Implement `Object#with` #46681

byroot · 2022-12-09T08:17:45Z

Use case

A very common pattern in Ruby, especially in testing is to save the value of an attribute, set a new value, and then restore the old value in an ensure clause.

e.g. in unit tests

def test_something_when_enabled
  enabled_was, SomeLibrary.enabled = SomeLibrary.enabled, true
  # test things
ensure
  SomeLibrary.enabled = enabled_was
end

Or sometime in actual APIs:

def with_something_enabled
  enabled_was = @enabled
  @enabled = true
  yield
ensure
  @enabled = enabled_was
end

There is no inherent problem with this pattern, but it can be easy to make a mistake, for instance the unit test example:

def test_something_when_enabled
  some_call_that_may_raise
  enabled_was, SomeLibrary.enabled = SomeLibrary.enabled, true
  # test things
ensure
  SomeLibrary.enabled = enabled_was
end

In the above if some_call_that_may_raise actually raises, SomeLibrary.enabled is set back to nil rather than its original value. I've seen this mistake quite frequently.

Object#with

I think it would be very useful to have a method on Object to implement this pattern in a correct and easy to use way.

NB: public_send is used because I don't think such method should be usable if the accessors are private.

With usage:

def test_something_when_enabled
  SomeLibrary.with(enabled: true) do
    # test things
  end
end

GC.with(measure_total_time: true, auto_compact: false) do
  # do something
end

Lots of tests in Rails's codebase could be simplified, e.g.:

Changing Thread.report_on_exception:

rails/activerecord/test/cases/connection_pool_test.rb

Lines 583 to 595 in 2d2fdc9

    
           def test_non_bang_disconnect_and_clear_reloadable_connections_throw_exception_if_threads_dont_return_their_conns 
        
             Thread.report_on_exception, original_report_on_exception = false, Thread.report_on_exception 
        
             @pool.checkout_timeout = 0.001 # no need to delay test suite by waiting the whole full default timeout 
        
             [:disconnect, :clear_reloadable_connections].each do |group_action_method| 
        
               @pool.with_connection do |connection| 
        
                 assert_raises(ExclusiveConnectionTimeoutError) do 
        
                   new_thread { @pool.public_send(group_action_method) }.join 
        
                 end 
        
               end 
        
             end 
        
           ensure 
        
             Thread.report_on_exception = original_report_on_exception 
        
           end

Changing a class attribute:

rails/activerecord/test/cases/associations/belongs_to_associations_test.rb

Lines 136 to 150 in 2d2fdc9

    
           def test_optional_relation 
        
             original_value = ActiveRecord::Base.belongs_to_required_by_default 
        
             ActiveRecord::Base.belongs_to_required_by_default = true 
        
             model = Class.new(ActiveRecord::Base) do 
        
               self.table_name = "accounts" 
        
               def self.name; "Temp"; end 
        
               belongs_to :company, optional: true 
        
             end 
        
             account = model.new 
        
             assert_predicate account, :valid? 
        
           ensure 
        
             ActiveRecord::Base.belongs_to_required_by_default = original_value 
        
           end

fatkodima · 2022-12-09T15:04:34Z

def with_something_enabled
  enabled_was = @enabled
  @enabled = true
  yield
ensure
  @enabled = enabled_was
end

Looking at the code, something like this is not currently supported?

with(:@enabled => true)
  yield
end

Found many examples of this in the rails codebase itself. For example,

rails/actionview/lib/action_view/base.rb

Lines 273 to 290 in 59fe981

    
           def in_rendering_context(options) 
        
             old_view_renderer  = @view_renderer 
        
             old_lookup_context = @lookup_context 
        
             if !lookup_context.html_fallback_for_js && options[:formats] 
        
               formats = Array(options[:formats]) 
        
               if formats == [:js] 
        
                 formats << :html 
        
               end 
        
               @lookup_context = lookup_context.with_prepended_formats(formats) 
        
               @view_renderer = ActionView::Renderer.new @lookup_context 
        
             end 
        
             yield @view_renderer 
        
           ensure 
        
             @view_renderer = old_view_renderer 
        
             @lookup_context = old_lookup_context 
        
           end

or

rails/actionpack/test/controller/flash_test.rb

Lines 215 to 225 in 59fe981

    
           def test_redirect_to_with_adding_flash_types 
        
             original_controller = @controller 
        
             test_controller_with_flash_type_foo = Class.new(TestController) do 
        
               add_flash_types :foo 
        
             end 
        
             @controller = test_controller_with_flash_type_foo.new 
        
             get :redirect_with_foo_flash 
        
             assert_equal "for great justice", @controller.flash[:foo] 
        
           ensure 
        
             @controller = original_controller 
        
           end

byroot · 2022-12-09T15:06:18Z

something like this is not currently supported?

Nope, and I don't wish to, similarly to why it only allow to use public methods, instance variables are out of scope.

Otherwise we'd open a huge way to bypass encapsulation.

jonathanhefner · 2022-12-09T17:07:30Z

What do you think about adding require "active_support/core_ext/object/with" to active_support/test_case.rb, so that #with will be automatically available wherever ActiveSupport::TestCase is?

On one hand, active_support/test_case.rb doesn't include any other such extensions. But, on the other hand, #with is arguably tangential to active_support/testing/setup_and_teardown.

byroot · 2022-12-10T08:33:51Z

Just to clarify, this is a very common pattern in test but not only. It isn't meant as purely a test helper, otherwise AS::TestCase#with_attr(object, **attributes) would do just fine without adding to Object.

I think this is also commonly useful in production code, e.g.

Rails.logger.with(level: Logger::Debug) do
  # ...
end

etc.

jonathanhefner · 2022-12-10T16:59:05Z

Sure! I didn't mean to imply it was only for tests. I was just thinking of all the places it would be useful in Rails' own tests. But we could also add the require to test helper files (e.g. abstract_unit.rb) instead.

simi · 2022-12-10T17:14:06Z

Sad to see this "soft" rejected by Ruby itself (https://bugs.ruby-lang.org/issues/18951).

Would it be possible to make it compatible with ENV as well? That's also often used in tests.

byroot · 2022-12-12T09:39:45Z

For ENV since that is really only really used in test, an ActiveSupport::TestCase#with_envhelper would do fine (I have one copy pasted in a bunch of apps).

simi · 2022-12-12T09:55:50Z

Would it make sense to wait for https://bugs.ruby-lang.org/issues/18951 to be rejected (closed)? Do you think there is still any chance to unblock that?

byroot · 2022-12-12T09:56:45Z

I don't think so. I tried several times already.

simi · 2022-12-12T09:59:38Z

For ENV since that is really only really used in test, an ActiveSupport::TestCase#with_envhelper would do fine (I have one copy pasted in a bunch of apps).

🤔 If that's enough in tests and original description is also mostly test oriented, would it make sense to start with with test helper for now instead of introducing global Object#with?

byroot · 2022-12-12T10:00:39Z

original description is also mostly test oriented

Again, I pointed to test code, but that's not what I want it for.

simi · 2022-12-12T10:08:03Z

original description is also mostly test oriented

Again, I pointed to test code, but that's not what I want it for.

Ahh, sorry. I have missed #46681 (comment). Than the last concern from my side would be the adoption of with method in various gems (mentioned at https://bugs.ruby-lang.org/issues/18951#note-17 for example). Maybe start with less generic name like with_attr/with_attrs?

byroot · 2022-12-12T10:13:58Z

Than the last concern from my side would be the adoption of with method in various gems

I disagree with that assessment. If you look at the detail of these methods the vast majority is internal APIs, or objects you'd never set attributes on.

I also disagree with the comparison with __send__ and __object__, these methods are things you need to be able to call on arbitrary unknown for metaprogramming reasons. That's not the case of with.

dhh · 2023-03-04T16:15:56Z

Ohhh, I just wrote several of those damn ensure clauses for MRSK. Let's make this happen. Respect that matz don't want it in core, but I like both the naming and the placement for Active Support.

@enabled

Use case A very common pattern in Ruby, especially in testing is to save the value of an attribute, set a new value, and then restore the old value in an `ensure` clause. e.g. in unit tests ```ruby def test_something_when_enabled enabled_was, SomeLibrary.enabled = SomeLibrary.enabled, true # test things ensure SomeLibrary.enabled = enabled_was end ``` Or sometime in actual APIs: ```ruby def with_something_enabled enabled_was = @enabled @enabled = true yield ensure @enabled = enabled_was end ``` There is no inherent problem with this pattern, but it can be easy to make a mistake, for instance the unit test example: ```ruby def test_something_when_enabled some_call_that_may_raise enabled_was, SomeLibrary.enabled = SomeLibrary.enabled, true # test things ensure SomeLibrary.enabled = enabled_was end ``` In the above if `some_call_that_may_raise` actually raises, `SomeLibrary.enabled` is set back to `nil` rather than its original value. I've seen this mistake quite frequently. Object#with I think it would be very useful to have a method on Object to implement this pattern in a correct and easy to use way. NB: `public_send` is used because I don't think such method should be usable if the accessors are private. With usage: ```ruby def test_something_when_enabled SomeLibrary.with(enabled: true) do # test things end end ``` ```ruby GC.with(measure_total_time: true, auto_compact: false) do # do something end ``` Lots of tests in Rails's codebase could be simplified, e.g.: - Changing `Thread.report_on_exception`: https://github.com/rails/rails/blob/2d2fdc941e7497ca77f99ce5ad404b6e58f043ef/activerecord/test/cases/connection_pool_test.rb#L583-L595 - Changing a class attribute: https://github.com/rails/rails/blob/2d2fdc941e7497ca77f99ce5ad404b6e58f043ef/activerecord/test/cases/associations/belongs_to_associations_test.rb#L136-L150

byroot · 2023-03-06T08:17:15Z

Ok, I rebased and hardened the tests a bit.

dorianmariecom · 2023-03-13T05:00:06Z

What about exceptions that could be raised and put the object in the intermediate state?

What about thread-safety?

That's what I had to do in climate_control https://github.com/thoughtbot/climate_control

byroot · 2023-03-13T07:33:39Z

What about exceptions that could be raised and put the object in the intermediate state?

That's tested for: https://github.com/rails/rails/pull/46681/files#diff-07330a39026ef5856cdbc07bfeb738810459904fe6c663207b8dac9b96d97049R61-R73. If you think a case isn't covered, please open a PR.

What about thread-safety?

I don't see what thread safety there is to account for here. If you use Object#with on an object that is shared between threads, you'll have the exact same thread safety issues that if you used the setter.

That's what I had to do in climate_control https://github.com/thoughtbot/climate_control

I assume you mean: thoughtbot/climate_control@d556bf4, but I don't see what you're getting at.

gsamokovarov · 2023-03-15T13:25:42Z

Should we consider a different name for this core extension method? Something longer like Object.with_attribute? with is a good method name for DSLs, and we do those often in Ruby.

byroot · 2023-03-15T16:24:51Z

The method being defined in object doesn't prevent you in any way to define another method with the same name for your DSL.

solnic · 2023-03-16T09:48:00Z

@byroot adding this to NilClass will cause confusing errors in people's codebases as this method is very common in many libraries. You could at least ensure nil doesn't respond to it as it makes no sense to have that method on nil.

byroot · 2023-03-16T10:25:37Z

Sure, not a bad idea.

byroot · 2023-03-16T10:43:55Z

Done in #47688

joeldrapper · 2023-03-16T10:58:07Z

I strongly recommend renaming this to with_attr (aliased as with_attrs). The attr abbreviation brings to mind the related concepts of attr_reader, atter_writer, etc. and it's less likely to conflict with other methods.

For example, any Data object will accept with with a block and attributes, but the block will be ignored since Data#with returns a copy. It will fail silently and frustratingly for many people.

Additionally, with_attr / with_attrs could raise a helpful ArgumentError if the object is frozen, since you can’t use this technique with frozen objects.

@solnic mentioned, it doesn't make sense for NilClass to respond to with_attr / with_attrs at all — the same is true of TrueClass, FalseClass, Symbol, etc. We could maintain a list of these cases, but they’d also be covered by the frozen? check.

byroot · 2023-03-16T11:02:37Z

For example, any Data object will accept with with a block and attributes, but the block will be ignored since Data#with returns a copy. It will fail silently and frustratingly for many people.

This is totally irrelevant. Try this without active support loaded:

Data.define(:foo).new(1).with(foo: 2) do
  raise "Oops this block is useless"
end

Additionally, with_attr / with_attrs could raise a helpful ArgumentError

No need to rename to do that
frozen.with(foo: 1) will raise the same error than frozen.foo = 1 which IMO is perfectly fine.

it doesn't make sense for NilClass ...

Already changed.

joeldrapper · 2023-03-16T11:04:24Z

This is totally irrelevant. Try this without active support loaded:

No one is going to try that without ActiveSupport loaded. But if they get used to using ActiveSupport’s Object#with because they’re using ActiveSupport and reading ActiveSupport’s documentation, they may try it on a Data object.

byroot · 2023-03-16T11:08:55Z

they may try it on a Data object.

That's an hypothetical I don't agree with, and there is similar redefinition of Object methods in the stdlib, e.g. UDPSocket#send, etc.

joeldrapper · 2023-03-16T11:11:51Z

As far as I can tell, UDPSocket isn't typically used as a base class for a specialised class under a different name. Data#with won't look like Data#with since Data is an abstract class. It'll be User#with, Address#with, Event#with, etc.

akostadinov · 2025-07-26T08:56:42Z

activesupport/lib/active_support/core_ext/object/with.rb

+  #
+  # It can be used on any object as long as both the reader and writer methods
+  # are public.
+  def with(**attributes)


Tthis method name is super generic. Very easy to create confusion. Maybe call it #with_attributes instead of just a generic #with?

rails-bot bot added the activesupport label Dec 9, 2022

casperisfine force-pushed the object-with branch from db7276f to b510e9f Compare December 9, 2022 08:19

casperisfine force-pushed the object-with branch from b510e9f to 1884323 Compare March 6, 2023 08:02

byroot merged commit 091e04a into rails:main Mar 6, 2023

zzak added the thisweekinrails label Mar 6, 2023

palkan mentioned this pull request Mar 6, 2023

Rails 7.1 breaks receive().with rspec/rspec-mocks#1530

Closed

estolfo mentioned this pull request Mar 7, 2023

Fixing Faraday::RackBuilder::StackLocked elastic/apm-agent-ruby#1371

Merged

casperisfine mentioned this pull request Mar 16, 2023

Undefine with on NilClass and other common immediate types #47688

Merged

mattbrictson mentioned this pull request Oct 24, 2023

Fix Object.with test class name typo #49766

Merged

fatkodima mentioned this pull request Apr 29, 2024

Raise when a block is passed to ActiveRecord::Relation#with #51684

Merged

whtsht mentioned this pull request May 8, 2025

Redis adapter not available for certain rails and redis version flippercloud/flipper#926

Closed

mayorova mentioned this pull request Jul 26, 2025

THREESCALE-11853: Upgrade to Rails 7.1 3scale/porta#4089

Merged

akostadinov reviewed Jul 26, 2025

View reviewed changes

	def test_non_bang_disconnect_and_clear_reloadable_connections_throw_exception_if_threads_dont_return_their_conns
	Thread.report_on_exception, original_report_on_exception = false, Thread.report_on_exception
	@pool.checkout_timeout = 0.001 # no need to delay test suite by waiting the whole full default timeout
	[:disconnect, :clear_reloadable_connections].each do \|group_action_method\|
	@pool.with_connection do \|connection\|
	assert_raises(ExclusiveConnectionTimeoutError) do
	new_thread { @pool.public_send(group_action_method) }.join
	end
	end
	end
	ensure
	Thread.report_on_exception = original_report_on_exception
	end

	def test_optional_relation
	original_value = ActiveRecord::Base.belongs_to_required_by_default
	ActiveRecord::Base.belongs_to_required_by_default = true

	model = Class.new(ActiveRecord::Base) do
	self.table_name = "accounts"
	def self.name; "Temp"; end
	belongs_to :company, optional: true
	end

	account = model.new
	assert_predicate account, :valid?
	ensure
	ActiveRecord::Base.belongs_to_required_by_default = original_value
	end

Implement Object#with #46681

Implement Object#with #46681

Conversation

byroot commented Dec 9, 2022

Use case

Object#with

Uh oh!

fatkodima commented Dec 9, 2022

Uh oh!

byroot commented Dec 9, 2022

Uh oh!

jonathanhefner commented Dec 9, 2022

Uh oh!

byroot commented Dec 10, 2022

Uh oh!

jonathanhefner commented Dec 10, 2022

Uh oh!

simi commented Dec 10, 2022

Uh oh!

byroot commented Dec 12, 2022

Uh oh!

simi commented Dec 12, 2022

Uh oh!

byroot commented Dec 12, 2022

Uh oh!

simi commented Dec 12, 2022

Uh oh!

byroot commented Dec 12, 2022

Uh oh!

simi commented Dec 12, 2022

Uh oh!

byroot commented Dec 12, 2022

Uh oh!

dhh commented Mar 4, 2023

Uh oh!

byroot commented Mar 6, 2023

Uh oh!

dorianmariecom commented Mar 13, 2023

Uh oh!

byroot commented Mar 13, 2023

Uh oh!

gsamokovarov commented Mar 15, 2023

Uh oh!

byroot commented Mar 15, 2023

Uh oh!

solnic commented Mar 16, 2023

Uh oh!

byroot commented Mar 16, 2023

Uh oh!

byroot commented Mar 16, 2023

Uh oh!

joeldrapper commented Mar 16, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

byroot commented Mar 16, 2023

Uh oh!

joeldrapper commented Mar 16, 2023

Uh oh!

byroot commented Mar 16, 2023

Uh oh!

joeldrapper commented Mar 16, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

akostadinov Jul 26, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Implement `Object#with` #46681

Implement `Object#with` #46681

joeldrapper commented Mar 16, 2023 •

edited

Loading

joeldrapper commented Mar 16, 2023 •

edited

Loading