The name `ActiveModel::AttributeAssignment::UnknownAttributeError` is
too implementation specific so let's move the constant directly under
the ActiveModel namespace.
Also since this constant used to be under the ActiveRecord namespace, to
make the upgrade path easier, let's avoid raising the former constant
when we deal with this error on the Active Record side.
Previously, calling `User.model_name.to_json` would result in an infinite
recursion as `.model_name` inherited its `.as_json` behavior from Object. This
patch fixes that unexpected behavior by delegating `.as_json` to :name.
To be able to return type of validator, one can now call `details`
on Errors instance:
```ruby
class User < ActiveRecord::Base
validates :name, presence: true
end
```
```ruby
user = User.new; user.valid?; user.errors.details
=> {name: [{error: :blank}]}
```
`before_validation` and `after_validation` from
ActiveModel::Validation::Callbacks accept an optional `:on` parameter
that was not previously documented or tested. For instance given
before_validation :do_something, on: :create
then `object.valid?(:create)` will invoke `:do_something` while
`object.valid?` or `object.valid?(:anything_else)` will not.
Before this commit, returning `false` in an ActiveModel `before_` callback
such as `before_create` would halt the callback chain.
After this commit, the behavior is deprecated: will still work until
the next release of Rails but will also display a deprecation warning.
The preferred way to halt a callback chain is to explicitly `throw(:abort)`.
Before this commit, returning `false` in an ActiveModel validation
callback such as `before_validation` would halt the callback chain.
After this commit, the behavior is deprecated: will still work until
the next release of Rails but will also display a deprecation warning.
The preferred way to halt a callback chain is to explicitly `throw(:abort)`.
This stems from https://github.com/rails/rails/pull/17227#discussion_r21641358
It's simply a clarification of the current behavior by which if an
`after_` or `around_` ActiveModel callback returns +false+, then the callback
chain **is not halted**.
The callback chain in ActiveModel is only halted when a `before_`
callback returns `false`.
This stems from https://github.com/rails/rails/pull/17227#discussion_r21641358
It's simply a clarification of the current behavior by which if an
`after_validation` ActiveModel callback returns +false+, then further
`after_` callbacks **are not halted**.
I'm not sure what's the use case for this, but apparently it broke some apps.
Since it was not the intended result from #16210 I fixed it to not raise an
exception anymore. However, I didn't add documentation for it because I don't
know if this should be officially supported without knowing how it's meant to
be used.
In general, validations should be side-effect-free (other than adding to the
error message to `@errors`). Order-dependent validations seems like a bad idea.
Fixes#18002
Goals:
1. Default to :random for newly generated applications
2. Default to :sorted for existing applications with a warning
3. Only show the warning once
4. Only show the warning if the app actually uses AS::TestCase
Fixes#16769
Calling ActiveSupport::TestCase.i_suck_and_my_tests_are_order_dependent! in AS::TestCase makes
everyone's tests order dependent, which should never be done by the framework.
We are planning to remove mocha from our test suite because of
performance problems. To make this possible we should stop require mocha
on ActionSupport::TestCase.
This should not affect applications since users still need to add mocha
to Gemfile and this already load mocha.
Added FIXME notes to place that still need mocha removal
This change prevents a certain class of user error which results when
mistakenly using the `validate` class method instead of the `validates`
class method.
Only apply when all arguments are symbols, because some validations use
the `validate` method and pass in additional options, namely the
`LenghValidator` via the `ActiveMode::Validations::validates_with`
method.
These methods may cause confusion with the `reset_changes` that
behaves differently
of them.
Also rename undo_changes to restore_changes to match this new set of
methods.
This method name is causing confusion with the `reset_#{attribute}`
methods. While `reset_name` set the value of the name attribute for the
previous value the `reset_changes` only discard the changes and previous
changes.
It's unintuitive to call '#valid?' when you want to run validations but
don't care about the return value.
The alias in ActiveRecord isn't strictly necessary (the ActiveModel
alias is still in effect), but it clarifies.
This gets the whole ActiveModel test suites working even if
`self.i_suck_and_my_tests_are_order_dependent!` is disabled
in `ActiveSupport::TestCase`.
Two places are found that potentially leak global state. This patch
makes sure states are restored so that none of the changes happen in
a single test will be carried over to subsequence tests.
Now that Validator #setup is called from the initializer, we need a
reference to the model's class to be passed in to allow the validators
to continue functioning when used at the instance level.
Closes#14134.
The documentation of `#to_key` (http://api.rubyonrails.org/classes/ActiveModel/Conversion.html#method-i-to_key)
states that it returns `nil` if there are no key attributes. `to_param` needs
to be aware of that fact and return `nil` as well.
Previously it raised the following exception:
```
1) Error:
ConversionTest#test_to_param_returns_nil_if_to_key_is_nil:
NoMethodError: undefined method `join' for nil:NilClass
/Users/senny/Projects/rails/activemodel/lib/active_model/conversion.rb:65:in `to_param'
/Users/senny/Projects/rails/activemodel/test/cases/conversion_test.rb:34:in `block in <class:ConversionTest>'
```
Before this patch after the changes are applied the changes can be only
accessed using string keys, but before symbols are also accepted.
After this change every state of the model will be consistent.
of an instance. When using the attribute_will_change! method, you must
supply a string and not a symbol or the *_changed? method will break
(because it is looking for the attribute name as a string in the keys
of the underlying hash). To remedy this, I simply made the underlying
hash a HashWithIndifferentAccess so it won't matter if you supply
the attribute name as a symbol or string to attribute_will_change!.
+ Namespace changes, overhaul of runners.
+ Internal ivar name changes
- Removed a logger globally applied to tests that spew everywhere?!?
+ Override Minitest#__run to sort tests by name.
+ Reworked testing isolation to work with the new cleaner architecture.
- Removed a bunch of tests that just test minitest straight up. I think these changes were all merged to minitest 4 a long time ago.
- Minor report output differences.
Closes#9535.
With 692b3b6 the `password=` setter does no longer set blank passwords.
This triggered validation errors when assigning empty Strings to `password`
and `password_confirmation`.
This patch only sets the confirmation if it is not `blank?`.
with the current tests, if delete the assignment of is_empty in add_on_empty method
the tests not fail. With this test, if we delete is_empty, the test fails
* Clarifying what the #changed method returns
* Adding tests to describe what the #changed_attributes returns
* Updating test name based on pull request comment
* Moving the test lower in the file per pull request comment
* Verifying persisted? is false when Model initialized
* Passing nil to Model should not blow up
* Changing test name to the one suggested in pull request comment
When resetting an attribute, you expect it to return to the state it was
before any changes. Namely, this fixes this unexpected behavior:
~~~ruby
model.name = "Bob"
model.reset_name!
model.name_changed? #=> true
~~~
When serialising a class, specify the type of any singular associations, if
necessary. Rails already correctly specifies the :type of any enumerable
association (e.g. a has_many association), but made no attempt to do so for
non-enumerables (e.g. a has_one association).
We must specify the :type of any STI association. A has_one
association to a class which uses single-table inheritance is an example of
this type of association.
Fixes#7471
Log output from activemodel's railtie_test directly to STDOUT.
There's no logging going on here, but since we initialize the app, the
logger is set and the folder is automatically created. With this change,
the default logger is not created, so there is no logging folder anymore.
Conflicts:
activemodel/test/cases/railtie_test.rb
This reverts commit 4e9f53f973, reversing
changes made to 6b802cdb4f.
Revert "Don't use tap in this case."
This reverts commit 454d820bf0.
Reason: Is not a good idea to add options to this method since we can do
the same thing using method composition.
Person.validators_on(:name).select { |v| v.kind == :presence }
Also it avoids to change the method again to add more options.
Due to a change in builder, nil values and empty strings now generates
closed tags, so instead of this:
<pseudonyms nil=\"true\"></pseudonyms>
It generates this:
<pseudonyms nil=\"true\"/>
Document this change in Rails so that people can track it down easily if
necessary.
- Enable propagation of :skip_types, :dasherize and :camelize on included models by default
- Adding the option to override this propagation on a per-include basis (:include => { :model => { :dasherize => false } }
- Enough tests to prove it works
- Updated activemodel CHANGELOG.md
Squashed my commits
ActiveRecord json/xml serialization should use as base
serializable_hash, provided by ActiveModel. Add some more coverage
around options :only and :except for both json and xml serialization.
Changes:
* Update `include_root_in_json` default value to false for default value
to false for `ActiveModel::Serializers::JSON`.
* Remove unnecessary change to include_root_in_json option in
wrap_parameters template.
* Update `as_json` documentation.
* Fix JSONSerialization tests.
Problem:
It's confusing that AM serializers behave differently from AR,
even when AR objects include AM serializers module.
class User < ActiveRecord::Base; end
class Person
include ActiveModel::Model
include ActiveModel::AttributeMethods
include ActiveModel::Serializers::JSON
attr_accessor :name, :age
def attributes
instance_values
end
end
user.as_json
=> {"id"=>1, "name"=>"Konata Izumi", "age"=>16, "awesome"=>true}
# root is not included
person.as_json
=> {"person"=>{"name"=>"Francesco", "age"=>22}}
# root is included
ActiveRecord::Base.include_root_in_json
=> false
Person.include_root_in_json
=> true
# different default values for include_root_in_json
Proposal:
Change the default value of AM serializers to false, update
the misleading documentation and remove unnecessary change
to false of include_root_in_json option with AR objects.
class User < ActiveRecord::Base; end
class Person
include ActiveModel::Model
include ActiveModel::AttributeMethods
include ActiveModel::Serializers::JSON
attr_accessor :name, :age
def attributes
instance_values
end
end
user.as_json
=> {"id"=>1, "name"=>"Konata Izumi", "age"=>16, "awesome"=>true}
# root is not included
person.as_json
=> {"name"=>"Francesco", "age"=>22}
# root is not included
ActiveRecord::Base.include_root_in_json
=> false
Person.include_root_in_json
=> false
# same behaviour, more consistent
Fixes#6578.
Passing a falsey option value for a validator currently causes that validator to
be enabled, just like "true":
ActiveModel.validates :foo, :presence => false
This is rather counterintuitive, and makes it inconvenient to wrap `validates` in
methods which may conditionally enable different validators.
As an example, one is currently forced to write:
def has_slug(source_field, options={:unique => true})
slugger = Proc.new { |r| r[:slug] = self.class.sluggify(r[source_field]) if r[:slug].blank? }
before_validation slugger
validations = { :presence => true, :slug => true }
if options[:unique]
validations[:uniqueness] = true
end
validates :slug, validations
end
because the following reasonable-looking alternative fails to work as expected:
def has_slug(source_field, options={:unique => true})
slugger = Proc.new { |r| r[:slug] = self.class.sluggify(r[source_field]) if r[:slug].blank? }
before_validation slugger
validates :slug, :presence => true, :slug => true, :uniqueness => options[:unique]
end
(This commit includes a test, and all activemodel and activerecord tests pass as before.)
XmlMini define the xml 'datatime', but according to
http://www.w3.org/TR/xmlschema-2/#dateTime could be better
change this to 'dateTime' with upper case letter 'T.
So 'DateTime' and 'Time' are redefined from 'datetime' to 'dateTime'
add the changing to the changelog
When a model nests another model that also nests yet another model
using accepts_nested_attributes_for method, its Errors object can
have an attribute name with "contacts.addresses.street" style.
In this case, the dots within the namespace should be substituted
with slashes so that we can provide the translation under the
"activemodel.attributes.person/contacts/addresses.street" key.
This commit is related to #3859.
This will render the error message on :#{attribute}_confirmation instead
of on attribute itself. When rendering confirmation errors inline on the
form with form builders such as SimpleForm and Formtastic it is
confusing to the ender user to see the confirmation error message on the
attribute element. Instead it makes more sense to have this validation
error render on the confirmation field instead.
The i18n message has been updated for the confirmation validator error
message to include the original attribute name.
There is an example in Rails documentation that suggests implementing
assign_attributes method for ActiveModel interface, that by default
sends option role with nil. Since mass_assignment_authorizer never
is called without args, we can move the default value internally.
Allow infinite values for validates_length_of. Particularly useful
for prettily defining an open ended range such as
validates_length_of :human_stupidity, :within => 0..Float::INFINITY
Robin Dupret