Home

Awesome

AttrJson

CI Status CI Status Gem Version

ActiveRecord attributes stored serialized in a json column, super smooth. For Rails 6.0.x through 7.0.x. Ruby 2.7+.

Typed and cast like Active Record. Supporting nested models, dirty tracking, some querying (with postgres jsonb contains), and working smoothy with form builders.

Use your database as a typed object store via ActiveRecord, in the same models right next to ordinary ActiveRecord column-backed attributes and associations. Your json-serialized attr_json attributes use as much of the existing ActiveRecord architecture as we can.

Why might you want or not want this?

Developed for postgres, but most features should work with MySQL json columns too, although has not yet been tested with MySQL.

Basic Use

# migration, default column used is `json_attributes, but this can be changed
class CreatMyModels < ActiveRecord::Migration[5.0]
  def change
    create_table :my_models do |t|
      t.jsonb :json_attributes
    end

    # If you plan to do any querying with jsonb_contains below..
    add_index :my_models, :json_attributes, using: :gin
  end
end

# An embedded model, if desired
class LangAndValue
  include AttrJson::Model

  attr_json :lang, :string, default: "en"
  attr_json :value, :string
end

class MyModel < ActiveRecord::Base
   include AttrJson::Record

   # use any ActiveModel::Type types: string, integer, decimal (BigDecimal),
   # float, datetime, boolean.
   attr_json :my_string, :string
   attr_json :my_integer, :integer
   attr_json :my_datetime, :datetime

   # You can have an _array_ of those things too. It will ordinarily default to empty array.
   attr_json :int_array, :integer, array: true

   # The empty array default can be disabled with the following setting
   attr_json :str_array, :string, array: true, default: AttrJson::AttributeDefinition::NO_DEFAULT_PROVIDED

   #and/or defaults
   attr_json :str_with_default, :string, default: "default value"

   attr_json :embedded_lang_and_val, LangAndValue.to_type
end

model = MyModel.create!(
  my_integer: 101,
  my_datetime: DateTime.new(2001,2,3,4,5,6),
  embedded_lang_and_val: LangAndValue.new(value: "a sentance in default language english")
  )

What will get serialized to your json_attributes column will look like:

{
  "my_integer":101,
  "my_datetime":"2001-02-03T04:05:06Z",
  "str_with_default":"default value",
  "embedded_lang_and_val": {
    "lang":"en",
    "value":"a sentance in default language english"
  }
}

These attributes have type-casting behavior very much like ordinary ActiveRecord values.

model = MyModel.new
model.my_integer = "12"
model.my_integer # => 12
model.int_array = "12"
model.int_array # => [12]
model.my_datetime = "2016-01-01 17:45"
model.my_datetime # => a Time object representing that, just like AR would cast
model.embedded_lang_and_val = { value: "val"}
model.embedded_lang_and_val #=> #<LangAndVal:0x000000010c9a7ad8 @attributes={"value"=>"val"...>

You can use ordinary ActiveRecord validation methods with attr_json attributes.

All the attr_json attributes are serialized to json as keys in a hash, in a database jsonb/json column. By default, in a column json_attributes. If you look at model.json_attributes, you'll see values already cast to their ruby representations.

To see JSON representations, we can use Rails *_before_type_cast methods, *-in_database and [*_for_database] methods (Rails 7.0+ only).

These methods can all be called on the container json_attributes json hash attribute (generally showing serialized JSON to string), or any individual attribute (generally showing in-memory JSON-able object). [This is a bit confusing and possibly not entirely consistent, needs more investigation.]

Specifying db column to use

While the default is to assume you want to serialize in a column called json_attributes, no worries, of course you can pick whatever named jsonb column you like, class-wide or per-attribute.

class OtherModel < ActiveRecord::Base
  include AttrJson::Record

  # as a default for the model
  attr_json_config(default_container_attribute: :some_other_column_name)

  # now this is going to serialize to column 'some_other_column_name'
  attr_json :my_int, :integer

  # Or on a per-attribute basis
  attr_json :my_int, :integer, container_attribute: "yet_another_column_name"
end

Store key different than attribute name/methods

You can also specify that the serialized JSON key should be different than the attribute name/methods, by using the store_key argument.

class MyModel < ActiveRecord::Base
  include AttrJson::Record

  attr_json :special_string, :string, store_key: "__my_string"
end

model = MyModel.new
model.special_string = "foo"
model.json_attributes # => {"__my_string"=>"foo"}
model.save!
model.json_attributes_before_type_cast # => string containing: {"__my_string":"foo"}

You can of course combine array, default, store_key, and container_attribute params however you like, with whatever types you like: symbols resolvable with ActiveRecord::Type.lookup, or any ActiveModel::Type::Value subclass, built-in or custom.

You can register your custom ActiveModel::Type::Value in a Rails initializer or early on in your app boot sequence:

ActiveRecord::Type.register(:my_type, MyActiveModelTypeSubclass)

<a name="querying"></a>

Querying

There is some built-in support for querying using postgres jsonb containment (@>) operator. (or see here or here). For now you need to additionally include AttrJson::Record::QueryScopes to get this behavior.

model = MyModel.create(my_string: "foo", my_integer: 100)

MyModel.jsonb_contains(my_string: "foo", my_integer: 100).to_sql
# SELECT "products".* FROM "products" WHERE (products.json_attributes @> ('{"my_string":"foo","my_integer":100}')::jsonb)
MyModel.jsonb_contains(my_string: "foo", my_integer: 100).first
# Implemented with scopes, this is an ordinary relation, you can
# combine it with whatever, just like ordinary `where`.

MyModel.not_jsonb_contains(my:string: "foo", my_integer: 100).to_sql
# SELECT "products".* FROM "products" WHERE NOT (products.json_attributes @> ('{"my_string":"foo","my_integer":100}')::jsonb)

# typecasts much like ActiveRecord on query too:
MyModel.jsonb_contains(my_string: "foo", my_integer: "100")
# no problem

# works for arrays too
model = MyModel.create(int_array: [10, 20, 30])
MyModel.jsonb_contains(int_array: 10) # finds it
MyModel.jsonb_contains(int_array: [10]) # still finds it
MyModel.jsonb_contains(int_array: [10, 20]) # it contains both, so still finds it
MyModel.jsonb_contains(int_array: [10, 1000]) # nope, returns nil, has to contain ALL listed in query for array args

jsonb_contains will handle any store_key you have set -- you should specify attribute name, it'll actually query on store_key. And properly handles any container_attribute -- it'll look in the proper jsonb column.

Anything you can do with jsonb_contains should be handled by a postgres USING GIN index. Figuring out how to use indexes for jsonb queries can be confusing, here is a good blog post.

<a name="nested"></a>

Nested models -- Structured/compound data

The AttrJson::Model mix-in lets you make ActiveModel::Model objects that can be round-trip serialized to a json hash, and they can be used as types for your top-level AttrJson::Record. AttrJson::Models can contain other AJ::Models, singly or as arrays, nested as many levels as you like.

That is, you can serialize complex object-oriented graphs of models into a single jsonb column, and get them back as they went in.

AttrJson::Model has an identical attr_json api to AttrJson::Record, with the exception that container_attribute is not supported.

class LangAndValue
  include AttrJson::Model

  attr_json :lang, :string, default: "en"
  attr_json :value, :string

  # Validations work fine, and will post up to parent record
  validates :lang, inclusion_in: I18n.config.available_locales.collect(&:to_s)
end

class MyModel < ActiveRecord::Base
  include AttrJson::Record
  include AttrJson::Record::QueryScopes

  attr_json :lang_and_value, LangAndValue.to_type

  # YES, you can even have an array of them
  attr_json :lang_and_value_array, LangAndValue.to_type, array: true
end

# Set with a model object, in initializer or writer
m = MyModel.new(lang_and_value: LangAndValue.new(lang: "fr", value: "S'il vous plaît"))
m.lang_and_value = LangAndValue.new(lang: "es", value: "hola")
m.lang_and_value
# => #<LangAndValue:0x007fb64f12bb70 @attributes={"lang"=>"es", "value"=>"hola"}>
m.save!
m.attr_jsons_before_type_cast
# => string containing: {"lang_and_value":{"lang":"es","value":"hola"}}

# Or with a hash, no problem.

m = MyModel.new(lang_and_value: { lang: 'fr', value: "S'il vous plaît"})
m.lang_and_value = { lang: 'en', value: "Hey there" }
m.save!
m.attr_jsons_before_type_cast
# => string containing: {"lang_and_value":{"lang":"en","value":"Hey there"}}
found = MyModel.find(m.id)
m.lang_and_value
# => #<LangAndValue:0x007fb64eb78e58 @attributes={"lang"=>"en", "value"=>"Hey there"}>

# Arrays too, yup

m = MyModel.new(lang_and_value_array: [{ lang: 'fr', value: "S'il vous plaît"}, { lang: 'en', value: "Hey there" }])
m.lang_and_value_array
# => [#<LangAndValue:0x007f89b4f08f30 @attributes={"lang"=>"fr", "value"=>"S'il vous plaît"}>, #<LangAndValue:0x007f89b4f086e8 @attributes={"lang"=>"en", "value"=>"Hey there"}>]
m.save!
m.attr_jsons_before_type_cast
# => string containing: {"lang_and_value_array":[{"lang":"fr","value":"S'il vous plaît"},{"lang":"en","value":"Hey there"}]}

You can nest AttrJson::Model objects inside each other, as deeply as you like.

You can edit nested models "in place", they will be properly saved.

m.lang_and_value.lang = "de"
m.save! # no problem!

For use with Rails forms, you may want to use attr_json_accepts_nested_attributes_for (like Rails accepts_nested_attributes_for, see doc page on Use with Forms and Form Builders.

Model-type defaults

If you want to set a default for an AttrJson::Model type, you should use a proc argument for the default, to avoid accidentally re-using a shared global default value, similar to issues people have with ruby Hash default.

  attr_json :lang_and_value, LangAndValue.to_type, default: -> { LangAndValue.new(lang: "en", value: "default") }

You can also use a Hash value that will be cast to your model, no need for proc argument in this case.

  attr_json :lang_and_value, LangAndValue.to_type, default: { lang: "en", value: "default" }

Polymorphic model types

There is some support for "polymorphic" attributes that can hetereogenously contain instances of different AttrJson::Model classes, see comment docs at AttrJson::Type::PolymorphicModel.

class SomeLabels
  include AttrJson::Model

  attr_json :hello, LangAndValue.to_type, array: true
  attr_json :goodbye, LangAndValue.to_type, array: true
end
class MyModel < ActiveRecord::Base
  include AttrJson::Record
  include AttrJson::Record::QueryScopes

  attr_json :my_labels, SomeLabels.to_type
end

m = MyModel.new
m.my_labels = {}
m.my_labels
# => #<SomeLabels:0x007fed2a3b1a18>
m.my_labels.hello = [{lang: 'en', value: 'hello'}, {lang: 'es', value: 'hola'}]
m.my_labels
# => #<SomeLabels:0x007fed2a3b1a18 @attributes={"hello"=>[#<LangAndValue:0x007fed2a0eafc8 @attributes={"lang"=>"en", "value"=>"hello"}>, #<LangAndValue:0x007fed2a0bb4d0 @attributes={"lang"=>"es", "value"=>"hola"}>]}>
m.my_labels.hello.find { |l| l.lang == "en" }.value = "Howdy"
m.save!
m.attr_jsons
# => {"my_labels"=>#<SomeLabels:0x007fed2a714e80 @attributes={"hello"=>[#<LangAndValue:0x007fed2a714cf0 @attributes={"lang"=>"en", "value"=>"Howdy"}>, #<LangAndValue:0x007fed2a714ac0 @attributes={"lang"=>"es", "value"=>"hola"}>]}>}
m.attr_jsons_before_type_cast
# => string containing: {"my_labels":{"hello":[{"lang":"en","value":"Howdy"},{"lang":"es","value":"hola"}]}}

GUESS WHAT? You can QUERY nested structures with jsonb_contains, using a dot-keypath notation, even through arrays as in this case. Your specific defined attr_json types determine the query and type-casting.

MyModel.jsonb_contains("my_labels.hello.lang" => "en").to_sql
# => SELECT "products".* FROM "products" WHERE (products.json_attributes @> ('{"my_labels":{"hello":[{"lang":"en"}]}}')::jsonb)
MyModel.jsonb_contains("my_labels.hello.lang" => "en").first


# also can give hashes, at any level, or models themselves. They will
# be cast. Trying to make everything super consistent with no surprises.

MyModel.jsonb_contains("my_labels.hello" => LangAndValue.new(lang: 'en')).to_sql
# => SELECT "products".* FROM "products" WHERE (products.json_attributes @> ('{"my_labels":{"hello":[{"lang":"en"}]}}')::jsonb)

MyModel.jsonb_contains("my_labels.hello" => {"lang" => "en"}).to_sql
# => SELECT "products".* FROM "products" WHERE (products.json_attributes @> ('{"my_labels":{"hello":[{"lang":"en"}]}}')::jsonb)

Remember, we're using a postgres containment (@>) operator, so queries always mean 'contains' -- the previous query needs a my_labels.hello which is a hash that includes the key/value, lang: en, it can have other key/values in it too. String values will need to match exactly.

Single AttrJson::Model serialized to an entire json column

The main use case of the gem is set up to let you combine multiple primitives and nested models under different keys combined in a single json or jsonb column.

But you may also want to have one AttrJson::Model class that serializes to map one model class, as a hash, to an entire json column on it's own.

AttrJson::Model can supply a simple coder for the ActiveRecord serialization feature to easily do that.

class MyModel
  include AttrJson::Model

  attr_json :some_string, :string
  attr_json :some_int, :int
end

class MyTable < ApplicationRecord
  serialize :some_json_column, MyModel.to_serialization_coder

  # NOTE: In Rails 7.1+, write:
  # serialize :some_json_column, coder: MyModel.to_serialization_coder
end

MyTable.create(some_json_column: MyModel.new(some_string: "string"))

# will cast from hash for you
MyTable.create(some_json_column: { some_int: 12 })

# etc

To avoid errors raised at inconvenient times, we recommend you set these settings to make 'bad' data turn into nil, consistent with most ActiveRecord types:

class MyModel
  include AttrJson::Model

  attr_json_config(bad_cast: :as_nil, unknown_key: :strip)
  # ...
end

And/or define a setter method to cast, and raise early on data problems:

class MyTable < ApplicationRecord
  serialize :some_json_column, MyModel.to_serialization_coder

  def some_json_column=(val)
    super(   )
  end
end

Serializing a model to an entire json column is a relatively recent feature, please let us know how it's working for you.

<a name="arbitrary-json-data"></a>

Storing Arbitrary JSON data

Arbitrary JSON data (hashes, arrays, primitives of any depth) can be stored within attributes by using the rails built in ActiveModel::Type::Value as the attribute type. This is basically a "no-op" value type -- JSON alone will be used to serialize/deserialize whatever values you put there, because of the json type on the container field.

class MyModel < ActiveRecord::Base
  include AttrJson::Record

  attr_json :arbitrary_hash, ActiveModel::Type::Value.new
end

<a name="forms"></a>

Forms and Form Builders

Use with Rails form builders is supported pretty painlessly. Including with simple_form and cocoon (integration-tested in CI).

If you have nested AttrJson::Models you'd like to use in your forms much like Rails associated records: Where you would use Rails accepts_nested_attributes_for, instead include AttrJson::NestedAttributes and use attr_json_accepts_nested_attributes_for. Multiple levels of nesting are supported.

For more info, see doc page on Use with Forms and Form Builders.

<a name="ar_attributes"></a>

ActiveRecord Attributes and Dirty tracking

We endeavor to make record-level attr_json attributes available as standard ActiveRecord attributes, supporting that full API.

Standard Rails dirty tracking should work properly with AttrJson::Record attributes! We have a test suite demonstrating.

We actually keep the "canonical" copy of data inside the "container attribute" hash in the ActiveRecord model. This is because this is what will actually get saved when you save. So we have two copies, that we do our best to keep in sync.

They get out of sync if you are doing unusual things like using the ActiveRecord attribute API directly (like calling write_attribute with an attr_json attribute). Even if this happens, mostly you won't notice. But one thing it will effect is dirty tracking.

If you ever need to sync the ActiveRecord attribute values from the AttrJson "canonical" copies, you can call active_record_model.attr_json_sync_to_rails_attributes. If you wanted to be 100% sure of dirty tracking, I suppose you could always call this method first. Sorry, this is the best we could do!

Note that ActiveRecord DirtyTracking will give you ruby objects, for instance for nested models, you might get:

record_obj.attribute_change_to_be_saved(:nested_model)
# => [#<object>, #<object>]

If you want to see JSON instead, you could call #as_json on the values. The Rails *_before_type_cast and *-in_database methods may also be useful.

<a name="why"></a>

Do you want this?

Why might you want this?

Why might you not want this?

Note on Optimistic Locking

When you save a record with any changes to any attr_jsons, it will overwrite the whole json structure in the relevant column for that row. Unlike ordinary AR attributes where updates just touch changed attributes.

Becuase of this, you probably want to seriously consider using ActiveRecord Optimistic Locking to prevent overwriting other updates from processes.

State of Code, and To Be Done

This code is solid and stable and is being used in production. If you don't see a lot of activity, it might be because it's stable, rather than abandoned. Check to see if it's passing/supported on recent Rails? We test on "edge" unreleased rails to try to stay ahead of compatibility, and has worked through multiple major Rails verisons with few if any changes needed.

In order to keep the low-maintenace scenario sustainable, I am very cautious accepting new features, especially if they increase code complexity at all. Even if you have a working PR, I may be reluctant to accept it. I'm prioritizing sustainability and stability over new features, and so far this is working out well. However, discussion is always welcome! Especially when paired with code (failing tests for the bugfix or feature you want are super helpful on their own!).

We are committed to semantic versioning and will endeavor to release no backwards breaking changes without a major version. We are also serious about minimizing backwards incompat releases altogether (ie minimiing major version releases).

Feedback of any kind of very welcome, please feel free to use the issue tracker. It is hard to get a sense of how many people are actually using this, which is helpful both for my own sense of reward and for anyone to get a sense of the size of the userbase -- feel free to say hi and let us know how you are using it!

Except for the jsonb_contains stuff using postgres jsonb contains operator, I don't believe any postgres-specific features are used. It ought to work with MySQL, testing and feedback welcome. (Or a PR to test on MySQL?). My own interest is postgres.

This is still mostly a single-maintainer operation, so has all the sustainability risks of that. Although there are other people using and contributing to it, check out the Github Issues and Pull Request tabs yourself to get a sense.

Possible future features:

Development

While attr_json depends only on active_record, we run integration tests in the context of a full Rails app, in order to test working with simple_form and cocoon, among other things. (Via combustion, with app skeleton at ./spec/internal).

At present this does mean that all our automated tests are run in a full Rails environment, which is not great (any suggestions or PR's to fix this while still running integration tests under CI with full Rails app).

Tests are in rspec, run tests simply with ./bin/rspec.

We use appraisal to test with multiple rails versions, including on travis. Locally you can run bundle exec appraisal rspec to run tests multiple times for each rails version, or eg bundle exec appraisal rails-5-1 rspec. If the Gemfile or Appraisal file changes, you may need to re-run bundle exec appraisal install and commit changes. (Try to put dev dependencies in gemspec instead of Gemfile, but sometimes it gets weird.)

There is a ./bin/console that will give you a console in the context of attr_json and all it's dependencies, including the combustion rails app, and the models defined there.

Acknowledements, Prior Art, alternatives