Skip to content

CHANGELOG.md

Latest commit

 

History

History
188 lines (132 loc) · 10.9 KB

CHANGELOG.md

File metadata and controls

188 lines (132 loc) · 10.9 KB

Version 1.0.1

  • Fix bug that broke saving associated objects when declaring the id attribute with Virtus.

Version 1.0.0

  • Invoke identity map when using mapper.select and mapper.find with a block.
  • When finding objects with an array of ids, if they are all contained in the identity map, we don't issue a query.
  • Allow indexes on multiple attributes.
  • Add support to Perpetuity::RailsModel for ActiveModel's human class method.

Version 1.0.0.beta5

  • Don't stringify ids in IdentityMap. This should result in a little better performance. To be honest, I'm not even sure anymore why I put that in there.
  • Use the IdentityMap when retrieving multiple objects by id.

Version 1.0.0.beta4

  • Remove in-memory updating of objects using the Mapper#update method.
  • Separate dirty tracking from IdentityMap and return to a real Identity Map.
  • Pass IdentityMap around when loading associations
  • Add Perpetuity.register_adapter method to allow various data-source adapters to register without Perpetuity knowing about them.

Version 1.0.0.beta3

  • Fix title-case -> snake-case support to convert something like UserRegistration to user_registration. Previously, it would return userregistration.

Version 1.0.0.beta2

  • Allow data source to be configured with a URI. This lets you say Perpetuity.data_source "postgres://user:password@postgres.example.com:5432/my_database", keeping you from having to split up the URI yourself.
  • Remove overreaching rescue clauses to let exceptions get rescued inside the adapters
  • Allow AttributeSet to be initialized w/ attributes
  • Remove Mongo-specific code

Version 1.0.0.beta

  • Duplicate objects properly for the identity map. The identity map holds references to objects pulled out of the DB. We need to duplicate these references on insertion and retrieval from the identity map to make sure that modifications to an object do not pollute the version in the identity map.
  • Add support for the Amazon DynamoDB adapter by Cardspring. This adapter is still in very early development.
  • Allow passing in the type of the id attribute for databases that enforce column/field types. This allows us to say id(String) { ... }, telling the database that id needs to be a String.
  • Use hashes for AttributeSets to reduce serialization time.
  • Extract all database-specific logic to their own gems. The existing MongoDB adapter has been moved to the perpetuity-mongodb gem and there is a PostgreSQL adapter that lives under perpetuity-postgres.
  • Allow specs to be run on top of perpetuity-postgres
  • Remove validations.

Version 0.7.3

  • Only save attributes which have changed

Version 0.7.2

Version 0.7.1

  • Only unmarshal attributes that we marshaled to begin with. This disallows the use of false marshaled objects. — with Kevin Sjöberg
  • Allow insertion of multiple objects in Mapper#insert
  • Alias Retrieval#limit as Retrieval#take for Enumerable compatibility
  • Leave result cache when branching to new retrievals if previous retrieval had triggered a query
  • Silence warnings (some still exist in Moped, unfortunately)
  • Add finding based on attribute truthiness. For example: mapper.find { |obj| obj.name } finds objects whose name is neither nil nor false.
  • When you remove an index call from the mapper DSL, Mapper#reindex! now removes that index from the DB
  • Previously activated indexes in the DB are converted to Perpetuity::Attributes rather than stored in the format specific to the DB driver

Version 0.7.0

  • Add Perpetuity::RailsModel, an ActiveModel-compliant mixin
  • Check Rails app directory for mappers dynamically if the mapper registry is missing a particular model class

Version 0.6.2

  • Make sure Retrieval#first limits the query to a single item
  • Add Mapper#sample/Retrieval#sample
  • Add Retrieval#drop to allow skipping a specific number of objects

Version 0.6.1

  • Fix reference equality bug

Version 0.6.0

  • Stop extending persisted/retrieved objects with Perpetuity::PersistedObject. Extending these objects invalidates method caches inside the VM, which degrades performance.
  • Cache all objects returned from Mapper#find(id) by default into the mapper object. This can be bypassed with Mapper#find(id, false)
  • Do not store uninitialized attributes with the MongoDB serializer. This shrinks the size of documents when possible.

Version 0.5.0

  • Allow querying based on referenced class/id or embedded-object data
  • Remove duplicate references when loading associations on referenced objects
  • Optimize loading associations with zero/one/many objects. It is similar to Array#detect vs Array#select. detect is faster if you only need one.
  • Fixed a bug in defining methods on generated mapper classes which would mistakenly define them on the class instead of the mapper objects
  • Add none?/one?/all?/any? methods to mappers.
  • Add block functionality to Mapper#count, similar to Enumerable#count
  • Alias Mapper#find_all to Mapper#select
  • Alias Mapper#detect to Mapper#find
  • Add Mapper#reject method to negate queries
  • Allow Mapper#find to take a block like Mapper#select
  • Add atomic incrementation

Version 0.4.8

  • Provide configuration one-liner ability for simple configs
  • Memoize results for Retrieval objects
  • Retrieval#count sends a count query rather than retrieving the result set and returning the count of the array
  • Allow updating of objects that contain Perpetuity::References to other persisted objects without first loading the association
  • Fix bug preventing objects with an array of referenced objects from being updated
  • Move serialization into the MongoDB adapter, allowing future adapters to implement their own serializations

Version 0.4.7

  • Use instance variables rather than attr_readers in IdentityMap (avoids calling methods during load_association!).
  • Fixed a bug with MongoDB authentication.
  • Use attribute names rather than a stringified attribute for MongoDB indexes.
  • Fixed a bug with deserializing attributes of namespaced classes. — Jaap Groeneveld
  • Fixed serialization of Perpetuity::Reference attributes caused by resaving an object that did not have all of its referenced associations loaded.

Version 0.4.6

  • Fix an inconsistency between mongo and moped gems. Previously, the MongoDB adapter relied on the fail-fast behavior of the mongo gem so we had no need to check for errors. Moped does not do that by default, which resulted in incorrect behavior of the gem when an object was not persisted.
  • The Mapper#select DSL now allows the use of != as criteria. Not all supported Ruby implementations allowed that operator to be overridden when the DSL was created, but they do now.

Version 0.4.5

  • Move from the mongo gem from 10gen to the moped gem for talking to MongoDB. This resulted in performance gains of 30-80%, depending on query size.
  • Make persisted objects marshalable.
    • Previously, the id method was defined on individual objects that were either persisted to or retrieved from the DB. This made them unable to be marshaled with Marshal.dump.
    • Now we extend the objects with Perpetuity::PersistedObject to keep them marshalable while still providing id. This keeps them marshalable (unmarshalled objects will still be extended with Perpetuity::PersistedObject).
  • Provided a partial fix for a bug that kept us from being able to persist hash attributes properly. See the first comments in the linked GitHub issue for an explanation of why it is only a partial fix.
  • Stopped testing on MRI 1.9.2 with Travis CI. Moped requires 1.9.3 or higher.

Version 0.4.4

  • Automatically persist all referenced objects if they are not already persisted. Previously, referenced objects were required to be persisted before persisting the referencing object.

Version 0.4.3

  • Made Mapper#load_association! more friendly. It now loads the associated objects for all objects passed in and works with arrays of referenced objects.
  • Mapper#load_association! is also now more efficient — the N+1 queries have been optimized. The number of queries it fires off is now equal to the quantity of different classes of associated objects.
    • For example, if a user can have either a UserProfile or an AdminProfile as its profile attribute, Perpetuity will use two queries in #load_association! if and only if both types of profiles are used.
    • In that example, if you only query users with a UserProfile, only one DB query will be triggered.

New query example:

user_mapper = Perpetuity[User]
users = user_mapper.all.to_a
user_mapper.load_association! users, :profile

Each of the users in the DB will have their profiles loaded with a single DB query per profile type and stored in their profile attributes.

Version 0.4.2

  • Improved speed and stability of Mapper#first and Mapper#all

Version 0.4.1

  • Add Mapper#remove_index! to remove DB indexes
  • Fix index declarations in mappers to take unique/order options
    • These were intended to be there from the beginning, but the API wasn't made available in the mapper

Version 0.4

  • Mapper select DSL now more closely resembles Enumerable syntax
    • Instead of mapper.select { published_at < Time.now } you would write mapper.select { |article| article.published_at < Time.now }
    • It's a little more typing, but it eliminates a lot of problems that came from calling instance_exec on the block. It meant we couldn't call methods on the current object (they would be invoked on the Query object) or use instance variables from the current object.
    • Explained more in issue #19

Version 0.3.1

  • Access object state via instance methods to store actual object state, rather than treating all objects as value objects
  • Use Object.allocate rather than Object.new when restoring state to an object from the database

Version 0.3

  • Use Perpetuity[] instead of Perpetuity::Mapper[] to get mapper instances
  • Allow subclassing of Perpetuity::Mapper with map macro
  • Fix select calls using id as criteria from returning invalid or missing data
  • Make MongoDB reads and writes fail fast
  • Add indexing API to mappers
  • Raise an exception when calling Perpetuity[] with a class that has no mapper
  • Add unions and intersections to select queries for MongoDB adapter
    • This allows for queries like Perpetuity[Article].select { (created_at < Time.now) & (published == true) }
  • Update object in memory when calling Mapper#update
  • Add Mapper#save to update an object's current state in the DB. Previously, we only had #update (primarily for update actions in Rails controllers).
  • Silenced some ugly warnings

Version 0.2 and 0.1 have no changelog because I am a terrible developer