2. Better queries with Arel

on Jul 20, 2019

Learn how Arel can help you build complex queries without having to write raw SQL queries manually and thus keeping them database agnostic.

Source code: github.com/code-snack/002-arel

ActiveRecord is a very powerful tool. It goes a long way allowing you to organize your database access and build queries in the Ruby way. However, in some scenarios we end up having to write SQL statements in Strings because ActiveRecord just won't generate the SQL we need for us.

Under the hood, ActiveRecord uses a tool called Arel to build SQL statements. We can leverage that to build our own advanced queries.

Before we dive into coding mode, be aware that there are gonna be dozens of code examples in this episode. Feel free to pause the video to read and understand them thoroughly. You can also see those examples in the repository I'll link in the description.

Let's get to it. We have here two simple models with a simple relationship: Category and Product. Product belongs to Category. Now let's jump to the Rails console where I can show you some queries we can build with Arel using these two models.

Every ActiveRecord model class has a method called arel_table.

# => #<Arel::Table:0x00007fd18cbb1ce0
#  @name="products",
#  @table_alias=nil,
#  @type_caster=#<ActiveRecord::TypeCaster::Map:0x00007fd18cbb1d30 @types=Product(id: integer, name: string, price: float, active: boolean, created_at: datetime, updated_at: datetime)>>

This method returns an object representing that model's table in the database. With that object we can access columns like this:

# => #<struct Arel::Attributes::Attribute
#  relation=
#   #<Arel::Table:0x00007fd18cbb1ce0
#    @name="products",
#    @table_alias=nil,
#    @type_caster=#<ActiveRecord::TypeCaster::Map:0x00007fd18cbb1d30 @types=Product(id: integer, name: string, price: float, active: boolean, created_at: datetime, updated_at: datetime)>>,
#  name=:id>

# => #<struct Arel::Attributes::Attribute
#  relation=
#   #<Arel::Table:0x00007fd18cbb1ce0
#    @name="products",
#    @table_alias=nil,
#    @type_caster=#<ActiveRecord::TypeCaster::Map:0x00007fd18cbb1d30 @types=Product(id: integer, name: string, price: float, active: boolean, created_at: datetime, updated_at: datetime)>>,
#  name=:name>

With those columns we can build conditions. Let's see an example.

Look at this query. We are using a raw string here because ActiveRecord doesn't provide us a way to do this in the object oriented manner. But Arel does!

Product.where('name ILIKE ?', '%shoe%')
# SELECT "products".* FROM "products" WHERE (name ILIKE '%shoe%')

We can use matches

# SELECT "products".* FROM "products" WHERE "products"."name" ILIKE '%shoe%'

What's the benefit? Well, for starters, Arel took care of quoting the column name for us and also prefixed it with the table name to avoid ambiguous matches. Also, now we're not writing SQL inside a ruby file anymore and if we switch database providers, the new adapter will take care of generating the right SQL for us. Keep in mind that ILIKE is not the right keyword for every database.

Moving on. You can build several different conditions with Arel. We have methods for =, IN, >, >=, <, <=.

# SELECT "products".* FROM "products" WHERE "products"."name" = 'shoe'

Product.where(Product.arel_table[:name].in(['shoe', 'sneakers']))
# SELECT "products".* FROM "products" WHERE "products"."name" IN ('shoe', 'sneakers')

# SELECT "products".* FROM "products" WHERE "products"."price" > 100.0

# SELECT "products".* FROM "products" WHERE "products"."price" >= 100.0

# SELECT "products".* FROM "products" WHERE "products"."price" < 100.0

# SELECT "products".* FROM "products" WHERE "products"."price" <= 100.0

Some of them have their negative forms like != and NOT IN.

# SELECT "products".* FROM "products" WHERE "products"."name" != 'shoe'

Product.where(Product.arel_table[:name].not_in(['shoe', 'sneakers']))
# SELECT "products".* FROM "products" WHERE "products"."name" NOT IN ('shoe', 'sneakers')

There are many predications you can use with arel columns. To get a list of them all, run Arel::Predications.instance_methods. This list varies depending on the Rails version you're using.

We can also combine conditions together. If you use where clauses from ActiveRecord you know they are gonna generate you AND conditions.

# SELECT "products".* FROM "products" WHERE "products"."name" = 'shoe' AND "products"."id" = 1

If you need an OR condition, you can do that with Arel like this.

# SELECT "products".* FROM "products" WHERE ("products"."name" = 'shoe' OR "products"."id" = 1)

You can even combine ANDs and ORs.

name = Product.arel_table[:name]
# SELECT "products".* FROM "products" WHERE ("products"."name" = 'shoe' OR "products"."id" = 1 AND "products"."name" ILIKE '%sneakers%')

Now let's get fancy with functions. Have you ever had to call a database function from your Rails code? Arel can help you there!

Product.where(Arel::Nodes::NamedFunction.new('LENGTH', [Product.arel_table[:name]]).gteq(50))
# SELECT "products".* FROM "products" WHERE LENGTH("products"."name") >= 50

The objects we are passing to the where calls in all these examples are Arel::Nodes.

Arel::Nodes::NamedFunction.new('LENGTH', [Product.arel_table[:name]]).gteq(50).class
# => Arel::Nodes::GreaterThanOrEqual

# => Arel::Nodes::Equality

Nodes can be used in select and order calls too!

Product.select(Arel::Nodes::NamedFunction.new('LENGTH', [Product.arel_table[:name]]))
# SELECT LENGTH("products"."name") FROM "products"

Product.order(Arel::Nodes::NamedFunction.new('LENGTH', [Product.arel_table[:name]]))
# SELECT "products".* FROM "products" ORDER BY LENGTH("products"."name")

In case of ordering, you can also call desc on a node to reverse it:

Product.order(Arel::Nodes::NamedFunction.new('LENGTH', [Product.arel_table[:name]]).desc)
# SELECT "products".* FROM "products" ORDER BY LENGTH("products"."name") DESC

The next feature we're gonna see allows us to create better joins. Have you ever had to do something like this to get a LEFT OUTER JOIN?

Category.joins('LEFT OUTER JOIN products ON products.category_id = categories.id')
# SELECT "categories".* FROM "categories" LEFT OUTER JOIN products ON products.category_id = categories.id

Arel can do this for us too!

categories = Category.arel_table
products = Product.arel_table
# SELECT "categories".* FROM "categories" LEFT OUTER JOIN "products" ON "categories"."id" = "products"."category_id"

You can also combine conditions here:

# SELECT "categories".* FROM "categories" LEFT OUTER JOIN "products" ON "categories"."id" = "products"."category_id" AND "products"."name" IS NOT NULL

Now that you've learned Arel, let's use what we saw here to refactor some scopes that I added to our models.

First let's improve these price filters to use Arel.

Now for the search scope we'll convert it into a class method because it's gonna get bigger. Remember: any class level methods in an ActiveRecord model can be used as a scope.

Now it's time for the active scope.

In this case in particular, I'm going through another model's arel table. Usually when I see that happening, either with arel_table or just where clauses I recommend creating a scope inside that model, in this case: Category. As you can see here we already have an active scope inside category. How can we reuse it in the Product model? Simple, we just use another ActiveRecord trick: the merge method.

This method allows us to combine together conditions from two different ActiveRecord relations.

And there you have it: no more raw SQL in our models.

I know this code is verbose and a lot longer than the previous version using a String, but it's a lot more reusable and high level. If the verbosity bothers you, take a look at the gem called arel_helpers. It provides several methods to use Arel with less code.

And that's it for this episode. Thank you for watching and I hope you enjoyed this snack!


Sign In or Sign Up to comment.