Understand Clojure Transducers 1

2 minute read

In this post, we will define what is a transducer.

First, let’s take a closer look at map.

(map inc [1 2 3])

A straightforward way to explain what has happened is this: Increment each item by one in a list.

At first thought, this sounds like one step operation.

Now let’s change the requirement a bit. We want to increment each element by one and then sum them together.

This is very similar to what map is except it needs more processing after.

Let’s do it by using a reduce.

(reduce (fn [acc n] (+ acc n 1)) 0 [1 2 3])

Now this looks like a two step operation:

  1. Increment each element by one
  2. Sum them together

Emm… Now if we rethink the map example, it can actually be expressed as a two step operation as well:

  1. Increment each element by one
  2. Create a new list out of the incremented elements

To make it more clear, let’s also write map using reduce:

(reduce (fn [acc n] (conj acc (inc n))) [] [1 2 3])

Now realize that the first step is the same. Can we somehow extract that out, so it can be reused in both examples? The ideal format would be something like:

((mapping inc) *second-step*)

This (mapping inc) is a function that takes a second-step and returns a reducing function that can be used by reduce. Also note that (mapping inc) is only saying map the inc function into each item, and doesn’t specify what should we do next. The second-step defines what to do with the result of increment every item by one.

(defn mapping [f]
  (fn [second-step]
    (fn [acc n]
      (second-step acc (f n)))))

Now what the mapping function returns is called a transducer. It takes a second-step function and returns a reducing function. Also note that the second-step function is also a reducing function. Thus, we can define a transducer to be a function that takes a reducing function and returns a new reducing function.

~To be continued~

comments powered by Disqus