Why do most functions pattern match structs on the left hand side?

Noob question. But I was just wondering, is there any advantage to pattern matching a struct on the left hand side vs right hand side?

For e.g

defmodule Test do
  defstruct name: nil

  def fun1(x = %Test{name: name}) do
    IO.inspect(name)
  end

  def fun2(%Test{name: name} = x) do
    IO.inspect(name)
  end
end

It seems like both are equivalent, but the style of “struct on left hand side” is more common.

Here’s some other instances of what I mean

2 Likes

AFAIK It’s just a stylistic choice.

1 Like

You are right, they are functionally equivalent.

I tend to use the left hand side because when I’m scanning code, its the pattern match I want to focus on first. The binding is the second consideration - especially when there are multiple function heads matching on different constructs.

10 Likes

Another reason why it’s likely more common is that it keeps things consistent.

When pattern matching outside a function head, the variable is always on the right of the = match operator since the other way around is for variable assignment

defmodule Test do
  defstruct name: nil

  def fun(x) do
    %Test{name: name} = x
    IO.inspect(name)
  end
end
13 Likes

This is the best reason I think :slight_smile:

More specifically, the variables on the left side get bound (unless you use the pin operator to indicate you want to use an existing variable) based on the data on the right. In a function head, the variables on both sides are bound based on the argument to the function.

4 Likes

+1 to this.

One other consideration: the x = %Test{name: name} shape looks similar to how other languages declare default arguments, versus the struct-on-the-left one which always means “pattern-match”.

5 Likes

Most Erlang code I’ve seen use the right hand side though. Pattern matching in the function haed has 2 purposes:

  • to de-structure the parameter. Left hand side feels more natual for this purpose.
  • to conditionally match one function clause out of several. Right hand side feels more natual for this purpose.

I guess Erlang and Elixir just pick different style convention out of the emphasizing which of the purposes are more important.

I think part of what makes it make sense in Erlang is that it doesn’t have re-binding. I feel like it would almost make more sense in Elixir if you had to use a pin to make if you want to put it on the left: def foo(^user = %User{}) since it’s sorta like, “this already has a value”. To be clear, I’m not suggesting that’s how I think it should actually be.

2 Likes

Just to be different I prefer in Erlang to use the left hand side. To me it looks more like a pattern match which is what I am doing. :smile:

6 Likes