CHANGELOG.md

praxis-mapper changelog

4.5 (6/25/2019)

• Support wrapping associations in resources for targets that do not explicitly contain the Enumerable module. For example, ActiveRecord::Associations::CollectionProxy are actually not enumerable, yet they proxy the calls directly to the underlying objects when necessary.

4.4.1 (6/25/2019)

• Fix Resource to not override association methods already defined
• Migrate to FactoryBot

4.4 (12/10/2017)

• Fix selector generator to properly handle composite key associations.
• Fix default_selection in Query::Base to accountt for models with composite identities.
• Support many_to_many relationships without requiring the :through annotations (for ORMs that can transparently load the intermediate relationships)

4.3

• Added a :through option to Resource.property to specify that the property effectively traverses the association(s) given so that it may continue to process additional subfield selections.
• Fixed SystemStackError from IdentityMap#add_selectors if the provided field hashes are self-referencing. I.e. {author: {posts: {author: {...}}}}
• Fixed SystemStackError from SelectorGenerator for circular Resource.property definitions like property :foo, dependencies: [:foo, :bar].
• Fixed Query::Sequel not working with raw SQL queries.
• Fixed SQL logging in Query::Sequel for query execution passing a dataset.

4.2

• Added Resource.property to specify the fields and associations from the underlying model that are needed by methods added to the Resource.
  • For example:
    • property :full_name, dependencies: [:first, :last] specifies that the full_name method on the resource depends upon the values of the first and last fields from the model.
    • property :owner_name, dependencies: ['owner.first', 'owner.last'] defines a dependency on the first and last fields from the associated owner.
  • Dependencies must always be an array. Symbol values should be used for fields or associations from the same resource, and strings can be used for fields from associated resources.
  • :* or 'association.*' may be used to specify a dependency on all of the fields of the resource (analagous to 'SELECT *').
  • Additionally, property dependencies may specify other properties on the same or related resources.
• Added IdentityMap#add_selectors(resource, field_selector) to define fields to select and associations to track given resource and field_selectors.
  • For example:
    • identity_map#add_selectors(PersonResource, full_name: true) would ensure any loads done for the model the PersonResource describes will select any fields required by the full_name property (first and last in the example above).
    • identity_map#add_selectors(PersonResource, id: true, address: true) could result in selecting the id and address_id fields of PersonModel, as well as tracking the address association, for any queries involving PersonModel.
• Added the ability to explicitly perform a "SELECT *" by specifying select :* or select '*' in a query.
• Added :through option to many_to_many associations to define the series of associations necessary to resolve it. This is unused by Sequel, but is required by the SelectorGenerator to handle such associations. Specify it as an array of symbols.
  • For example: many_to_many :commented_posts, through: [:comments, :post], ...

4.1.2

• Remove finalizer from IdentityMap to fix memory leak that was preventing them from being GC'd properly.

4.1.1

• Fixed critical bug in IdentityMap#finalize!=

4.1

• Added instrumentation through ActiveSupport::Notifications
  • Use the 'praxis.mapper.load' to subscribe to IdentityMap.load calls
  • Use the 'praxis.mapper.finalize' to subscribe to IdentityMap.finalize! calls

4.0

• Optimization on handling repeated ids (especially noticeable when using subloads)
• Refactored ConnectionManager repository handling to improve integration with other connection-pooling (specifically Sequel's at present).
• Added two types of connection factory under Praxis::Mapper::ConnectionFactories:::
  • Simple: Takes a connection: option to specify the raw object to return for all checkout. Also, preserves current behavior with proc-based uses when ConnectionManager.repository is given a block. This is the default factory type if one is not specified for the repository.
  • Sequel: Takes connection: option to specify a Sequel::Database, or hash of options to pass to Sequel.connect.
• IdentityMap#finalize! now calls ConnectionManager#release to ensure any connections are returned to their respective pools, if applicable.
• Added SequelCompat module, which provides support for using Sequel::Model objects with an IdentityMap when included in model class.
  • This overrides the association accessors on instances associated with an IdentityMap (see below for more on this) to query the map instead of database.
  • See (spec/support/spec_sequel_models.rb) for example definition.
• Added prototype for write-path support to IdentityMap (for Sequel::Model models):
  • IdentityMap#attach(record): adds the record to the identity map, saving it to the database first if it does not have a value for its primary key (or other identities).
  • IdentityMap#detatch(record): removes the record from the identity map
  • IdentityMap#flush!(record_or_class=nil): depending on the value of record_or_class it:
    • with no argument, or nil, given it saves all modified records in the identity map.
    • with an instance of a model, it saves just that record.
    • with a model class, it saves all modified records for that class in the identity map.
  • IdentityMap#remove: calls detatch with the record, and then callsrecord.delete to delete it.

3.4.0

• Add preliminary Sequel-centric Query::Sequel class that exposes a a Sequel dataset with an interface compatible with other Query types.
• Removed member_name and collection_name methods from Resource, as well as the type alias.
• Fixed IdentityMap#load to return the records from the identity map corresponding to rows returned from the database. This fixes a bug where loading rows already existing in the identity map would return new record objects rather than the objects in the map.

3.3.0

• Tightened up handling of missing records in IdentityMap#all and Resource.wrap. They now will never return nil values.

3.2.0

• Add Resource.decorate for extending/overriding methods on Resource associations. See PersonResource in spec_resources.rb for usage.

3.1.2

• Fixed handling of loads where :staged, which could cause an incorrect underlying query using a "staged" column.
• Relaxed ActiveSupport version dependency from 4 to >=3

3.1

• Begin migration to Sequel for Query::Sql.
  • #_execute uses it for a database-agnostic adapter for running raw SQL.
  • #_multi_get uses it generate a datbase-agnostic where clause
• Added accessor generation for one_to_many associations in resources.
• Imported Model extensions for FactoryGirl to praxis-mapper/support/factory_girl.
• This also adds IdentityMap#persist! to insert all rows in the identity map into the database.
• Added SchemaDumper and SchemaLoader for dumping and loading schema information in praxis-mapper/support/schema_dumper and praxis-mapper/support/schema_loader respectively.
• IdentityMap#all now accepts non-identity keys as conditions, and will create a secondary index for the key the first time it's queried.
• Added Model#identities to get a hash of identities and values for a record.
• Auto-generated attribute accessors on Blueprint do not load (coerce) the value using the attribute. Ensure values passed in already have the appropriate types. Blueprint attributes will still be wrapped properly, however.
• Performance and memory use optimizations.
• IdentityMap#load now supports eagerly-loading associated records in a query, and supports the full set of options on the inner query, including
• Tracked :one_to_many associations now support where clauses. Using where clauses when tracking other association types is not supported and will raise an exception.

3.0

• Moved Blueprint inner attribute creation to finalize!
• Added :dsl_compiler and :identity as valid options for Blueprint.
• Praxis::Mapper::Model
  • Model.context allows named query parameters to be reused across queries. See documentation for details.
  • Model#_query references original query that loaded the record.
  • Model.serialized_fields returns hash of fields defined with either json or yaml directives as serialized fields.
  • Fixed accessors will to raise KeyError if the record does not the field (i.e., if it was not loaded from the database), rather than silently returning nil.
• Praxis::Mapper::Query:
  • Multiple select and track calls within one query are now additive.
  • track option now takes a block that will be applied to the query used to load the tracked records.
  • context directive to apply the named context from the model to the query. See documentation for more.
• Praxis::Mapper::IdentityMap
  • Removed :track option from #add_records in favor of using Model#_query to determine tracked associations.
• Added Praxis::Mapper::Support with MemoryQuery and MemoryRepository for use in testing data loading without requiring a database.

Notices

The Model accessor changes may break existing applications that (incorrectly) expect unloaded attributes to return nil.

2.1.0

• Blueprint.validate now only accepts objects of the its type.
• Improved handling of Blueprints with circular relations.

2.0.0

• First pass at reworking model associations.
  • Split belongs_to into many_to_one and array_to_many associations
  • Added one_to_many and many_to_array associations (the inverse of the above associations, aka: has_many)
  • Added association configuration DSL to replace previous hash-based configuration.

0.3.1

• Added support for code coverage
• Added support for 'guard' gem (use 'bundle exec guard')
• Fixed bug in Praxis::Mapper::Model.undefine_data_accessors when method doesn't exist.
• Safer version checking with Gem::Version class
• Cleaned up Gemfile
• Updated 'slop' gem

0.3.0

• identity map hotfix

0.2.0

• don't know what happened here, check git log I guess

0.1.0

• initial release