Skip to content

GeorgeKaraszi/ActiveRecordExtended

Repository files navigation

Gem Version Build Status Maintainability Test Coverage

Index

Description and History

Active Record Extended is the continuation of maintaining and improving the work done by Dan McClain, the original author of postgres_ext.

Overtime the lack of updating to support the latest versions of ActiveRecord 5.x has caused quite a bit of users forking off the project to create their own patches jobs to maintain compatibility. The only problem is that this has created a wild west of environments of sorts. The problem has grown to the point no one is attempting to directly contribute to the original source. And forked repositories are finding themselves as equally as dead with little to no activity.

Active Record Extended is essentially providing users with the other half of Postgreses querying abilities. Due to Rails/ActiveRecord/Arel being designed to be DB agnostic, there are a lot of left out features; Either by choice or the simple lack of supporting API's for other databases. However some features are not exactly PG explicit. Some are just helper methods to express an idea much more easily.

Compatibility

This package is designed align and work with any officially supported Ruby and Rails versions.

  • Minimum Ruby Version: 3.0.x (EOL warning!)
  • Minimum Rails Version: 6.0.x (EOL warning!)
  • Minimum Postgres Version: 12.x (EOL warning!)
  • Latest Ruby supported: 3.2.x
  • Latest Rails supported: 7.1.x
  • Postgres: 11-current(15) (probably works with most older versions to a certain point)

Installation

Add this line to your application's Gemfile:

gem 'active_record_extended'

And then execute:

$ bundle

Usage

Predicate Query Methods

Any

Postgres 'ANY' expression

In Postgres the ANY expression is used for gather record's that have an Array column type that contain a single matchable value within its array.

alice = User.create!(tags: [1])
bob   = User.create!(tags: [1, 2])
randy = User.create!(tags: [3])

User.where.any(tags: 1) #=> [alice, bob]

This only accepts a single value. So querying for example multiple tag numbers [1,2] will return nothing.

All

Postgres 'ALL' expression

In Postgres the ALL expression is used for gather record's that have an Array column type that contains only a single and matchable element.

alice = User.create!(tags: [1])
bob   = User.create!(tags: [1, 2])
randy = User.create!(tags: [3])

User.where.all(tags: 1) #=> [alice]

This only accepts a single value to a given attribute. So querying for example multiple tag numbers [1,2] will return nothing.

Contains

Postgres '@>' (Array type) Contains expression

Postgres '@>' (JSONB/HSTORE type) Contains expression

The contains/1 method is used for finding any elements in an Array, JSONB, or HSTORE column type. That contains all of the provided values.

Array Type:

alice = User.create!(tags: [1, 4])
bob   = User.create!(tags: [3, 1])
randy = User.create!(tags: [4, 1])

User.where.contains(tags: [1, 4]) #=> [alice, randy]

HSTORE / JSONB Type:

alice = User.create!(data: { nickname: "ARExtend" })
bob   = User.create!(data: { nickname: "ARExtended" })
randy = User.create!(data: { nickname: "ARExtended" })

User.where.contains(data: { nickname: "ARExtended" }) #=> [bob, randy]

Overlap

Postgres && (overlap) Expression

The overlap/1 method will match an Array column type that contains any of the provided values within its column.

alice = User.create!(tags: [1, 4])
bob   = User.create!(tags: [3, 4])
randy = User.create!(tags: [4, 8])

User.where.overlap(tags: [4]) #=> [alice, bob, randy]
User.where.overlap(tags: [1, 8]) #=> [alice, randy]
User.where.overlap(tags: [1, 3, 8]) #=> [alice, bob, randy]

Inet / IP Address

Inet Contains

Postgres >> (contains) Network Expression

The inet_contains method works by taking a column(inet type) that has a submask prepended to it. And tries to find related records that fall within a given IP's range.

alice = User.create!(ip: "127.0.0.1/16")
bob   = User.create!(ip: "192.168.0.1/16")

User.where.inet_contains(ip: "127.0.0.254") #=> [alice]
User.where.inet_contains(ip: "192.168.20.44") #=> [bob]
User.where.inet_contains(ip: "192.255.1.1") #=> []
Inet Contains or Equals

Postgres >>= (contains or equals) Network Expression

The inet_contains_or_equals method works much like the Inet Contains method, but will also accept a submask range.

alice = User.create!(ip: "127.0.0.1/10")
bob   = User.create!(ip: "127.0.0.44/24")

User.where.inet_contains_or_equals(ip: "127.0.0.1/16") #=> [alice]
User.where.inet_contains_or_equals(ip: "127.0.0.1/10") #=> [alice]
User.where.inet_contains_or_equals(ip: "127.0.0.1/32") #=> [alice, bob]
Inet Contained Within

Postgres << (contained by) Network Expression

For the inet_contained_within method, we try to find IP's that fall within a submasking range we provide.

alice = User.create!(ip: "127.0.0.1")
bob   = User.create!(ip: "127.0.0.44")
randy = User.create!(ip: "127.0.55.20")

User.where.inet_contained_within(ip: "127.0.0.1/24") #=> [alice, bob]
User.where.inet_contained_within(ip: "127.0.0.1/16") #=> [alice, bob, randy]
Inet Contained Within or Equals

Postgres <<= (contained by or equals) Network Expression

The inet_contained_within_or_equals method works much like the Inet Contained Within method, but will also accept a submask range.

alice = User.create!(ip: "127.0.0.1/10")
bob   = User.create!(ip: "127.0.0.44/32")
randy = User.create!(ip: "127.0.99.1")

User.where.inet_contained_within_or_equals(ip: "127.0.0.44/32") #=> [bob]
User.where.inet_contained_within_or_equals(ip: "127.0.0.1/16") #=> [bob, randy]
User.where.inet_contained_within_or_equals(ip: "127.0.0.44/8") #=> [alice, bob, randy]
Inet Contains or Contained Within

Postgres && (contains or is contained by) Network Expression

The inet_contains_or_contained_within method is a combination of Inet Contains and Inet Contained Within. It essentially (the database) tries to use both methods to find as many records as possible that match either condition on both sides.

alice = User.create!(ip: "127.0.0.1/24")
bob   = User.create!(ip: "127.0.22.44/8")
randy = User.create!(ip: "127.0.99.1")

User.where.inet_contains_or_is_contained_within(ip: "127.0.255.80") #=> [bob]
User.where.inet_contains_or_is_contained_within(ip: "127.0.0.80") #=> [alice, bob]
User.where.inet_contains_or_is_contained_within(ip: "127.0.0.80/8") #=> [alice, bob, randy]

Conditional Methods

Any_of / None_of

any_of/1 simplifies the process of finding records that require multiple or conditions.

none_of/1 is the inverse of any_of/1. It'll find records where none of the contains are matched.

Both accepts An array of: ActiveRecord Objects, Query Strings, and basic attribute names.

Querying With Attributes:

alice = User.create!(uid: 1)
bob   = User.create!(uid: 2)
randy = User.create!(uid: 3)

User.where.any_of({ uid: 1 }, { uid: 2 }) #=> [alice, bob]

Querying With ActiveRecord Objects:

alice = User.create!(uid: 1)
bob   = User.create!(uid: 2)
randy = User.create!(uid: 3)

uid_one = User.where(uid: 1)
uid_two = User.where(uid: 2)

User.where.any_of(uid_one, uid_two) #=> [alice, bob]

Querying with Joined Relationships:

alice     = User.create!(uid: 1)
bob       = User.create!(uid: 2)
randy     = User.create!(uid: 3)
tag_alice = Tag.create!(user_id: alice.id)
tag_bob   = Tag.create!(user_id: bob.id)
tag_randy = Tag.create!(user_id: randy.id)

bob_tag_query   = Tag.where(users: { id: bob.id }).includes(:user)
randy_tag_query = Tag.where(users: { id: randy.id }).joins(:user)

Tag.joins(:user).where.any_of(bob_tag_query, randy_tag_query) #=> [tag_bob, tag_randy] (with users table joined)

Either Join

The #either_join/2 method is a base ActiveRecord querying method that will joins records based on a set of conditionally joinable tables.

class User < ActiveRecord::Base
  has_one :profile_l, class: "ProfileL"
  has_one :profile_r, class: "ProfileR"

  scope :completed_profile, -> { either_joins(:profile_l, :profile_r) }
end

alice = User.create!
bob   = User.create!
randy = User.create! # Does not have a single completed profile type
ProfileL.create!(user_id: alice.id)
ProfileR.create!(user_id: bob.id)

User.completed_profile #=> [alice, bob]
# alternatively
User.either_joins(:profile_l, :profile_r) #=> [alice, bob]

Either Order

The #either_order/3 method is a base ActiveRecord querying method that will order a set of columns that may or may not exist for each record. This works similar to how Either Join works. This does not however exclude records that do not have relationships.

alice = User.create!
bob   = User.create!
ProfileL.create!(user_id: alice.id, left_turns: 100)
ProfileR.create!(user_id: bob.id, right_turns: 50)

User.either_order(:asc, profile_l: :left_turns, profile_r: :right_turns) #=> [bob, alice]
User.either_order(:desc, profile_l: :left_turns, profile_r: :right_turns) #=> [alice, bob]

randy = User.create!
User.either_order(:asc, profile_l: :left_turns, profile_r: :right_turns) #=> [bob, alice, randy]
User.either_order(:desc, profile_l: :left_turns, profile_r: :right_turns) #=> [randy, alice, bob]

Common Table Expressions (CTE)

Postgres WITH (CTE) Statement

The .with/1 method is a base ActiveRecord querying method that will aid in creating complex queries.

alice = User.create!
bob   = User.create!
randy = User.create!
ProfileL.create!(user_id: alice.id, likes: 200)
ProfileL.create!(user_id: bob.id,   likes: 400)
ProfileL.create!(user_id: randy.id, likes: 600)

User.with(highly_liked: ProfileL.where("likes > 300"))
    .joins("JOIN highly_liked ON highly_liked.user_id = users.id") #=> [bob, randy]

Query output:

WITH "highly_liked" AS (SELECT "profile_ls".* FROM "profile_ls" WHERE (likes >= 300))
SELECT "users".*
FROM "users"
JOIN highly_liked ON highly_liked.user_id = users.id

You can also chain or provide additional arguments to the with/1 method for it to merge into a single, WITH statement.

User.with(highly_liked: ProfileL.where("likes > 300"), less_liked: ProfileL.where("likes <= 200"))
    .joins("JOIN highly_liked ON highly_liked.user_id = users.id")
    .joins("JOIN less_liked ON less_liked.user_id = users.id")

# OR

User.with(highly_liked: ProfileL.where("likes > 300"))
    .with(less_liked: ProfileL.where("likes <= 200"))
    .joins("JOIN highly_liked ON highly_liked.user_id = users.id")
    .joins("JOIN less_liked ON less_liked.user_id = users.id")

Query output:

WITH "highly_liked" AS (SELECT "profile_ls".* FROM "profile_ls" WHERE (likes > 300)),
     "less_liked" AS (SELECT "profile_ls".* FROM "profile_ls" WHERE (likes <= 200))
SELECT "users".*
FROM "users"
JOIN highly_liked ON highly_liked.user_id = users.id
JOIN less_liked ON less_liked.user_id = users.id

There are three methods you can chain to the with/1 to add modifiers to the query.

recursive

User.with.recursive(highly_liked: ProfileL.where("likes > 300"))
    .joins("JOIN highly_liked ON highly_liked.user_id = users.id")

Query output:

WITH RECURSIVE "highly_liked" AS (SELECT "profile_ls".* FROM "profile_ls" WHERE (likes >= 300))
SELECT "users".*
FROM "users"
JOIN highly_liked ON highly_liked.user_id = users.id

materialized (Note: MATERIALIZED modifier is only available in PG versions 12+.)

User.with.materialized(highly_liked: ProfileL.where("likes > 300"))
    .joins("JOIN highly_liked ON highly_liked.user_id = users.id")

Query output:

WITH "highly_liked" AS MATERIALIZED (SELECT "profile_ls".* FROM "profile_ls" WHERE (likes >= 300))
SELECT "users".*
FROM "users"
JOIN highly_liked ON highly_liked.user_id = users.id

not_materialized (Note: NOT MATERIALIZED modifier is only available in PG versions 12+.)

User.with.not_materialized(highly_liked: ProfileL.where("likes > 300"))
    .joins("JOIN highly_liked ON highly_liked.user_id = users.id")

Query output:

WITH "highly_liked" AS NOT MATERIALIZED (SELECT "profile_ls".* FROM "profile_ls" WHERE (likes >= 300))
SELECT "users".*
FROM "users"
JOIN highly_liked ON highly_liked.user_id = users.id

Subquery CTE Gotchas

In order keep queries PG valid, subquery explicit methods (like Unions and JSON methods) will be subject to "Piping" the CTE clauses up to the parents query level.

This also means there's potential for having duplicate CTE names. In order to combat duplicate CTE references with the same name, piping will favor the parents CTE over the nested sub-queries.

This also means that this is a "First come First Served" implementation. So if you have a parent with no CTE's but two sub-queries with the same CTE name but with different querying statements. It will process and favor the one that comes first.

Example:

   sub_query      = Person.with(dupped_cte: Person.where(id: 1)).select("dup_cte.id").from(:dup_cte)
   other_subquery = Person.with(unique_cte: Person.where(id: 5)).select("unique_cte.id").from(:unique_cte)

   # Will favor this CTE below, over the `sub_query`'s CTE
   Person.with(dupped_cte: Person.where.not(id: 1..4)).union(sub_query, other_subquery)

Query Output

WITH "unique_cte" AS (
  SELECT "people".*
  FROM "people"
  WHERE "people"."id" = 5
), "dupped_cte" AS (
  SELECT "people".*
  FROM "people"
  WHERE NOT ("people"."id" BETWEEN 1 AND 4)
)
  SELECT "people".*
  FROM (( (
    SELECT dup_cte.id
    FROM dup_cte
  ) UNION (
    SELECT unique_cte.id
    FROM unique_cte
  ) )) people

JSON Query Methods

If any or all of your json sub-queries include a CTE, read the Subquery CTE Gotchas warnings.

Row To JSON

Postgres 'ROW_TO_JSON' function

The implementation of the.select_row_to_json/2 method is designed to be used with sub-queries. As a means for taking complex query logic and transform them into a single or multiple json responses. These responses are required to be assigned to an aliased column on the parent(callee) level.

While quite the mouthful of an explanation. The implementation of combining unrelated or semi-related queries is quite smooth(imo).

Arguments:

  • from [String, Arel, or ActiveRecord::Relation]: A subquery that can be nested into a ROW_TO_JSON clause

Options:

  • as [Symbol or String] (default="results"): What the column will be aliased to
  • key [Symbol or String] (default=[random letter]): Internal query alias name.
    • This is useful if you would like to add additional mid-level predicate clauses
  • cast_with [Symbol or Array<Symbol>]:
    • :to_jsonb
    • :array
    • :array_agg
    • :distinct (auto applies :array_agg & :to_jsonb)
  • order_by [Symbol or Hash]: Applies an ordering operation (similar to ActiveRecord #order)
    • Note: this option will be ignored if you need to order a DISTINCT Aggregated Array.
    physical_cat  = Category.create!(name: "Physical")
    products      = 3.times.map { Product.create! }
    products.each { |product|  100.times { Variant.create!(product: product, category: physical_cat) } }

    # Since we plan to nest this query, you have access top level information. (I.E categories table)
    item_query = Variant.select(:name, :id, :category_id, :product_id).where("categories.id = variants.category_id")

    # You can provide addition scopes that will be applied to the nested query (but will not effect the actual inner query)
    # This is ideal if you are dealing with but not limited to, CTE's being applied multiple times and require additional constraints
    product_query  =
    Product.select(:id)
            .joins(:items)
            .select_row_to_json(item_query, key: :outer_items, as: :items, cast_with: :array) do |item_scope|
              item_scope.where("outer_items.product_id = products.id")
                # Results to:
                #  SELECT ..., ARRAY(SELECT ROW_TO_JSON("outer_items")
                #   FROM ([:item_query:]) outer_items
                #   WHERE outer_items.product_id = products.id
                # ) AS items
            end

    # Not defining a key will automatically generate a random key between a-z
    category_query = Category.select(:name, :id).select_row_to_json(product_query, as: :products, cast_with: :array)
    Category.json_build_object(:physical_category, category_query.where(id: physical_cat.id)).results
    #=> {
    #        "physical_category" => {
    #            "name" => "Physical",
    #            "id" => 1,
    #            "products" => [
    #              {
    #                "id" => 2,
    #                "items" => [{"name" => "Bojangels", "id" => 3, "category_id" => 1, "product_id" => 2}, ...]
    #              },
    #              ...
    #            ]
    #        }
    #  }
    #

Query Output

SELECT (JSON_BUILD_OBJECT('physical_category', "physical_category")) AS "results"
FROM (
     SELECT "categories"."name", "categories"."id", (ARRAY(
         SELECT ROW_TO_JSON("j")
         FROM (
              SELECT "products"."id", (ARRAY(
                  SELECT ROW_TO_JSON("outer_item")
                  FROM (
                       SELECT "variants"."name", "variants"."id", "variants"."category_id", "variants"."product_id"
                       FROM "variants"
                       WHERE (categories.id = variants.category_id)
                       ) outer_items
                  WHERE (outer_items.product_id = products.id)
                )) AS "items"
              FROM "products"
              INNER JOIN "items" ON "products"."id" = "items"."product_id"
              ) j
       )) AS "products"
     FROM "categories"
     WHERE "categories"."id" = 1
     ) AS "physical_category"

JSON/B Build Object

Postgres 'json(b)_build_object' function

The implementation of the.json_build_object/2 and .jsonb_build_object/2 methods are designed to be used with sub-queries. As a means for taking complex query logic and transform them into a single or multiple json responses.

Arguments:

  • key: [Symbol or String]: What should this response return as
  • from: [String, Arel, or ActiveRecord::Relation] : A subquery that can be nested into the top-level from clause

Options:

  • as: [Symbol or String] (defaults to "results"): What the column will be aliased to
  • value: [Symbol or String] (defaults to key argument): How the response should handel the json value return

See the included example on Row To JSON to see it in action.

JSON/B Build Literal

Postgres 'json(b)_build_object' function

The implementation of the.json_build_literal/1 and .jsonb_build_literal/1 is designed for creating static json objects that don't require subquery interfacing.

Arguments:

  • Requires an Array or Hash set of values

Options:

  • as: [Symbol or String] (defaults to "results"): What the column will be aliased to
    User.json_build_literal(number: 1, last_name: "json", pi: 3.14).take.results
     #=> { "number" => 1, "last_name" => "json", "pi" => 3.14 }

    # Or as array elements
    User.json_build_literal(:number, 1, :last_name, "json", :pi, 3.14).take.results
      #=> { "number" => 1, "last_name" => "json", "pi" => 3.14 }

Query Output

SELECT (JSON_BUILD_OBJECT('number', 1, 'last_name', 'json', 'pi', 3.14)) AS "results"
  FROM "users"

Unionization

If any or all of your union queries include a CTE, read the Subquery CTE Gotchas warnings.

SQL-Query Helpers

  • .to_union_sql : Will return a string of the constructed union query without being nested in the from clause.
  • .to_nice_union_sql(requires NiceQL Gem to be install): A formatted .to_union_sql

Known issue

There's an issue with providing a single union clause and chaining it with a different union clause. This is due to requirements of grouping SQL statements. The issue is being working on, but with no ETA.

This issue only applies to the first initial set of unions and is recommended that you union two relations right off the bat. Afterwords you can union/chain single relations.

Example

Person.union(Person.where(id: 1..4)).union_except(Person.where(id: 3..4)).union(Person.where(id: 4))
#=> Will include all people with an ID between 1 & 3 (throwing the except on ID 4)

# This can be fixed by doing something like

Person.union_except(Person.where(id: 1..4), Person.where(id: 3..4)).union(Person.where(id: 4))
#=> Will include people with the ids of 1, 2, and 4 (properly excluding the user with the ID of 3)

Problem Query Output

( ( (
  SELECT "people".*
  FROM "people"
  WHERE "people"."id" BETWEEN 1 AND 4
) UNION (
  SELECT "people".*
  FROM "people"
  WHERE "people"."id" BETWEEN 3 AND 4
) ) EXCEPT (
  SELECT "people".*
  FROM "people"
  WHERE "people"."id" = 4
) )

Union

Postgres 'UNION' combination

user_1 = Person.where(id: 1)
user_2 = Person.where(id: 2)
users  = Person.where(id: 1..3)

Person.union(user_1, user_2, users) #=> [#<Person id: 1, ..>, #<Person id: 2,..>, #<Person id: 3,..>]

# You can also chain union's
Person.union(user_1).union(user_2).union(users)

Query Output

SELECT "people".*
  FROM (( ( (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 1
  ) UNION (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 2
  ) ) UNION (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 1 AND 3
  ) )) people

Union ALL

Postgres 'UNION ALL' combination

user_1 = Person.where(id: 1)
user_2 = Person.where(id: 2)
users  = Person.where(id: 1..3)

Person.union_all(user_1, user_2, users)
  #=> [#<Person id: 1, ..>, #<Person id: 2,..>, #<Person id: 1, ..>, #<Person id: 2,..>, #<Person id: 3,..>]

# You can also chain union's
Person.union_all(user_1).union_all(user_2).union_all(users)
# Or
Person.union.all(user1, user_2).union.all(users)

Query Output

SELECT "people".*
  FROM (( ( (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 1
  ) UNION ALL (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 2
  ) ) UNION ALL (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 1 AND 3
  ) )) people

Union Except

Postgres 'EXCEPT' combination

users               = Person.where(id: 1..5)
except_these_users  = Person.where(id: 2..4)

Person.union_except(users, except_these_users) #=> [#<Person id: 1, ..>, #<Person id: 5,..>]

# You can also chain union's
Person.union.except(users, except_these_users).union(Person.where(id: 20))

Query Output

SELECT "people".*
  FROM (( ( (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 1 AND 5
  ) EXCEPT (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 2 AND 4
  )) people

Union Intersect

Postgres 'INTERSECT' combination

randy = Person.create!
alice = Person.create!
ProfileL.create!(person: randy, likes: 100)
ProfileL.create!(person: alice, likes: 120)

likes_100           = Person.select(:id, "profile_ls.likes").joins(:profile_l).where(profile_ls: { likes: 100 })
likes_less_than_150 = Person.select(:id, "profile_ls.likes").joins(:profile_l).where("profile_ls.likes < 150")
Person.union_intersect(likes_100, likes_less_than_150) #=> [randy]



# You can also chain union's
Person.union_intersect(likes_100).union_intersect(likes_less_than_150) #=> [randy]
# Or
Person.union.intersect(likes_100, likes_less_than_150) #=> [randy]

Query Output

SELECT "people".*
  FROM (( (
    SELECT "people"."id", profile_ls.likes
    FROM "people"
    INNER JOIN "profile_ls" ON "profile_ls"."person_id" = "people"."id"
    WHERE "profile_ls"."likes" = 100
  ) INTERSECT (
    SELECT "people"."id", profile_ls.likes
    FROM "people"
    INNER JOIN "profile_ls" ON "profile_ls"."person_id" = "people"."id"
    WHERE (profile_ls.likes < 150)
  ) )) people

Union As

By default unions are nested in the from clause and are aliased to the parents table name. We can change this behavior by chaining the method .union_as/1

Person.select("good_people.id").union(Person.where(id: 1), Person.where(id: 2)).union_as(:good_people)

Query Output

SELECT good_people.id
  FROM (( (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 1
  ) UNION (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 2
  ) )) good_people

Union Order

Unions allow for a final outside ORDER BY clause. This will ensure that all the results that come back are ordered in an expected return.

query_1 = Person.where(id: 1..3)
query_2 = Person.where(id: 3)
query_3 = Person.where(id: 3..10)
Person.union_except(query_1, query_2).union(query_3).order_union(:id, tags: :desc)

Query Output

SELECT "people".*
  FROM (( ( (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 1 AND 3
  ) EXCEPT (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 3
  ) ) UNION (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 3 AND 10
  ) ) ORDER BY id ASC, tags DESC) people

Union Reorder

much like Rails .reorder; .reorder_union/1 will clear the previous order in a new instance and/or apply a new ordering scheme

query_1     = Person.where(id: 1..3)
query_2     = Person.where(id: 3)
query_3     = Person.where(id: 3..10)
union_query = Person.union_except(query_1, query_2).union(query_3).order_union(:id, tags: :desc)
union_query.reorder_union(personal_id: :desc, id: :desc)

Query Output

SELECT "people".*
  FROM (( ( (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 1 AND 3
  ) EXCEPT (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" = 3
  ) ) UNION (
    SELECT "people".*
    FROM "people"
    WHERE "people"."id" BETWEEN 3 AND 10
  ) ) ORDER BY personal_id DESC, id DESC) people

Window Functions

Postgres Window Functions

Let's address the elephant in the room. Arel has had, for a long time now, window function capabilities; However they've never seen the lime light in ActiveRecord's query logic. The following brings the dormant Arel methods up to the ActiveRecord Querying level.

Define Window

To set up a window function, we first must establish the window and we do this by using the .define_window/1 method. This method also requires you to call chain .partition_by/2

.define_window/1 - Establishes the name of the window you'll reference later on in .select_window

  • Aliased name of window

.partition_by/2 - Establishes the windows operations a pre-defined window function will leverage.

  • column name being partitioned against
  • (optional) order_by: Processes how the window should be ordered
User
.define_window(:number_window).partition_by(:number, order_by: { id: :desc })
.define_window(:name_window).partition_by(:name, order_by: :id)
.define_window(:no_order_name).partition_by(:name)

Query Output

SELECT *
FROM users
WINDOW number_window AS (PARTITION BY number ORDER BY id DESC),
       name_window   AS (PARTITION BY name ORDER BY id),
       no_order_name AS (PARTITION BY name)

Select Window

Once you've define a window, the next step to to utilize it on one of the many provided postgres window functions.

.select_window/3

User.create!(name: "Alice", number: 100) #=> id: 1
User.create!(name: "Randy", number: 100) #=> id: 2
User.create!(name: "Bob", number: 300)   #=> id: 3

User
.define_window(:number_window).partition_by(:number, order_by: { id: :desc })
.select(:id, :name)
.select_window(:row_number, over: :number_window, as: :row_id)
.select_window(:first_value, :name, over: :number_window, as: :first_value_name)
#=> [
 #  { id: 1, name: "Alice", row_id: 2, first_value_name: "Randy" }
 #  { id: 2, name: "Randy", row_id: 1, first_value_name: "Randy" }
 #  { id: 3, name: "Bob",   row_id: 1, first_value_name: "Bob" }
 # ]
 #

Query Output

SELECT "users"."id",
        "users"."name",
        (ROW_NUMBER() OVER number_window)      AS "row_id",
        (FIRST_VALUE(name) OVER number_window) AS "first_value_name"
FROM "users"
WINDOW number_window AS (PARTITION BY number ORDER BY id DESC)

License

The gem is available as open source under the terms of the MIT License.