Skip to content

Latest commit

 

History

History
720 lines (530 loc) · 24 KB

README.md

File metadata and controls

720 lines (530 loc) · 24 KB

Riak Ruby Client (riak-client)

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 .

Build Status

Build Status

Dependencies

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 install

Run the RSpec suite using bundle exec:

$ bundle exec rake

Basic Example

require 'riak'

# Create a client interface
client = Riak::Client.new

# Create a client that uses secure Protocol Buffers
client = Riak::Client.new(authentication: {
      # certificate authority to validate the server cert
      ca_file: '/home/zedo/ca.crt',

      # username, required
      user: 'zedo',

      # password for password-based authentication
      password: 'catnip',

      # client-cert authentication parameters support filenames,
      # OpenSSL-compatible string data, or properly initialized
      # OpenSSL objects
      client_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 nodes
client = 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 bucket
bucket = client.bucket("doc")  # a Riak::Bucket

# Get an object from the bucket
object = bucket.get_or_new("index.html")   # a Riak::RObject

# Change the object's data and save
object.raw_data = "<html><body>Hello, world!</body></html>"
object.content_type = "text/html"
object.store

# Reload an object you already have
object.reload                  # Works if you have the key and vclock, using conditional GET
object.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 object
new_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.store

Bucket Types

Riak 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 type

untyped_coffees.get 'chapadao' # raises error, not found
coffees.get 'chapadao' # succeeds

chapadao.reload # succeeds

untyped_coffees.delete 'chapadao' # silently fails to delete it
chapadao.delete # deletes it
coffees.delete 'chapadao' # deletes it

Client 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.store type: 'beverages' # stores this in the "beverages" bucket type

coffees.get 'chapadao' # raises error, not found
coffees.get 'chapadao', type: 'beverages' # succeeds

chapadao.reload # raises error, not found
chapadao.reload type: 'beverages' # succeeds

chapadao.delete # silently fails to delete it
coffees.delete 'chapadao' # silently fails to delete it

chapadao.delete type: 'beverages' # deletes it
coffees.delete 'chapadao', type: 'beverages' # deletes it

Map-Reduce Example

# Assuming you've already instantiated a client, get the album titles for The Beatles
results = Riak::MapReduce.new(client).
                add("artists","Beatles").
                link(:bucket => "albums").
                map("function(v){ return [JSON.parse(v.values[0].data).title]; }", :keep => true).run

p results # => ["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"]

Riak Search Examples

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.new
bucket_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.new client, 'pizzas'
index.exist? #=> false
index.create!
client.set_bucket_props bucket, {search_index: 'pizzas'}, 'yokozuna'

# Store some records for indexing
meat = bucket.new 'meat'
meat.data = {toppings_ss: %w{pepperoni ham sausage}}
meat.store type: 'yokozuna'

hawaiian = bucket.new 'hawaiian'
hawaiian.data = {toppings_ss: %w{ham pineapple}}
hawaiian.store type: 'yokozuna'

# Search the pizzas index for hashes that have a "ham" entry in the toppings_ss array
query = Riak::Search::Query.new client, index, 'toppings_ss:ham'
query.rows = 5 # return the first five pizzas

result = query.results # returns a ResultsCollection object
result.length # number of results returned
result.num_found # total number of results found, including ones not returned

pizza_result = result.first # a ResultDocument of the first pizza
pizza_result.score # score of the match
pizza_result.key # also pizza.bucket and pizza.bucket_type

pizza = pizza_result.robject # load the actual RObject for the match

Secondary Index Examples

Riak 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.

Tagging Objects

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{lettuce tomato bacon egg chives}

# 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.store

Finding Objects

Secondary 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: 5
c.length # => 5
c.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.new bucket, '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? # => true
q2 = q.next_page

Riak 2 Data Types

Riak 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.new bucket, 'brews'
brews.add 'espresso'
brews.add 'aeropress'

bucket.delete brews.key, type: brews.bucket_type

Counters

Riak 2 integer counters have one operation: increment by an integer.

counter = Riak::Crdt::Counter.new bucket, key

counter.value #=> 15

counter.increment

counter.value #=> 16

counter.increment 3

counter.value #=> 19

counter.decrement

counter.value #=> 18

Counter operations can be batched:

counter.batch do |c|
  c.increment
  c.increment 5
end

Maps

Riak 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.new bucket, key

map.counters['potatoes'].value #=> 5
map.sets['potatoes'].include? 'yukon gold' #=> true

map.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'] = true

map.maps['atlantis'].registers['location'] #=> 'kennedy space center'

map.counters.delete 'thermometers'

Maps are a prime candidate for batched operations:

map.batch do |m|
  m.counters['hits'].increment
  m.sets['followers'].add 'basho_elevator'
end

Frequently, you might want a map with a Riak-assigned name instead of one you come up with yourself:

map = Riak::Crdt::Map.new bucket, nil

map.registers['coat_pattern'] = 'tabby'

map.key #=> "2do4NvcurWhXYNQg8HoIR9zedJV"

Sets

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.new bucket, key

set.members #=> #<Set: {"Edinburgh", "Leeds", "London"}>

set.add "Newcastle"
set.remove "London"

set.include? "Leeds" #=> true

Sets support batched operations:

set.batch do |s|
  s.add "York"
  s.add "Aberdeen"
  s.remove "Newcastle"
end

Client Implementation Notes

The client code for these types is in the Riak::Crdt namespace, and mostly in the lib/riak/crdt directory.

Riak 1.4 Counters

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 true
bucket = client.bucket "counters"
bucket.allow_mult = true

# You can create a counter by using the bucket's counter method
counter = bucket.counter("counter-key-here")
counter.increment
=> nil

p counter.value
1
=> 1

# Let's increment one more time and then retrieve it from the database
counter.increment

# Retrieval is similar to creation
persisted_counter = Riak::Counter.new(bucket, "counter-key-here")

p persisted_counter.value
2
=> 2

# We can also increment by a specified number
persisted_counter.increment(20)
p persisted_counter.value
22
=> 22

# Decrement works much the same
persisted_counter.decrement
persisted_counter.value
=> 21

persisted_counter.decrement(6)
persisted_counter.value
=> 15

# Incrementing by anything other than integer will throw an ArgumentError
persisted_counter.increment "nonsense"
ArgumentError: Counters can only be incremented or decremented by integers.

That's about it. PN Counters in Riak are distributed, so each node will receive the proper increment/decrement operation. Enjoy using them.

Instrumentation

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.

Enabling

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 execution
  finish  # => Time, when the instrumented block ended execution
  id      # => String, unique ID for this notification
  payload # => Hash, the payload
end

The 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, for riak.get_object, this value would resemble [<Riak::Bucket ...>, 'key', {}]

Running Specs

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.yml
pb_port: 10017
# UNCOMMENT AUTHENTICATION SECTION WHEN RIAK HAS SECURITY ENABLED
# authentication:
#   user: user
#   password: password
#   ca_file: spec/support/certs/ca.crt

Spec dependencies

Specs 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 yokozuna

Client 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 user

How to Contribute

  • Fork the project on Github. If you have already forked, use git pull --rebase to 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_whitespace Rake 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.

License & Copyright

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.

Auxillary Licenses

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. "Creative Commons"

gzip encoding support is copied from the ActiveSupport project.

Contributors

Thank you to all of our contributors!