Getting to Know Pluck and Select

Let’s take some time to get familiar with pluck and select. They’re two very useful methods in ActiveRecord and using both effectively can really contribute to improving performance of your app.

Before there was pluck there was select. select was THE WAY for you to query for a single field from your database.

1
2
3
4

User.select(:id).to_a

# => User Load (0.9ms)  SELECT id FROM "users"
# => [#<User id: 12>, #<User id: 42>, #<User id: 1>, #<User id: 24>, #<User id: 200>, ...]

In this example, select creates an array of User models where only id is returned. (You need to include to_a so that the query runs, without it you’ll simply get an ActiveRecord::Relation object.)

pluck was a introduced in Rails 3.2 and with it you can perform a select and skip the overhead of building ActiveRecord models.

1
2
3
4

User.pluck(:id)

# => (0.9ms)  SELECT "users"."id" FROM "users"
# => [12, 42, 1, 24, 200, ..., 365]

The results are nearly the same from an iteration standpoint. With pluck your results are an Array of integers, instead of User models. Now I can hear you saying: “Wait just a minute there! Both queries took 0.9ms, so there’s no difference between the two methods”. And to that I would reply…

Not so!

The difference is in the cost of object construction, which Rails is hiding from you!

Let’s use Benchmark to find out the real cost of select and pluck. For each method we’ll do 5 runs, reject the lowest and highest times, then average the 3 results.

1
2
3

puts Benchmark.measure { User.select(:id).to_a }
    user    system     total        real
0.883333  0.006667  0.890000 (  1.113816)

And pluck:

1
2
3

puts Benchmark.measure { User.pluck(:id) }
    user    system     total        real
0.043333  0.000000  0.046667 (  0.089493)

I ran the two queries across 10,000 user records, and it’s easy to see that select was 91% slower than pluck. For the record, both queries take the same amount of time to run, it’s the object creation that eats up that additional time with select

This is no big secret. pluck was introduced to remove the overhead caused by Rails object creation, which was commonly seen when running a query like this: User.select(:id).map(&:id)

Something I didn’t know prior to writing this post is that you can pluck on multiple columns! (Note: that this is available as of Rails 4.0)

1
2
3

User.pluck(:first_name, :last_name, :email)
# =>   (2.2ms)  SELECT "users"."first_name", "users"."last_name", "users"."email" FROM "users"
# => [[nil, nil, "jeff@gmail.com"], ['joe', nil, "joe.brown@gmail.com"], ["Edward", "Stanza", "ed@email.com"], ...]

pluck is however slightly different than select when it comes to multiple columns. With select you can pass :first_name, :last_name, :email or [:first_name, :last_name, :email] whereas if you pass an array to pluck it’s going to barf on you – oddly, this was a deliberate choice.

This post wouldn’t be complete without talking about the use case for select. Namely when you need to run model methods on the results that come back from select. There’s no way to do that with pluck. In one simple sentence:

pluck for model values, select for model objects

That’s it for now. In the future you can look forward to regular posts on difference ways to improve the performance of your Rails application!