Waistband

Build Status

Configuration and sensible defaults for ElasticSearch on Ruby. Handles configuration, index creation, quality of life, etc, of Elastic Search in Ruby.

Waistband doens't handle connections or API requests, it merely acts as a translator between commonly used patterns and the underlying elasticsearch gems.

Installation

Install ElasticSearch:

brew install elasticsearch

Add this line to your application's Gemfile:

gem 'waistband'

And then execute:

$ bundle

Or install it yourself as:

$ gem install waistband

Configuration

Configuration is generally pretty simple. First, create a folder where you'll store your Waistband configuration docs, usually under #{APP_DIR}/config/waistband/, you can also just throw it under #{APP_DIR}/config/ if you want. The baseline config contains something like this:

# #{APP_DIR}/config/waistband/waistband.yml
development:
    retries: 5
    timeout: 2
    reload_on_failure: true
    servers:
        server1:
            protocol: http
            host: localhost
            port: 9200

You can name the servers whatever you want. The connection, retries, timeouts, etc are handled by the elasticsearch-transport gem.

Here's an example with two servers:

# #{APP_DIR}/config/waistband/waistband.yml
development:
    retries: 5
    timeout: 2
    reload_on_failure: true
    servers:
        server1:
            protocol: http
            host: 173.247.192.214
            port: 9200
        server2:
            protocol: http
            host: 173.247.192.215
            port: 9200

You'll need a separate config file for each index you use, containing the index settings and mappings. For example, for my search index, I use something akin to this:

# #{APP_DIR}/config/waistband/waistband_search.yml
development:
    stringify: false
    settings:
        index:
            number_of_shards: 4
    mappings:
        event:
            _source:
                includes: ["*"]

Note that you can configure specific connection settings for a particular index, if for some reason or another that index in particular uses a different set of servers:

development:
  connection:
    timeout: 2
    retries: 5
    reload_on_failure: true
    servers:
      server1:
        host: 192.168.31.112
        port: 9200
        protocol: http
      server2:
        host: 192.168.31.113
        port: 9200
        protocol: http
  settings:
    index:
      number_of_shards: 1
      number_of_replicas: 1
  mappings:
    event:
      _source:
        includes: ["*"]

List of config settings:

  • settings: settings for the Elastic Search index. Refer to the "admin indices update settings" document for more info.
  • mappings: the index mappings. More often than not you'll want to include all of the document attribute, so you'll do something like in the example above. For more info, refer to the mapping reference.
  • retries: number of times to retry before moving on to the next server node.
  • reload_on_failure: should we reload the node list from the server on failure.
  • timeout: seconds till a timeout exception is raise when trying to connect to the node.
  • name: optional - name of the index. You can (and probably should) have a different name for the index for your test environment. If not specified, it defaults to the name of the yml file minus the waistband_ portion, so in the above example, the index name would become search_#{env}, where env is your environment variable as defined in Waistband::Configuration#setup (determined by RAILS_ENV or RACK_ENV).
  • stringify: optional - determines wether whatever is stored into the index is going to be converted to a string before storage. Usually false unless you need it to be true for specific cases, like if for some key => value pairs the value is of different types some times.
  • connection: optional - determines which server/s the index uses. If not present, it'll use the default connection settings specified in config/waistband.yml.
  • permissions: optional - determines which permissions to allow on this index, please refer to the permissions section for more information.

Initializer

Waistband will look for config files by default in File.join(Rails.root, 'config'). You can override the default location of the config folder. After getting all the YML config files in place, you'll just need to hook up an initializer to these files:

# #{APP_DIR}/config/initializers/waistband.rb
Waistband.configure do |c|
    c.config_dir = "#{APP_DIR}/spec/config/waistband"
end

Usage

Indexes

Creating and destroying the indexes

For each index you have, you'll probably want to make sure it's created on initialization, or in a Rake task, so either in the same waistband initializer or in another initializer, depending on your preferences, you'll have to create them. For our search example:

# #{APP_DIR}/config/initializers/waistband.rb
# ...
Waistband::Index.new('search').create

This will create the index if it's not been created already or return nil if it already exists. If you want to raise an exception if it already exists, use the #create! method.

Destroying an index is equally easy:

Waistband::Index.new('search').delete

When writing tests, it would generally be advisable to destroy and create the indexes in a before(:each) or before(:all) depending in your circumstances. Also, remember for testing that replication and data availability is not inmediate on the indexes, so if you create an immediate expectation for data to be there, you should refresh the index before it:

Waistband::Index.new('search').refresh

Note: most index methods such as create, delete, etc, have an equivalent bang method (delete!) that will actually throw an exception if something goes wrong. For example, delete will return nil if the index doesn't exist, but will raise any other unrelated exceptions, whereas delete! will raise even the Index Not Found exception.

Writing, reading and deleting from an index

index = Waistband::Index.new('search')

# writing
index.save('my_data', {'important' => true, 'valuable' => {'always' => true}}) # => true

# reading
index.find('my_data') # => {"important"=>true, "valuable"=>{"always"=>true}}

# reading with all the internal data
index.read('my_data') # => {'_id' => '123123', '_source' => {"important"=>true, "valuable"=>{"always"=>true}}, ...}

# deleting
index.destroy('my_data') # => "{\"ok\":true,\"found\":true,\"_index\":\"search\",\"_type\":\"search\",\"_id\":\"my_data\",\"_version\":2}"

# reading non-existent data
index.find('my_data') # => nil

Searching

For searching, you construct a search from your index:

index = Waistband::Index.new('search')
results = index.search({
    query: {
        term: { hidden: false }
    },
    sort: { created_at: {order: 'desc' } },
    page: 1,
    page_size: 5
})

results.hits # => returns a search results object
results.total_hits # => 28481

For paginating the results, you can use the #paginated_results method, which will provide an array object compatible with the kaminari gem. If you use another gem, you can just override the method, etc.

For more information and extra methods, take a peek into the class docs.

Also, for convenience, the gem provides the Result class, which just provides some quality-of-life methods for working with search result hashes or their inner _source hashes, for example:

search = index.search({
    query: {
        term: { hidden: false }
    }
})
results = search.results
result = result.first

result._id # => '123123'
result._type # => 'search_result'
result._index # => 'search'
result.task_id # => 991122 -- note that this is a method missing interface directly either to the search result hash, or to the _source sub-hash

The Result class is directly exposed via two methods in the SearchResults class: #results and #paginated_results. You can use #paginated_results if you're using Kaminari for pagination and wish to use the awesomeness it provides.

Index Aliasing

Sometimes it can be useful to sub-divide your index into smaller indexes based on dates or other partitioning schemes. To do this, the Index class exposes the subs option on instantiation:

index = Waistband::Index.new('events', subs: %w(2013 01))
index.create!

This creates the index events__2013_01, which in your application logic you could design to store all event data for Jan 2013. You'd do the same for Feb, etc., and when you no longer need one of the older ones, you could delete just that sub-index, instead of things getting more complicated.

We've also found quite a bit of usefulness in using index versioning, so you can add/remove fields to your object without much worry. Waistband accomodates this pattern as follows:

index = Waistband::Index.new('events', version: 1)
index.create!

Aliasing

Part of subbing is gonna be creating the correct aliases that group up your sub-indexes.

index = Waistband::Index.new('events', subs: %w(2013 01))
index.create!
index.alias('my_super_events_alias') # => true
index.alias_exists?('my_super_events_alias') # => true

The alias methods receives a param to define the alias name. The same pattern can be used when using index versions.

Permissions

We've found it safer to tighten up the permissions to determine which actions can be done on each index based on environments. For example, you might want to allow an index to be created and deleted at will on the development or staging environments, but you probably don't want to allow it to be deleted on production. To that effect, you're able to set permissions on each index's config file:

production:
    permissions:
        create: true
        delete_index: false
        destroy: true
        read: true
        write: true

By default, all permissions are true unless set otherwise.

The specific permissions are:

  • create: can Waistband create the index?
  • delete_index: can Waistband delete the entire index?
  • destroy: can Waistband destroy an object in the index?
  • read: can Waistband read, find, or find_result in the index?
  • write: can Waistband save (create or update) an object to the index?

Logging

It's very useful to log what the output from all operations is, with that in mind, you can specify a log for Waistband easily:

Waistband.configure do |c|
  c.logger = Rails.logger.dup
end

You can do this in an initializer, even the same initializer you used to specify the config_dir ealirer if you did so.

Contributing

  1. Fork it
  2. Create your feature branch (git checkout -b my-new-feature)
  3. Commit your changes (git commit -am 'Add some feature')
  4. Push to the branch (git push origin my-new-feature)
  5. Create new Pull Request