Skip to content

snusnu/ramom

Repository files navigation

ramom

Gem Version Build Status Dependency Status Code Climate Coverage Status

Usage

Currently, ramom's focus is on reading database content. If you have defined your database with datamapper, ramom provides all the necessary integrations to get you going in no time.

Assuming that you have a set of datamapper models and a few records stored already

uri = 'postgres://localhost/test'.freeze

DataMapper.logger = DataMapper::Logger.new($stdout, :debug) if ::ENV['LOG']
DataMapper.setup(:default, uri)

class Account
  include DataMapper::Resource

  property :id,    Serial
  property :email, String, unique: true
end

class Person
  include DataMapper::Resource

  property :id,         Serial
  property :name,       String, unique_index: :test_compound_index
  property :nickname,   String, unique_index: :test_compound_index

  belongs_to :account
  has n, :tasks
end

class Task
  include DataMapper::Resource

  property :id,        Serial
  property :name,      String

  belongs_to :person
end

DataMapper.finalize.auto_migrate!

Person.create(
  name:     'snusnu',
  nickname: 'gams',
  account:  { email: '[email protected]' },
  tasks:    [{name: 'test'}]
)

You can then initialize a new Ramom::Schema ready to be queried by first inferring the information needed from DataMapper::Model.descendants and passing that as options to Ramom::Schema.define.

models        = DataMapper::Model.descendants
dm_definition = Ramom::Schema::Definition::Builder::DM.call(models)

options = {
  base:           dm_definition[:base_relations],
  fk_constraints: dm_definition[:fk_constraints]
}

ramom will automatically generate all base relations based on the information found in DataMapper::Model.descendants. Base relations correspond with tables in an SQL database, their equivalents in DataMapper code are therefore all descendants of DataMapper::Model. ramom infers all column names, their types and all the keys for every base relation. It also infers all foreign key constraints based on the belongs_to definitions in the DataMapper models.

Only querying raw tables doesn't help much tho, you need a way to write arbitrary queries, joining a few tables, restricting on some predicates and so on. ramom is based on relational algebra (by using axiom) internally, so from now on, we will use relational algebra terms instead of their SQL counterparts. Specifically, we'll say base relation instead of table, and virtual relation instead of query.

In contrast to most ORMs, ramom (currently) defines no API for ad-hoc queries, instead, you register any number of virtual relations with your schema. You can think of preregistered virtual relations much like methods that encapsulate a query. You can chain them arbitrarily, further composing more complex relations (by reusing simpler ones). Much like with regular methods, you typically don't want/need to expose all of them (i.e. make them public). ramom allows to define "public" relations using the external method inside a schema definition, and supports "private" relations using the internal method. While external relations will typically be called from client code (i.e. your application code), internal relations aren't exposed, but simply help with defining the external ones.

Just as regular ruby methods take parameters, virtual relations oftentimes need to accept parameters too, in order to be context aware. By context aware, we mean that oftentimes we need to restrict a relation based on, say, the currently logged in account_id. It's easy to do that with ramom, just make the block that defines a relation accept any parameters you like. You can still compose these relations arbitrarily, you just need to remember to pass all the relevant parameters to any relation invocation.

schema_definition = Ramom::Schema.define(options) do

  # These have been inferred from DM1 models
  # fk_constraint :people, :accounts, account_id: :account_id
  # fk_constraint :tasks,  :people,   person_id:  :person_id

  # Here we define a context aware relation that needs
  # an account_id in order to perform a restriction
  external :actors do |account_id|
    people.
      join(accounts.restrict(account_id: account_id)).
      wrap(account: [:account_id, :account_email])
  end

  external :person_details do |account_id|
    task_actors(account_id).
      group(tasks: [:task_id, :task_name])
  end

  external :task_details do |account_id|
    task_actors(account_id).
      wrap(person: [:person_id, :person_name, :account])
  end

  internal :task_actors do |account_id|
    tasks.join(actors(account_id))
  end

end

Once we have our relation schema set up, we can (optionally) define mappers (by using mom) in order to be able to access aptly named objects when iterating over the tuples returned from the database.

Mappers mainly act as a way to prettify access to the tuple's attributes, so for example, in order to be able to call actor.account.id instead of actor[:account_id] you tell ramom which mappers to use for your virtual relations.

definition_options  = Ramom::Schema::Mapping.default_options(schema_definition)
definition_registry = Mom::Definition::Registry.build(definition_options) do

  register :detailed_person, relation: :person_details, prefix: :person do
    map :id
    map :name

    wrap :account do
      map :id
      map :email
    end

    group :tasks do
      map :id
      map :name
    end
  end

  register :detailed_task, relation: :task_details, prefix: :task do
    map :id
    map :name

    wrap :person do
      map :id
      map :name

      wrap :account do
        map :id
        map :email
      end
    end
  end

  register :actor, prefix: :person do
    map :id
    map :name

    wrap :account do
      map :id
      map :email
    end
  end

end

Typically, you will need mappers for a few selected base relations too, but it'd be tedious to manually define them, also, ramom already has all the necessary information to generate them for you (it knows all base relations and their attributes). The following snippet shows how you can take advantage of that feature, and have ramom generate selected base relation mappers for you on the fly.

# This mutates +definition_registry+ and adds base relation mappers
Ramom::EntityBuilder.call(schema_definition, definition_registry) #, [
#
# Passing a whitelist of base relation names to generate
# mappers from is also supported. If no relation names
# are given, mappers for all base relations are generated.
#
# This is useful for automatically generating mappers
# that are used for mapping base relation tuples that
# get returned from successful write operations. Most
# of the time, not all base relations need respective
# mappers.
#
#  :people,
#  :tasks,
#])

Once we have defined all the mappers we're going to use in our app, we can finally connect all the pieces together. At this point you might have noticed that we never defined the actual objects our mappers are going to use. Until now, we've simply assumed that there will be, say, an Actor class that will expose all the relevant attributes.

This wasn't by mistake. ramom (in fact, mom) automatically generates these classes for you, as it has all the information it needs to do so, available from the mapper definitions. Since it is the author's belief that this kind of objects should never have any business logic methods attached, it's reasonably to generate them automatically, only exposing their attributes. Business logic can, and imo always should, be implemented somewhere else.

In order to tell mom exactly "how" to generate these classes, you must tell it which "model builder" it shall use. The only currently supported model builder is :anima, which builds models using the anima gem.

The first statement in the example code below, shows how to generate model classes from your mapper definitions.

models             = definition_registry.models(:anima)
entity_environment = definition_registry.environment(models)

The entity_environment created above can then be used to either automatically generate, or further refine the mapping of relation names to mapper names. By default, if you don't pass a block to Ramom::Mapping.new, it will simply assume that all your mappers are named in the singular form of your relations names. The block gets evaluated after inferring these mappings, so you can fine tune / overwrite any of the default mappings.

# The commented mappings are inferred automatically
mapping = Ramom::Mapping.new(entity_environment) # do
#  map :accounts,       :account
#  map :people,         :person
#  map :tasks,          :task
#  map :person_details, :detailed_person
#  map :task_details,   :detailed_task
#  map :actors,         :actor
#end

Now we're good to go! By passing a database adapter, a schema definition and a mapping to Ramom::Reader.build, we can finally start displaying our database content.

describe Ramom do

  adapter = Axiom::Adapter::DataObjects.new(uri)

  let(:db) { Ramom::Reader.build(adapter, schema_definition, mapping) }

  let(:account) { db.one(:accounts) }
  let(:person)  { db.one(:people) }
  let(:task)    { db.one(:tasks) }

  it 'provides access to base relations' do
    expect(account.id).to_not be(nil)
    expect(account.email).to eq('[email protected]')

    expect(person.id).to_not be(nil)
    expect(person.name).to eq('snusnu')
    expect(person.account_id).to eql(account.id)

    expect(db.read(:accounts).sort.one).to eql(account)
    expect(db.read(:people).sort.one).to eql(person)
  end

  it 'provides access to virtual relations' do
    a = db.one(:actors, 1)
    expect(a.id).to eq(person.id)
    expect(a.name).to eq(person.name)
    expect(a.account.id).to eq(account.id)
    expect(a.account.email).to eq(account.email)

    dt = db.one(:task_details, 1)
    expect(dt.id).to eq(task.id)
    expect(dt.name).to eq(task.name)
    expect(dt.person.id).to eq(person.id)
    expect(dt.person.name).to eq(person.name)
    expect(dt.person.account.id).to eql(account.id)
    expect(dt.person.account.email).to eql(account.email)

    dp = db.one(:person_details, 1)
    expect(dp.id).to eq(person.id)
    expect(dp.name).to eq(person.name)
    expect(dp.account.id).to eq(account.id)
    expect(dp.account.email).to eq(account.email)

    t = dp.tasks.first
    expect(t.id).to eq(task.id)
    expect(t.name).to eq(task.name)

    tuple  = db.schema.actors(1).sort.one
    mapper = db.mapping[:actors]

    expect(mapper.load(tuple)).to eql(a)

    expect {
      db.schema.task_actors(1).one
    }.to raise_error(NoMethodError)
  end

Credits

Contributing

See CONTRIBUTING.md for details.

Copyright

Copyright © 2014 Martin Gamsjaeger (snusnu). See LICENSE for details.

About

relational algebra meets object mapping

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages