Ensure #second (and others) acts like #first AR finder

[terracatta]

This PR addresses issue #13743 where the ordinal ActiveSupport Array extentions were not assuming a default order like first was.

As suggested by @dhh this commit brings the famous ordinal Array instance methods defined
in ActiveSupport into ActiveRecord as fully-fledged finders.

These finders ensure a default ascending order of the table's primary key, and utilize the OFFSET SQL verb to locate the user's desired record. If an offset is defined in the query, calling #second adds
to the offset to get the actual desired record.

Example:

    User.all.second

    # Before
    # => 'SELECT  "users".* FROM "users"'

    # After
    # => SELECT  "users".* FROM "users"   ORDER BY "users"."id" ASC LIMIT 1 OFFSET 1'

    User.offset(3).second

    # Before
    # => 'SELECT "users".* FROM "users"  LIMIT -1 OFFSET 3' # sqlite3 gem
    # => 'SELECT "users".* FROM "users"  OFFSET 3' # pg gem
    # => 'SELECT `users`.* FROM `users`  LIMIT 18446744073709551615 OFFSET 3' # mysql2 gem"'

    # After
    # => SELECT  "users".* FROM "users"   ORDER BY "users"."id" ASC LIMIT 1 OFFSET 4'

[dhh]

Looking good to me! 👍

[josevalim]

This function should not have knowledge of :first, :second and friends. We should probably just pass the offset_value as argument.

[terracatta]

@josevalim Ok, fixed that up for you! Any other concerns?

[senny]

Can you add:

Fixes #13743.

[terracatta]

Added it after the examples! Thanks for the heads up!

[senny]

@terracatta this looks good. Can you squash your commits?

[carlosantoniodasilva]

I have a question: how problematic can this cached variable be now that we are using it for different "finds"?

[rafaelfranca]

Very problematic. It will cache only the first call.

[terracatta]

Ok so I removed the caching here, but obviously that isn't the right way to go as that will cause multiple queries to fire unnecessarily when calling .first a bunch of times. I think the only option here is cache based on offset being passed into the private method.

Any other suggestions?

[terracatta]

ok I just went with that

[terracatta]

@rafaelfranca @carlosantoniodasilva @senny, commits are squashed, updated the changelog, fixed the caching issue by dynamically setting and manually memoizing the instance variable based on the offset passed in. Tested it again on several personal rails projects and the test suite looks good!

Any other feedback?

[rafaelfranca]

It is better to initialize a empty hash in the initialize of Relation and use the offset as the key to memoize it.

Since we are adding an internal variable to cache the value we have to update reset method

rails/activerecord/lib/active_record/relation.rb

Line 497 in 080fc9c

def reset

[terracatta]

Great suggestion, I'll implement it ASAP

[terracatta]

@rafaelfranca, I made the requested changes, but couldn't reproduce your issue or create a failing test case (still trying), can you try again with the latest code and see if you still get a negative limit?

[terracatta]

@rafaelfranca Ok great, added better before examples for pg, sqlite3, and mysql2 gems. I also added some extra test cases for the offset functionality just in case!

Anything else? Sorry for all the back and forth!

[rafaelfranca]

Looks good. Thank you so much for working on this.

[rafaelfranca]

Merged ❤️

[carlosantoniodasilva]

This is apparently not going to work with already loaded associations. Just testing around I got two different issues:

# consider the models `Author` has many `Post`s

## 1st scenario - raises exception
>> carlos.posts.reload.first
  Post Load (0.1ms)  SELECT "posts".* FROM "posts"  WHERE "posts"."author_id" = ?  [["author_id", 1]]
=> #<Post id: 1, author_id: 1, title: "ZOMG", body: nil, created_at: "2014-01-21 11:44:22", updated_at: "2014-01-21 11:44:22">
>> carlos.posts.reload.second
  Post Load (0.1ms)  SELECT "posts".* FROM "posts"  WHERE "posts"."author_id" = ?  [["author_id", 1]]
NoMethodError: undefined method 'first' for nil:NilClass
  from....rails/rails/activerecord/lib/active_record/relation/finder_methods.rb:449:in 'find_nth'


## 2nd scenario - returns the same record
>> carlos.posts.first
=> #<Post id: 1, author_id: 1, title: "ZOMG", body: nil, created_at: "2014-01-21 11:44:22", updated_at: "2014-01-21 11:44:22">
>> carlos.posts.second
=> #<Post id: 1, author_id: 1, title: "ZOMG", body: nil, created_at: "2014-01-21 11:44:22", updated_at: "2014-01-21 11:44:22">
>> carlos.posts.third
=> #<Post id: 1, author_id: 1, title: "ZOMG", body: nil, created_at: "2014-01-21 11:44:22", updated_at: "2014-01-21 11:44:22">

@terracatta can you take a look please? It'd be good to provide some tests for already loaded associations with some of these methods. Thanks.

[terracatta]

Absolutely! Thanks for testing!