riak-client is a rich Ruby client/toolkit for Riak, Basho's
distributed database that contains a basic wrapper around typical
operations, including bucket manipulation, object CRUD, link-walking,
and map-reduce.
Exhaustive documentation is available at http://basho.github.io/riak-ruby-client/ , and API documentation can be read at http://www.rubydoc.info/gems/riak-client/frames .
Ruby 1.9.3, 2.0, 2.1, and 2.2 are supported. JRuby in 2.0 mode is also supported.
Travis CI is configured to run unit tests on 1.9.3, 2.X versions of Ruby and JRuby.
NOTE: official support for the Ruby 1.9 series ended February of 2015.
In JRuby 1.7.13, OCSP validation is absent, and CRL validation always fails. This issue is being tracked and this document will be updated when it is fixed. Additionally, client certificate authentication doesn't work in JRuby. This issue is also being tracked, and this document will be updated when it works.
riak-client requires beefcake, cert_validator, i18n, innertube, and
multi_json.
Development dependencies are handled with bundler. Install bundler
(gem install bundler) and run this command to get started:
$ bundle installRun the RSpec suite using bundle exec:
$ bundle exec rakerequire'riak'# Create a client interfaceclient=Riak::Client.new# Create a client that uses secure Protocol Buffersclient=Riak::Client.new(authentication: {# certificate authority to validate the server certca_file: '/home/zedo/ca.crt',# username, requireduser: 'zedo',# password for password-based authenticationpassword: 'catnip',# client-cert authentication parameters support filenames,# OpenSSL-compatible string data, or properly initialized# OpenSSL objectsclient_ca: '/home/zedo/ca.crt',cert: File.read'/home/zedo/zedo.crt',key: OpenSSL::PKey::RSA.new(File.read'/home/zedo/zedo.key')})# Automatically balance between multiple nodesclient=Riak::Client.new(:nodes=>[{:host=>'10.0.0.1'},{:host=>'10.0.0.2',:pb_port=>1234},{:host=>'10.0.0.3',:pb_port=>5678}])# Retrieve a bucketbucket=client.bucket("doc")# a Riak::Bucket# Get an object from the bucketobject=bucket.get_or_new("index.html")# a Riak::RObject# Change the object's data and saveobject.raw_data="<html><body>Hello, world!</body></html>"object.content_type="text/html"object.store# Reload an object you already haveobject.reload# Works if you have the key and vclock, using conditional GETobject.reload:force=>true# Reloads whether you have the vclock or not# Access more like a hash, client[bucket][key]client['doc']['index.html']# the Riak::RObject# Create a new objectnew_one=Riak::RObject.new(bucket,"application.js")new_one.content_type="application/javascript"# You must set the content type.new_one.raw_data="alert('Hello, World!')"new_one.storeRiak 2 uses bucket types to enable groups of similar buckets to share properties, configuration, and to namespace values within those buckets. Bucket type support is integral to how CRDTs work.
In versions 2.2.0 and newer of this client, bucket types have a first-class
representation, and can be used to create BucketTyped::Bucket objects that are
namespaced differently from regular Riak::Bucket objects.
# This example assumes you have a "beverages" bucket type.beverages=client.bucket_type'beverages'coffees=beverages.bucket'coffees'untyped_coffees=client.bucket'coffees'chapadao=coffees.new'chapadao'chapadao.data="Chapadao de Ferro"chapadao.store# stores this in the "beverages" bucket typeuntyped_coffees.get'chapadao'# raises error, not foundcoffees.get'chapadao'# succeedschapadao.reload# succeedsuntyped_coffees.delete'chapadao'# silently fails to delete itchapadao.delete# deletes itcoffees.delete'chapadao'# deletes itClient 2.0 and 2.1 code that uses the type argument to methods still works:
coffees=client.bucket'coffees'chapadao=coffees.new'chapadao'chapadao.data="Chapadao de Ferro"chapadao.storetype: 'beverages'# stores this in the "beverages" bucket typecoffees.get'chapadao'# raises error, not foundcoffees.get'chapadao',type: 'beverages'# succeedschapadao.reload# raises error, not foundchapadao.reloadtype: 'beverages'# succeedschapadao.delete# silently fails to delete itcoffees.delete'chapadao'# silently fails to delete itchapadao.deletetype: 'beverages'# deletes itcoffees.delete'chapadao',type: 'beverages'# deletes it# Assuming you've already instantiated a client, get the album titles for The Beatlesresults=Riak::MapReduce.new(client).add("artists","Beatles").link(:bucket=>"albums").map("function(v){ return [JSON.parse(v.values[0].data).title]; }",:keep=>true).runpresults# => ["Please Please Me", "With The Beatles", "A Hard Day's Night",# "Beatles For Sale", "Help!", "Rubber Soul",# "Revolver", "Sgt. Pepper's Lonely Hearts Club Band", "Magical Mystery Tour",# "The Beatles", "Yellow Submarine", "Abbey Road", "Let It Be"]This client supports the new Riak Search 2 (codenamed "Yokozuna"). For more information about Riak Search, see the Riak documentation.
This documentation assumes there's a yokozuna bucket type created and activated.
# Create a client and bucket.client=Riak::Client.newbucket_type=client.bucket_type'yokozuna'bucket=bucket_type.bucket'pizzas'# Create an index and add it to a typed bucket. Setting the index on the bucket# may fail until the index creation has propagated.index=Riak::Search::Index.newclient,'pizzas'index.exist?#=> falseindex.create!client.set_bucket_propsbucket,{search_index: 'pizzas'},'yokozuna'# Store some records for indexingmeat=bucket.new'meat'meat.data={toppings_ss: %w{pepperonihamsausage}}meat.storetype: 'yokozuna'hawaiian=bucket.new'hawaiian'hawaiian.data={toppings_ss: %w{hampineapple}}hawaiian.storetype: 'yokozuna'# Search the pizzas index for hashes that have a "ham" entry in the toppings_ss arrayquery=Riak::Search::Query.newclient,index,'toppings_ss:ham'query.rows=5# return the first five pizzasresult=query.results# returns a ResultsCollection objectresult.length# number of results returnedresult.num_found# total number of results found, including ones not returnedpizza_result=result.first# a ResultDocument of the first pizzapizza_result.score# score of the matchpizza_result.key# also pizza.bucket and pizza.bucket_typepizza=pizza_result.robject# load the actual RObject for the matchRiak supports secondary indexes. Secondary indexing, or "2i," gives you the ability to tag objects with multiple queryable values at write time, and then query them later.
Objects are tagged with a hash kept behind the indexes method. Secondary index
storage logic is in lib/riak/rcontent.rb.
object=bucket.get_or_new'cobb.salad'# Indexes end with the "_bin" suffix to indicate they're binary or string# indexes. They can have multiple values.object.indexes['ingredients_bin']=%w{lettucetomatobaconeggchives}# Indexes ending with the "_int" suffix are indexed as integers. They can# have multiple values too.object.indexes['calories_int']=[220]# You must re-save the object to store indexes.object.storeSecondary index queries return a list of keys exactly matching a scalar or within a range.
# The Bucket#get_index method allows querying by scalar...bucket.get_index'calories_int',220# => ['cobb.salad']# or range.bucket.get_index'calories_int',100..300# => ['cobb.salad']# Binary indexes also support both ranges and scalars.bucket.get_index'ingredients_bin','tomata'..'tomatz'# => ['cobb.salad']# The collection from #get_index also provides a continuation for pagination:c=bucket.get_index'ingredients_bin','lettuce',max_results: 5c.length# => 5c.continuation# => "g2gCbQAAA="# You can use that continuation to get the next page of results:c2=bucket.get_index'ingredients_bin','lettuce',max_results: 5,continuation: c.continuation# More complicated operations may benefit by using the `SecondaryIndex` object:q=Riak::SecondaryIndex.newbucket,'ingredients_bin','lettuce',max_results: 5# SecondaryIndex objects give you access to the keys...q.keys# => ['cobb.salad', 'wedge.salad', 'buffalo_chicken.wrap', ...]# but can also fetch values for you in parallel.q.values# => [<RObject {recipes,cobb.salad} ...>, <RObject {recipes,wedge...# They also provide simpler pagination:q.has_next_page?# => trueq2=q.next_pageRiak 2 features new distributed data structures: counters, sets, and maps (containing counters, flags, maps, registers, and sets). These are implemented by the Riak database as Convergent Replicated Data Types.
Riak data type support requires bucket types to be configured to support each top-level data type. If you're just playing around, use the Riak Ruby Vagrant setup to get started with the appropriate configuration and bucket types quickly.
The examples below presume that the appropriate bucket types are named
counters, maps, and sets; these bucket type names are the client's defaults.
Viewing and changing the defaults is easy:
Riak::Crdt::DEFAULT_BUCKET_TYPES[:set]#=> "sets"Riak::Crdt::DEFAULT_BUCKET_TYPES[:set]="a_cooler_set"The top-level CRDT types have both immediate and batch mode. If you're doing multiple writes to a single top-level counter or set, or updating multiple map entries, batch mode will make fewer round-trips to Riak.
Top-level CRDT types accept nil as a key. This allows Riak to assign a random
key for them.
Deleting CRDTs requires you to use the key-value API for the time being.
brews=Riak::Crdt::Set.newbucket,'brews'brews.add'espresso'brews.add'aeropress'bucket.deletebrews.key,type: brews.bucket_typeRiak 2 integer counters have one operation: increment by an integer.
counter=Riak::Crdt::Counter.newbucket,keycounter.value#=> 15counter.incrementcounter.value#=> 16counter.increment3counter.value#=> 19counter.decrementcounter.value#=> 18Counter operations can be batched:
counter.batchdo |c|
c.incrementc.increment5endRiak 2 maps can contain counters, flags (booleans), registers (strings), sets, and other maps.
Maps are similar but distinct from the standard Ruby Hash. Entries are
segregated by both name and type, so you can have counters, registers, and sets inside a map that all have the same name.
map=Riak::Crdt::Map.newbucket,keymap.counters['potatoes'].value#=> 5map.sets['potatoes'].include?'yukon gold'#=> truemap.sets['cacti'].value#=> #<Set: {"saguaro", "prickly pear", "fishhook"}>map.sets['cacti'].remove'prickly pear'map.registers['favorite butterfly']='the mighty monarch'map.flags['pet cat']=truemap.maps['atlantis'].registers['location']#=> 'kennedy space center'map.counters.delete'thermometers'Maps are a prime candidate for batched operations:
map.batchdo |m|
m.counters['hits'].incrementm.sets['followers'].add'basho_elevator'endFrequently, you might want a map with a Riak-assigned name instead of one you come up with yourself:
map=Riak::Crdt::Map.newbucket,nilmap.registers['coat_pattern']='tabby'map.key#=> "2do4NvcurWhXYNQg8HoIR9zedJV"Sets are an unordered collection of entries.
PROTIP: Ruby and Riak Ruby Client both have classes called Set. Be careful
to refer to the Ruby version as ::Set and the Riak client version as
Riak::Crdt::Set.
set=Riak::Crdt::Set.newbucket,keyset.members#=> #<Set: {"Edinburgh", "Leeds", "London"}>set.add"Newcastle"set.remove"London"set.include?"Leeds"#=> trueSets support batched operations:
set.batchdo |s|
s.add"York"s.add"Aberdeen"s.remove"Newcastle"endThe client code for these types is in the Riak::Crdt namespace, and mostly
in the lib/riak/crdt directory.
For more information about 1.4-style counters in Riak, see the Basho documentation.
Counter records are automatically persisted on increment or decrement. The initial default value is 0.
# Firstly, ensure that your bucket is allow_mult set to truebucket=client.bucket"counters"bucket.allow_mult=true# You can create a counter by using the bucket's counter methodcounter=bucket.counter("counter-key-here")counter.increment=>nilpcounter.value1=>1# Let's increment one more time and then retrieve it from the databasecounter.increment# Retrieval is similar to creationpersisted_counter=Riak::Counter.new(bucket,"counter-key-here")ppersisted_counter.value2=>2# We can also increment by a specified numberpersisted_counter.increment(20)ppersisted_counter.value22=>22# Decrement works much the samepersisted_counter.decrementpersisted_counter.value=>21persisted_counter.decrement(6)persisted_counter.value=>15# Incrementing by anything other than integer will throw an ArgumentErrorpersisted_counter.increment"nonsense"ArgumentError: Counterscanonlybeincrementedordecrementedbyintegers.That's about it. PN Counters in Riak are distributed, so each node will receive the proper increment/decrement operation. Enjoy using them.
The Riak client has built-in event hooks for all major over-the-wire operations including:
- riak.list_buckets
- riak.list_keys
- riak.set_bucket_props
- riak.get_bucket_props
- riak.clear_bucket_props
- riak.get_index
- riak.store_object
- riak.get_object
- riak.reload_object
- riak.delete_object
- riak.map_reduce
- riak.ping
Events are propogated using ActiveSupport::Notifications, provided by the Instrumentable gem.
Instrumentation is opt-in. If instrumentable is not available, instrumentation will not be available. To turn on instrumentation, simply require the instrumentable gem in your app's Gemfile:
gem'instrumentable','~> 1.1.0'Then, to subscribe to events:
ActiveSupport::Notifications.subscribe(/^riak\.*/)do |name,start,finish,id,payload|
name# => String, name of the event (such as 'riak.get_object' from above)start# => Time, when the instrumented block started executionfinish# => Time, when the instrumented block ended executionid# => String, unique ID for this notificationpayload# => Hash, the payloadendThe payload for each event contains the following keys:
:client_id: The client_id of the Riak client:_method_args: The array of method arguments for the instrumented method. For instance, forriak.get_object, this value would resemble[<Riak::Bucket ...>, 'key', {}]
We aim to have a comprehensive and fast set of tests, implemented using a modern, well-supported version of RSpec. These tests include both unit specs for individual classes, and integration specs that ensure the client works properly with an actual Riak instance.
The Riak Ruby Vagrant virtual machine's Riak configuration is normally
used to test this client in development. Once it's up and running, configure
the Ruby test_client.yml on the host machine to connect to pb_port: 17017
and test away.
Configuring the Riak node the tests connect to is done via the
spec/support/test_client.yml file, which is loaded into a Ruby hash with
symbolized keys, and passed to Riak::Client.new.
# test_client.ymlpb_port: 10017# UNCOMMENT AUTHENTICATION SECTION WHEN RIAK HAS SECURITY ENABLED# authentication:# user: user# password: password# ca_file: spec/support/certs/ca.crtSpecs depend on the following Riak configurations:
- The LevelDB backend is necessary for testing secondary indexes.
- allow_mult is required for many features: conflict resolution, and legacy counters among them.
- Riak Search 2.0 ("Yokozuna") must be configured for testing full-text search.
The following bucket types are used during testing:
riak-admin bucket-type create counters '{"props":{"datatype":"counter", "allow_mult":true}}'
riak-admin bucket-type create other_counters '{"props":{"datatype":"counter", "allow_mult":true}}'
riak-admin bucket-type create maps '{"props":{"datatype":"map", "allow_mult":true}}'
riak-admin bucket-type create sets '{"props":{"datatype":"set", "allow_mult":true}}'
riak-admin bucket-type create yokozuna '{"props":{}}'
riak-admin bucket-type activate other_counters
riak-admin bucket-type activate counters
riak-admin bucket-type activate maps
riak-admin bucket-type activate sets
riak-admin bucket-type activate yokozunaClient tests run both with and without security enabled, as we have to test several positive and negative paths. The tests broadly depend on these users and roles:
riak-admin security add-user user password=password
riak-admin security add-user certuser
riak-admin security add-source user 0.0.0.0/0 password
riak-admin security add-source certuser 0.0.0.0/0 certificate
riak-admin security grant riak_kv.get,riak_kv.put,riak_kv.delete,\
riak_kv.index,riak_kv.list_keys,riak_kv.list_buckets,\
riak_core.get_bucket,riak_core.set_bucket,\
riak_core.get_bucket_type,riak_core.set_bucket_type,\
search.admin,search.query,riak_kv.mapreduce on any to userFork the project on Github. If you have already forked, use
git pull --rebaseto reapply your changes on top of the mainline. Example:$ git checkout master $ git pull --rebase basho master
Copy spec/support/test_server.yml.example to spec/support/test_server.yml and change that file according to your local installation of riak.
Create a topic branch. If you've already created a topic branch, rebase it on top of changes from the mainline "master" branch. Examples:
New branch:
``` bash $ git checkout -b topic ```Existing branch:
``` bash $ git rebase master ```
Write an RSpec example or set of examples that demonstrate the necessity and validity of your changes. Patches without specs will most often be ignored. Just do it, you'll thank me later. Documentation patches need no specs, of course.
Make your feature addition or bug fix. Make your specs and stories pass (green).
Run the suite using multiruby or rvm to ensure cross-version compatibility.
Cleanup any trailing whitespace in your code (try @whitespace-mode@ in Emacs, or "Remove Trailing Spaces in Document" in the "Text" bundle in Textmate). You can use the
clean_whitespaceRake task if you like.Commit, do not mess with Rakefile. If related to an existing issue in the tracker, include "Closes #X" in the commit message (where X is the issue number).
Send a pull request to the Basho repository.
Copyright ©2010-2016 Sean Cribbs and Basho Technologies, Inc.
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.
The included photo (spec/fixtures/cat.jpg) is Copyright ©2009 Sean Cribbs, and is licensed under the Creative Commons Attribution Non-Commercial 3.0 license.
gzip encoding support is copied from the ActiveSupport project.
Thank you to all of our contributors!
- Adam Hunter
- akm
- Alexander Lang
- Alex Moore
- André Silva
- Ashley Woodard
- batasrki
- benmills
- Brian Kaney
- Bryce Kerley
- Carl Hall
- Charl Matthee
- dan
- Daniel Reverri
- Dave Perrett
- David Czarnecki
- dn
- Duff OMelia
- Elias Levy
- Eric Cestari
- Eric Redmond
- Hidekazu Tanaka
- Hiroyasu Ohyama
- Jace
- Jack Dempsey
- Jay Adkisson
- Jeff Pollard
- Joe DeVivo
- John Axel Eriksson
- John Leach
- John Lynch
- Jordan Goldstein
- Michael Grosser
- Josh Nichols
- Justin Pease
- Kazuhiro Suzuki
- Kyle Kingsbury
- Lee Jensen
- Luke Bakken
- Marco Campana
- Mat Brown
- Mathias Meyer
- Michael Sullivan
- Misha Gorodnitzky
- Myron Marston
- Nathaniel Talbott
- Nicholas Rowe
- Nicolas Fouché
- oleg dashevskii
- PatrickMa
- Peter Garbers
- Randy Secrist
- Rusty Klophaus
- Ryan Daigle
- Sam Aarons
- Sean Cribbs
- Sebastian Röbke
- Shay Frendt
- Srdjan Pejic
- StabbyCutyou
- Stefan Sprenger
- Technorama, Ltd
- Tyler Hunt
- Wagner Camarao
- Woody Peterson
- Zshawn Syed
