kevsmith / reddy

Exploring writing a redis client in Erlang

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

What is reddy?

reddy is simply my attempt at writing a better redis client in Erlang. It tries to address shortcomings I saw in other clients like:

  • Too low-level: Yes, you can manually shove binaries over a socket and make it work but that's hardly a satisfying solution. reddy strives to provide a low-friction, self-documenting API. redis' excellent command docs should serve equally well as docs for reddy's API.

  • No support for multiple connections: While reddy is packaged as a proper OTP application it doesn't make any assumptions about how you'd like to use it. Start as many connections as you need.

  • No asynchronous API: reddy combines redis' native pipelining with Erlang's actor concurrency to provide solid support for executing redis operations asynchronously.

Synchronous API

1> application:start(reddy)
ok
2> {ok, Conn} = reddy_conn:connect("127.0.0.1", 6379).
{ok, <0.50.0>}
3> reddy_lists:lpush(Conn, <<"foo">>, <<"1">>).
1
4> reddy_lists:lpush(Conn, <<"foo">>, <<"2">>).
2
5> reddy_lists:lpop(Conn, <<"foo">>).
<<"2">>

Async API

1> application:start(reddy)
ok
2> {ok, Conn} = reddy_conn:connect("127.0.0.1", 6379).
{ok, <0.50.0>}
3> {ok, ResultId} = reddy_lists:lpush_(Conn, <<"bar">>, <<"1">>, true).
{ok, #Ref<0.0.100>}
4> receive {ResultId, Result} -> Result end.
1

Async fire n' forget API

1> application:start(reddy)
ok
2> {ok, Conn} = reddy_conn:connect("127.0.0.1", 6379).
{ok, <0.50.0>}
%% Last function arg indicates whether or not the return value
%% should be sent to the caller. Response is always parsed to prevent
%% memory consumption due to accumulated pipelined responses.
3> ok = reddy_lists:lpush_(Conn, <<"bar">>, <<"1">>, false).
ok
4>

Note: the trailing underscore on the function name indicates it is an async operation.

Connection pools

1> application:start(reddy)
ok
2> reddy_pool:new_pool(production, [{ip, "127.0.0.1"}, {port, 6379}, {count, 10}]).
{ok, <0.51.0>}
3> reddy_lists:lpush(production, <<"foo">>, 1).
1
4> {ok, ResultId} = reddy_lists:lpop_(production, <<"foo">>, true)
{ok,#Ref<0.0.0.159>}
5> receive {ResultId, Results} -> Results end.
<<"1">>
6>

Set hashes as proplists

1> application:start(reddy)
ok
2> {ok, Conn} = reddy_conn:connect("127.0.0.1", 6379).
{ok, <0.50.0>}
3> reddy_hashes:hmset(C, <<"foo">>, [{today, <<"Tuesday">>}, {redis_is, <<"awesome">>}]).
ok
4> reddy_hashes:hvals(C, <<"foo">>).
[<<"Tuesday">>,<<"awesome">>]

TODO

  • Support for ordered sets, pub/sub, transactions, and server commands
  • Connection pool support for keys
  • Pool management & introspection API
  • Keep-alive logic for pooled connections
  • More tests (!!!)

About

Exploring writing a redis client in Erlang


Languages

Language:Erlang 100.0%