Skip to content
/ ready Public
generated from treeform/nimtemplate

A Redis client for multi-threaded servers.

License

Notifications You must be signed in to change notification settings

guzba/ready

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

83 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Ready

nimble install ready

API reference

Ready is a Redis client that is built to work well in multi-threaded programs. A great use-case for Ready is in a multi-threaded HTTP server like Mummy.

Check out the examples/ folder for more sample code using Ready.

Using Ready

First you'll need to open a Redis connection. By default Ready connects to the default Redis server at localhost:6379. You can easily specify a different address and port in newRedisConn when needed.

import ready

let redis = newRedisConn() # Defaults to localhost:6379

After opening a connection you can start sending commands. You can send any of Redis's vast set of commands.

import ready, std/options

let redis = newRedisConn() # Defaults to localhost:6379

let value = redis.command("GET", "key").to(Option[string])

We use Option[string] above since the reply may be nil if the key is not present. Alternatively, if you know the key exists, you could just use string.

You can also easily work with replies to more complex commands:

import ready

let redis = newRedisConn() # Defaults to localhost:6379

let values = redis.command("MGET", "key1", "key2", "key3").to(seq[string])

Here we are using MGET to request multiple keys in one command. Since we expect multiple reply entries, we can use to to convert the reply to a seq[string].

Working with replies

A call to command or receive will return a RedisReply object. You'll want to convert that into the types you expect. Ready makes that easy by providing the to proc.

# Basic conversions:

echo reply.to(int)
echo reply.to(string)
echo reply.to(Option[string]) # If the reply can be nil

# Convert array replies to seq:

echo reply.to(seq[int])
echo reply.to(seq[string])
echo reply.to(seq[Option[string]])

# Convert array replies to tuples:

echo reply.to((int, string))
echo reply.to((int, Option[string]))
echo reply.to((string, Option[string], int))

# Mix and match:

echo reply.to((string, Option[string], seq[int]))

# Index access, if you know the reply is an array you can access its elements

echo reply[0].to(string)

A call to reply.to for a type Ready does not know how to convert to will fail at compile time.

If Ready is unable to convert the reply from Redis to your requested type, a RedisError is raised.

Connection pooling

Ready includes a built-in connection pool when compiled with --threads:on:

import ready

let redisPool = newRedisPool(3) # Defaults to localhost:6379

# This automatically removes a connection from the pool, runs the command
# and then returns it back to the pool
redisPool.command("PING")

Or, if you want to run more than one command with the same connection:

import ready

let redisPool = newRedisPool(3) # Defaults to localhost:6379

redisPool.withConnection conn:
    # `conn` is automatically recycled back into the pool after this block
    discard conn.command("PING")

Reusing Redis connections is much faster and more efficient than opening new connections for every command.

Pipelining commands and transactions

Ready also includes separate send and receive calls as an alternative to the command call. These commands make pipelining commands easy:

import ready

let redis = newRedisConn()

redis.send("MULTI")
redis.send("INCR", "mycount")
redis.send("SET", "mykey", "myvalue")
redis.send("EXEC")

## OR:

# redis.send([
#  ("MULTI", @[]),
#  ("INCR", @["mycount"]),
#  ("SET", @["mykey", "myvalue"]),
#  ("EXEC", @[])
#])

# Match the number of `receive` calls to the number of commands sent

discard redis.receive() # OK
discard redis.receive() # QUEUED
discard redis.receive() # QUEUED
let (num, _) = redis.receive().to((int, string))

Pipelining as an advanced technique when using Redis that can drastically increase performance when possible.

Important! Remember to match the number of receive calls to the number of commands sent.

Publish and subscribe (PubSub)

Ready makes it easy to use Redis's PubSub functionality.

Here we dedicate a thread to receiving messages on a PubSub connection while our other thread is free to send commands like SUBSCRIBE and UNSUBSCRIBE to manage the PubSub connection.

let pubsub = newRedisConn() # Defaults to localhost:6379

proc receiveThreadProc() =
  try:
    while true:
      let reply = pubsub.receive()
      echo "Event: ", reply[0].to(string)
      echo "Channel: ", reply[1].to(string)
      echo "Raw: ", reply
  except RedisError as e:
    echo e.msg

var receiveThread: Thread[void]
createThread(receiveThread, receiveThreadProc)

pubsub.send("SUBSCRIBE", "mychannel")

Note that using PubSub with Ready requires threads.

Pro Tips

You can use Ready in two ways, either by calling command or by calling send and receive. Calling command is the equivalent of calling send and then calling receive immediately.

Whenever a command or receive call gets an error reply from Redis a RedisError is raised. This means discarding the reply in discard redis.command("PING") is perfectly ok. If the reply was an error an exception would have been raised.

If you open a short-lived Redis connection, remember to call close when you no longer need it. The connections are not garbage collected. (For HTTP servers this is unlikely, see #1 for a brief discussion.)

Why use send and receive separately? Two reasons:

First, where possible, it is more efficient to pipeline many Redis commands. This is easy to do with Ready, just call send multiple times (or ideally call send with a seq of commands).

Second, you may want to have a separate thread be sending vs receiving. A common use of this is PubSub, where one thread is dedicated to receiving messages and the sending thread manages what channels are subscribed to. See this example.