NAME
Riak::Client - Fast and lightweight Perl client for Riak
VERSION
version 1.94
SYNOPSIS
use Riak::Client;
# normal mode
my $client = Riak::Client->new(
host => '127.0.0.1',
port => 8087,
r => 2,
w => 2,
dw => 1,
connection_timeout => 5,
read_timeout => 5,
write_timeout => 5,
no_auto_connect => 0,
);
# AnyEvent mode
my $client = Riak::Client->new(
host => '127.0.0.1',
port => 8087
anyevent_mode => 1,
);
$client->is_alive() or die "riak is not alive";
# store hashref. will be serialized as JSON
$client->put( 'bucket_name', 'key_name', { some => 'structure' } );
# store text
$client->put( 'bucket_name', 'key_name', 'sometext', 'text/plain' );
# store raw data
$client->put_raw( 'bucket_name', 'key_name', 'rawdata' );
# fetch hashref
my $hash = $client->get( 'bucket_name', 'key_name' );
# fetch raw data
my $text = $client->get_raw( 'bucket_name', 'key_name');
# delete data
$client->del( 'bucket_name', 'key_name');
# AnyEvent mode
my $cv = AE::cv
$client->get_raw( 'bucket_name', 'key_name'
sub { do_something_with($_[0]);
$cv->send();
}
);
# ... later one
$cv->recv();
# list keys in stream
$client->get_keys(foo => sub{
my ($key, $done) = @_;
# you should use another client inside this callback!
$another_client->del(foo => $key);
});
DESCRIPTION
Riak::Client is a fast and light Perl client for Riak using PBC interface, with optional AnyEvent mode.
It supports operations like ping, get, exists, put, del, secondary indexes (so-called 2i) setting and querying, and Map Reduce querying.
It has two modes, a traditional procedural mode, and an event based mode, using AnyEvent.
It started as a fork of Riak::Light to fix some bugs, but actually ended up in a complete rewrite with more features, but the same performance.
ATTRIBUTES
anyevent_mode
Enables the AnyEvent mode, allowing true asynchronous mode.
host
Str, Required. Riak IP or hostname.
port
Int, Required. Port of the PBC interface.
r
Int, Default 2. R value setting for this client.
w
Int, Default 2. W value setting for this client.
dw
Int, Default 1. DW value setting for this client.
connection_timeout
Float, Default 5. Timeout for connection operation, in seconds. Set to 0 for no timeout.
read_timeout
Float, Default 5. Timeout for read operation, in seconds. Set to 0 for no timeout.
no_delay
Boolean, Default 0. If set to a true value, TCP_NODELAY will be enabled on the socket, which means deactivating Nagle's algorithm. Use only if you know what you're doing.
no_auto_connect
Bool, Default 0. If set to true, then the module won't automatically connect upon instanciation. Instead, you'll have to call connect() yourself.
anyevent_mode
Bool, Default 0. If set to true, then all methods can receive a callback, as last argument. If present, the method will return immediately, and the callback will be executed upon completion of the operation, receiving a condvar as first and only argument. If set to false (the default), then the client instance will be synchronous.
METHODS
connect
$client->connect();
$client->connect($coderef);
Connects to the Riak server. On error, will raise an exception. This is automatically done when new() is called, unless the no_auto_connect attribute is set to true. Accepts an optional callback, that will be executed when connected.
# example in AnyEvent mode
$cv = AE::cv;
$client->connect(sub { print "connected!\n"; $cv->send(); });
# ...
$cv->recv();
ping
my $result = $client->ping();
$client->ping($coderef);
Performs a ping operation. On error, will raise an exception. Accepts an optional callback, that will be executed upon completion
# example in AnyEvent mode
$cv = AE::cv;
$client->ping(sub { print "got $_[0] \n"; $cv->send(); });
# ...
$cv->recv();
# an other example
use Try::Tiny;
try { $client->ping() } catch { "oops... something is wrong: $_" };
See also is_alive().
is_alive
my $is_alive = $client->is_alive();
$client->is_alive($coderef);
Checks if the connection is alive. Returns true or false. On error, will raise an exception. Accepts an optional callback, that will be executed upon completion. Even in AnyEvent mode, this operation is synchronous.
# example in AnyEvent mode
$cv = AE::cv;
$client->is_alive(sub { print($_[0] ? "alive\n" : "dead\n"); $cv->send(); });
# ...
$cv->recv();
get
my $value = $client->get($bucket, $key);
$client->get($bucket, $key, $coderef);
# example in AnyEvent mode
$cv = AE::cv;
$client->get('bucket', 'key', sub { do_stuff_with_value($_[0]); $cv->send() });
# ...
$cv->recv();
Performs a fetch operation. Expects bucket and key names. Returns the value. On error, will raise an exception. Accepts an optional callback, that will be called upon completion, with the value as first argument. If the content_type of the fetched value is 'application/json', automatically decodes the JSON into a Perl structure. If you need the raw data you can use get_raw.
get_raw
my $value = $client->get_raw($bucket, $key);
$client->get_raw($bucket, $key, $coderef);
Same as get, but no automatic JSON decoding will be performed. If you want JSON to be automatically decoded, you should use get() instead.
put
$client->put($bucket, $key, $value);
$client->put($bucket, $key, $value, $coderef);
$client->put($bucket, $key, $value, $mime_type, $coderef);
$client->put($bucket, $key, $value, $mime_type, $secondary_indexes, $coderef);
$client->put($bucket, $key, $value, $mime_type, $secondary_indexes, $links, $coderef);
Performs a store operation. Expects bucket and key names, the value, the content type (optional, default is 'application/json'), the indexes to set for this value (optional, default is none), the links to set for this value (optional, default is none), and an optional coderef. On error, will raise an exception
Will encode the structure in json string if necessary. If you need to store the raw data you should use put_raw instead.
IMPORTANT: all the index field names should end by either _int or _bin, depending if the index type is integer or binary.
To query secondary indexes, see query_index.
$client->put('bucket', 'key', { some_values => [1,2,3] });
$client->put('bucket', 'key', { some_values => [1,2,3] }, 'application/json);
$client->put('bucket', 'key', 'text', 'plain/text');
# you can set secondary indexes (2i)
$client->put( 'bucket', 'key', 'text_value', 'plain/text',
{ field1_bin => 'abc', field2_int => 42 }
);
$client->put( 'bucket', 'key', { some_values => [1,2,3] }, undef,
{ field1_bin => 'abc', field2_int => 42 }
);
# you can also set links
$client->put( 'bucket', 'key', 'text', 'plain/text', undef,
{ link_tag1 => 'bucket/key',
link_tag2 => 'other_bucket/key',
}
);
# you can set multiple links for the same tag
$client->put( 'bucket', 'key', 'text', 'plain/text', undef,
{ link_tag1 => [ qw( bucket/key bucket2/key2 ) ],
link_tag2 => 'other_bucket/key',
}
);
# you can also use this form (marginally faster)
$client->put( 'bucket', 'key', 'text', 'plain/text', undef,
[ { tag => 'link_tag1', bucket => 'bucket1', key => 'key1'},
{ tag => 'link_tag2', bucket => 'bucket2', key => 'key2'},
],
);
# example in AnyEvent mode
$cv = AE::cv;
$client->put( 'bucket', 'key', 'some_text', 'plain/text',
{ field1_bin => 'abc', field2_int => 42 },
{ next_key => 'bucket2/foo'},
sub { print "data is sent to Riak\n"; $cv->send() },
);
# ...
$cv->recv();
put_raw
$client->put_raw('bucket', 'key', encode_json({ some_values => [1,2,3] }), 'application/json');
$client->put_raw('bucket', 'key', 'text');
$client->put_raw('bucket', 'key', 'text', undef, {field_bin => 'foo'});
$client->put_raw('bucket', 'key', 'text', undef, {field_bin => 'foo'}, $links);
For more example, see put.
Perform a store operation. Expects bucket and key names, the value, the content type (optional, default is 'plain/text'), the indexes (optional, default is none), and links (optional, default is none) to set for this value
This method won't encode the data, but pass it as such, trusting it's in the type you've indicated in the passed content-type. If you want the structure to be automatically encoded, use put instead.
IMPORTANT: all the index field names should end by either _int or _bin, depending if the index type is integer or binary.
To query secondary indexes, see query_index.
del
$client->del(bucket => key);
Perform a delete operation. Expects bucket and key names.
get_keys
# in default mode
$client->get_keys(foo => sub{
my ($key, $done) = @_;
# you should use another client inside this callback!
$another_client->del(foo => $key);
});
# in anyevent mode
my $cv = AE::cv;
$client->get_keys(foo => sub{
my ($key, $done) = @_;
# ... do stuff with $key
$done and $cv->send;
});
$cv->recv();
WARNING, this method should not be called on a production Riak cluster, as it can have a big performance impact. See Riak's documentation.
WARNING, because Riak doesn't handles pipelining, you cannot use the same Riak::Client instance inside the callback, it would raise an exception.
Perform a list keys operation. Receive a callback and will call it for each key. The callback will receive two arguments: the key, and a boolean indicating if it's the last key
The callback is optional, in which case an ArrayRef of all the keys are returned. But don't do that, and always provide a callback, to avoid your RAM usage to skyrocket...
exists
$client->exists(bucket => 'key') or warn "key not found";
Perform a fetch operation but with head => 0, and the if there is something stored in the bucket/key.
query_index
Perform a secondary index (2i) query. Expects a bucket name, the index field name, the index value you're searching on, and optionally a callback.
If a callback has been provided, doesn't return anything, but execute the callback on each matching keys. callback will receive the key name as first argument. key name will also be in $_. If no callback is provided, returns and ArrayRef of matching keys.
The index value you're searching on can be of two types. If it's a Scalar, an exact match query will be performed. if the value is an ArrayRef, then a range query will be performed, the first element in the array will be the range_min, the second element the range_max. other elements will be ignored.
Based on the example in put, here is how to query it:
# exact match
my $matching_keys = $client->query_index( 'bucket', 'field2_int', 42 ),
# range match
my $matching_keys = $client->query_index( 'bucket', 'field2_int', [ 40, 50] ),
# range match with callback
$client->query_index( 'bucket', 'field2_int', [ 40, 50], sub { print "key : $_" } ),
get_buckets
WARNING, this method should not be called on a production Riak cluster, as it can have a big performance impact. See Riak's documentation.
get_bucket_props
set_bucket_props
map_reduce
map_reduce_raw
BENCHMARKS
Note: These benchmarks are the one provided by Riak::Light.
Note: the AnyEvent mode is a bit slower below, because we are forcing synchronous mode, even in AnyEvent, so the benchmark is paying the price of having AnyEvent enabled but not used.
GETS
Rate Data::Riak (REST) Riak::Tiny (REST) Net::Riak (REST) Data::Riak::Fast (REST) Net::Riak (PBC) Riak::Client (PBC + AnyEvent) Riak::Light (PBC) Riak::Client (PBC)
Data::Riak (REST) 427/s -- -30% -31% -43% -65% -85% -90% -91%
Riak::Tiny (REST) 611/s 43% -- -2% -19% -51% -79% -86% -87%
Net::Riak (REST) 623/s 46% 2% -- -17% -50% -78% -86% -87%
Data::Riak::Fast (REST) 755/s 77% 24% 21% -- -39% -74% -83% -84%
Net::Riak (PBC) 1238/s 190% 103% 99% 64% -- -57% -72% -74%
Riak::Client (PBC + AnyEvent) 2878/s 573% 371% 362% 281% 132% -- -34% -39%
Riak::Light (PBC) 4348/s 917% 612% 598% 476% 251% 51% -- -8%
Riak::Client (PBC) 4706/s 1001% 671% 655% 524% 280% 64% 8% --
PUTS
Rate Net::Riak (REST) Data::Riak (REST) Riak::Tiny (REST) Data::Riak::Fast (REST) Net::Riak (PBC) Riak::Light (PBC) Riak::Client (PBC + AnyEvent) Riak::Client (PBC)
Net::Riak (REST) 542/s -- -15% -29% -55% -57% -90% -92% -92%
Data::Riak (REST) 635/s 17% -- -17% -47% -49% -89% -90% -90%
Riak::Tiny (REST) 765/s 41% 20% -- -36% -39% -86% -88% -88%
Data::Riak::Fast (REST) 1198/s 121% 89% 57% -- -4% -79% -82% -82%
Net::Riak (PBC) 1254/s 131% 97% 64% 5% -- -78% -81% -81%
Riak::Light (PBC) 5634/s 939% 787% 637% 370% 349% -- -14% -14%
Riak::Client (PBC + AnyEvent) 6557/s 1110% 933% 757% 448% 423% 16% -- 0%
Riak::Client (PBC) 6557/s 1110% 933% 757% 448% 423% 16% 0% --
SEE ALSO
CONTRIBUTORS
Ivan Kruglov <ivan.kruglov@yahoo.com>
AUTHOR
Damien Krotkine <dams@cpan.org>
COPYRIGHT AND LICENSE
This software is copyright (c) 2014 by Damien Krotkine.
This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.