A Ruby ORM for RethinkDB


0.34.0 – Oct. 16, 2019

  • Add support for Rails 6
  • Add support for nonvoting_replica_tags table configuration

0.33.0 – Sep. 27, 2016

  • Allow connection passwords to have special characters through standard URL encodings. #230.
  • Detect connection failures when connected to a proxy.
  • For rails applications, also comment active_record related lines in config/initializers/*.rb during installation #222.
  • Removed .unscoped when fetching belongs_to associations to be consistent with find().
  • update_all() operates without ordering.
  • Fix model reloading with Rails 5.
  • Added a run_options configuration for options.
  • Bug fix where long keys (>255 chars) in NoBrainer::Lock are rejected #214.
  • Add support for user/password authentication on connections #213.
  • Bug fix with Rails5 where strong params where no longer compatible #211.
  • Bug fix where using large indexes caused systematic index rebuilds with sync_indexes #210.

0.32.0 – June 5, 2016

  • Added order() to be an alias of order_by() #194.
  • Added Rails 5 compatiblity #191, #209.
  • Added attribute_will_change!() for compatiblity as an no-op.

0.31.0 – Feb. 7, 2016

  • Allow belongs_to associations to be used in upsert() #186.
  • Added :uniq and :unique options shorthands to the belongs_to keyword.
  • Allow self-referential belongs_to association #184.
  • Primary key time offset operates on a fixed timezone.
  • Guard against partial model definitions.
  • Use index for where(XXX.defined => true) queries #180.
  • Fixed lazy fetching when fetching an undefined value.
  • Added back model.update_attributes() for compatibility with other gems.

0.30.0 – Oct. 3, 2015

  • Fix problem with joins and nil join keys.
  • Add virtual attributes that can be specified with custom RQL.
  • Implementation of NoBrainer::ReentrantLock. Like Lock, but can be locked multiple times before being unlocked #160.
  • Allow compound indexes to be declared with an implicit name.
  • Allow polymorphic queries with first_or_create under certain conditions.

0.29.0 – Aug. 19, 2015

  • When redefining the primary key, the default primary key generator is no longer used.
  • Try to emit validation errors when upsert() fails with missing keys.
  • upsert() updates the document when found.
  • Allow :class to be used as a shorthand for :class_name in association declarations.
  • Cleaned up the rake tasks. Added rake nobrainer:rebalance.
  • Removed the query retry on connection failure. This was useful when r.table().wait was not available.
  • Added support for EventMachine by using fibers.
  • Added support for the Enum type, which adds a few helper methods similar to ActiveRecord #153.
  • Added hooks nobrainer_field_defined and nobrainer_field_undefined for custom types which are triggered when a field is defined and undefined.

0.28.0 – Aug. 2, 2015

  • Added SSL options for encrypted connections to the server.
  • Added criteria.join() to perform inner joins on associations.
  • Added instance.unset(attr) to remove a field from the document.
  • Added Model.upsert() which uses first_or_create behind the scenes.
  • Added _where() to allow unsafe where() operation (which disables type checking, and enable dynamic attributes).
  • Added the where modifier undefined as a shortcut for where(:field.defined => !value).
  • Added the where modifier include as a shortcut for where(:field.any.eq => value).
  • Minor bug fixes with lazy fetching logic and atomic operations.
  • Provide a NoBrainer.eager_load(docs, ...) method to eager load associations on a given array of model instances.
  • Allowing to specify multiple rethinkdb_urls in the configuration to provide fault tolerance capabilities.
  • Added system table models (e.g. NoBrainer::System::TableConfig).
  • Allow tables to be configured (e.g. shards, replicas) with a task rake nobrainer:sync_schema to apply these settings.
  • Bug fix: fix conflicting keys logic between where() and first_or_create() #151.
  • Implemented touch() on model instances to update the updated_at field #150.

0.27.0 – July 12, 2015

  • belongs_to associations no longer use the primary key name of the target class to avoid depending on loading models in a certain order. #132.
  • Optimized boot time.
  • Bug fix: the installer was saving its configuration file in config/initializer/nobrainer.rb, but should have been config/initializers/nobrainer.rb.
  • Added Document#cache_key for compatiblity with Rails caching mechanisms #149.

0.26.0 – June 20, 2015

  • Added a rails g nobrainer:install task.
  • Updated reconnect logic to match latest RethinkDB messages.
  • Synchronizing indexes is now performed with a distributed lock to avoid races during deployments when multiple servers attempt to synchronize indexes.
  • Dirty tracking no longer go through user defined getters, but operates on internal attributes.
  • Bug fix: Previously assigned values wouldn’t necessarily receive updates when using queue_atomic { }.

0.25.1 – June 1, 2015

  • Added criteria.changes() which returns a changefeed.
  • Added criteria.first_or_create() for atomic creates.
  • Minor fix for uniqueness validators in the context of polymorphic classes.
  • Named scopes are cached when used in a chain (e.g. post.comments.recent).
  • Warn when not using .raw in a criteria when running under run_with(profile: true).
  • Added a synchronize { } method for distributed locks.
  • Removed criteria.scoped method, it seems useless.
  • NoBrainer will now raise if a critera is leaked from a NoBrainer.run_with() block.
  • Removed Model.store_in(db: 'xxx') due to hard to deal with conflict semantics with run_with().
  • NoBrainer.with_database() is deprecated. NoBrainer.run_with(db: 'xxx') should be used instead.
  • NoBrainer.with() has been renamed to NoBrainer.run_with().
  • Removed find_by() due to conflicting semnantics with ActiveRecord #145.
  • Allow nested hash queries (e.g. Model.where(:lang => { :en => 'yes' } }))

0.24.0 – May 20, 2015

  • save/update/create no longer raises on failed validations. We are back to the original ActiveRecord behavior #141.
  • Validations are now performed before calling the before_save callbacks. We are back to the original ActiveRecord behavior #142.
  • Allow circles to be used in where(:field.near => circle) queries.
  • Enforce valid fields in order_by().
  • Allow .not to be chainable (e.g. Model.where( => []). Removed the .nin and .ne criteria operators.
  • Allow Model.where( => []) to return an empty array #144.
  • Forbid a belongs_to’s foreign_key declaration to conflict with another belongs_to’s foreign_key.
  • Allow to find associations with classes within same module #136.
  • Make private #139.

0.23.0 – Apr. 18, 2015

  • Upgraded to RethinkDB 2.0.
  • Bug fix: criteria.to_json was triggering a stack overflow #134.
  • Added a modular profiler, which feeds the logger and the rails controller runtime instrumentation #130.
  • Validation of foreign keys in belongs_to associations are always validated #133.
  • Bug fix: Associations properly support custom primary keys #132.

0.22.0 – Mar. 7, 2015

  • find_by() introduced #127.
  • find() no longer applies unscoped.
  • store_in accepts symbols.
  • Bug fix: NoBrainer considered all validated attributes to be DB fields.

0.21.0 – Jan. 19, 2015

  • Regexp matches now support /i and /m properly: #120.
  • required => true validation shorthand uses the presence validator except for Boolean which uses the non null validator: #109, #110, #118.
  • Bug fix: the uniqueness validator would raise an exception when dealing with invalid types: #117.
  • auto_create_databases and auto_create_tables are no longer configurable. Automatic database/table creation is always on.
  • Added the validation shorthands min_length and max_length.
  • Added a new type Text meant to be used for long strings. Regular String types are now limited to 255 characters. This limit is configurable with config.max_string_length.
  • ActiveModel::ForbiddenAttributesError is raised if Rails controller params are passed directly to NoBrainer without using strong parameters.
  • Implementation of a distributed lock NoBrainer::Lock, which is used by the uniqueness validator.
  • Drop criteria cache when it has too many elements to avoid out of memory issues (configurable with config.criteria_cache_max_entries, defaults to 10,000).
  • Raise errors when using attributes from controller params without sanitizing them with strong params. The check is performed during attribute assignement and in where() queries.
  • preload() is now eager_load().
  • default value procs are now executed in the context of the document. This is useful to set values depending on other values.
  • Added a :scope option for has_many and has_many_through associations. The provided scope is run in the context of the target model, which is useful to reuse named scoped: #115.
  • has_many_through associations have a target_model method to access the underlying model: #114.
  • Allow default_scope to be defined multiple times on models and their subclasses. They all get to be applied: #113.
  • Validations are only run against changed fields.
  • Atomic operations use sane default values.
  • Allow atomic values to be read, it’s easier for debugging and callbacks handlers.
  • Detect potential miseuses with :or arguments ([{...},{...}] vs [...,...]). Using :_or does not check these misuses.
  • Added a :length validation shorthand on field definitions.
  • Added the not null validator. validates_not_null :field or validates :field, :not_null can be used to validates non null values.
  • Bug fix: Allow delete_all/update_all/replace_all to operate on multi indexes.
  • Introduced criteria.without_distinct keyword to ommit distinct on multi indexes.
  • Bug fix: where(:field.defined => value) properly cast value to a boolean.

0.20.0 – Dec. 3, 2014

  • Renamed find! to find. Sticking with ORMs established API to be safer overall.
  • Geo support. Introduced the Geo::Point type and others. Querying with where(:field.near => {}) and where(:field.intersects => geo_object).
  • Generated primary keys use a [A-Za-z0-9]{14} format.

0.19.0 – Nov. 17, 2014

  • :as is now :store_as.
  • To access a multi index, any must be used. e.g. Post.where(:tags.any => 'tag').
  • Introducing all and any modifiers in where() queries.
  • Bug fix: Table creation should use the aliased primary key: #80.
  • where() only allow queries on known fields, unless the model uses dynamic attributes.
  • created_at and updated_at can be updated.
  • Extracted the symbol decoration logic into a separate gem.
  • Alias create! to create to provide backward compatibility with gems.
  • Bug fix: indexed :or queries should return distinct elements.
  • Atomic operations seems now ready for prime time.

0.18.1 – Nov. 8, 2014

  • Added config.app_name and config.environment for non Rails apps.
  • config.max_reconnection_tries is now max_retries_on_connection_failure.
  • Added criteria.extend() to add chainable extend behavior.
  • Bug fix: more atomic operations can now be performed after save().
  • Added an :external option to indexes.
  • Renamed index synchronization rake task to nobrainer:sync_indexes.
  • Indexes are kept synchronized through an auxilary table.
  • Index deletions are no longer confirmed.
  • Added a :uniq validation shorthand (alias of :unique).
  • Added a :format validation shorthand on field definitions.
  • update is now an alias of update_attributes.
  • Renamed rake namespace from db: to nobrainer:.
  • config.reconnection_tries defaults to 1 in development mode, or 15 otherwise.
  • criteria is available in the middleware stack, useful for monitoring tools.
  • order_by accepts Arrays as arguments as well.
  • min/avg/max return nil by default.

0.17.0 – Sep. 25, 2014

  • Implementation of atomic operations.
  • save() is now save?() and save!() is now save(). Same for update_attributes().
  • Added a per_thread_connection configuration flag (until we have a real connection pool from the driver).

0.16.0 – Aug. 27, 2014

  • Added support for lazy fetching on given fields.
  • Added support for Binary types.
  • The logger shows non indexed queries.
  • Added support for pluck() and without(): #86.
  • :or queries are optimized to leverage indexes: #85.
  • Added aliases support with :as: #84.
  • with_index() influences order_by(): #82.
  • Added a connection lock for thread safety: #80.
  • with_index() can take no argument to force the use of an index.
  • Bug fix with the Rails reloader: #78.
  • Added aggregate functions on criteria (min/max/avg/sum).

0.15.0 – July 27, 2014

  • Allow custom types serialization and cast methods.
  • Added local/utc timezone translation for Time typed attributes, configurable with user_timezone and db_timezone.
  • Added ISO8601 Time casting.
  • Added Date casting. They are stored as UTC times in the DB.
  • Added a defined symbol modifier to test for undefined fields.
  • Added a sample method to pick a random document.
  • Added a :in field definition shorthand for inclusion validation.
  • Support types and indexes when using Rails scaffolding by Josh Kuhn for

0.14.0 – June 17, 2014

  • Added custom primary keys support.
  • Optimized certain queries.
  • Added some parallel testing (with rake)
  • Updated to the new RethinkDB driver and its new JSON transport mechanism.

0.13.1 – March 31, 2014

  • update_indexes waits for completion by default.
  • Added the nin operator for where() queries.
  • Renamed AssociationNotSaved to AssociationNotPersisted.
  • Added a :unique => true options on fields as a shorthand for the unique validation.
  • Added support for queries on belongs_to associations. e.g. Comment.where(:post => Post.first).
  • Added support for distributed locks when performing uniqueness validations.
  • Removed the document.attributes= accessor.
  • Bug fix: document dirtiness is now properly cleaned after a database read.
  • Symbol typed fields are now read back from the database as symbols and not strings.
  • Bug fix: before_validation callbacks no longer stop the validation chain when returning false.
  • Added support for Rails 4.1.

0.13.0 – Jan. 12, 2014

  • Removed update() and replace() for the model instance.
  • Removed inc_all() and dec_all() for criteria.
  • Reinstantiating an instance model from the database no longer goes through the setters to keep things consistent with the rest of the API.
  • Dirty tracking tracks changes from an undefined field, to a field set to nil.
  • Hashes are updated with r.literal() to avoid the use of replace.
  • Added a :required => true options on fields as a shorthand for the presence validation.
  • Improved the reconnection mechanism.
  • Added support for readonly fields.
  • Removed Rails3 compatibility.

0.12.0 – Jan. 8, 2014

  • Timestamps are no longer enabled by default.
  • where() validates and casts all the values with respect to their declared field types. This avoid potential query injections.
  • Saving a model will only update the attributes that have changed. No database update will be performed in case nothing has changed.
  • after_find() callbacks are available on models, and also on criteria.
  • Removed .to_xml. This feature is still available by including a module.
  • License changed from MIT to LGPLv3.

0.11.0 – Jan. 7, 2014

  • Using indexes for gt, gte, lt, lte operators when possible.
  • Renamed criteria.includes() to criteria.preload().
  • Added a :validates option on fields.
  • Fixed dirty tracking with mutable values such as hashes and array.
  • Added initialize callbacks.
  • Boolean types adds a field_name? method for convenience.

0.10.0 – Jan. 6, 2014

  • Fixed a NameError bug when trying to include the DynamicAttributes module. Issue: #54.
  • Make the associations hackable.
  • Implementation of the type checking/casting mechanism.

0.9.1 – Jan. 5, 2014

  • Added Rails generators for models
  • Removed unecessary ActiveSupport requires.

0.9.0 – Jan. 5, 2014

  • Removed the auto_include_timestamps and include_root_in_json settings. Because The order in which the models are declared and NoBrainer configured affected the result. Related issue #52
  • Removed the cache_documents setting because it should not be broken.
  • Bug fix with order_by() which would try to use an index after a RQL filter() or get_all().
  • includes() no longer kill the criteria cache.
  • Loading a has_many association will set the corresponding belongs_to association, with or without eager loading.
  • Added the has_many through association. The implementation is done through eager loading.
  • Added the has_one association.
  • Renamed with_options() -> with().

0.8.0 – Dec. 31, 2013

  • First documented release of NoBrainer.