalls, to increase performance.
Please checkout this document for better understanding of request options and response values of this library: https://github.com/memcached/memcached/blob/master/doc/protocol.txt
Examples
Examples can be found here: https://github.com/QuangTung97/go-memcache/tree/master/examples
pipeline := client.Pipeline()
defer pipeline.Finish()
fn1 := pipeline.MGet("KEY01", memcache.MGetOptions{})
fn2 := pipeline.MGet("KEY02", memcache.MGetOptions{})
getResp, err := fn1()
fmt.Printf("GET: %+v %+v\n", getResp, err)
getResp, err = fn2()
fmt.Printf("GET: %+v %+v\n", getResp, err)
In this example, two meta get commands will be sent only in the call:
getResp, err := fn1()
It will try to batch as much number of commands as possible to the underlining TCP Connections.
But if you do like this:
pipeline := client.Pipeline()
defer pipeline.Finish()
fn1 := pipeline.MGet("KEY01", memcache.MGetOptions{})
getResp, err := fn1()
fmt.Printf("GET: %+v %+v\n", getResp, err)
fn2 := pipeline.MGet("KEY02", memcache.MGetOptions{})
getResp, err = fn2()
fmt.Printf("GET: %+v %+v\n", getResp, err)
Then no batching is possible.
The line defer pipeline.Finish() is the best practise for preventing cases like this:
pipeline := client.Pipeline()
defer pipeline.Finish()
pipeline.MSet("KEY01", []byte("key data 01"), memcache.MSetOptions{})
pipeline.MSet("KEY02", []byte("key data 02"), memcache.MSetOptions{})
without pipeline.Finish() the two set commands will NOT be delivered to the memcached server.
Directories