cat /dev/brain |

Let assignments in list comprehensions

published on Saturday, March 11, 2017

List comprehensions and its kin (i.e. dict/set comprehensions or generator expressions) are a powerful means to write functions returning lists and such in a concise and abstract, functional style.

One thing that seems to be missing in python's list comprehensions is the ability to assign names to intermediate values. This can be useful to avoid recomputation of expensive functions or to make the code easier to read (e.g. when expressions become to lengthy) – without introducing separate functions.

It turns out, although this feature is not supported natively, it can be emulated easily, even though it is not quite as beautiful as the let assignments in Haskell.

To avoid recomputation of f(x) in list-comprehensions such as this:

[h(x, f(x))
 for x in X
 if g(x, f(x))]

You would like to have inline assignment to variables. But this doesn't exist in python:

[h(x, y)
 for x in X
 let y = f(x)
 if g(x, y)]

However, you can write instead:

[h(x, y)
 for x in X
 for y in [f(x)]
 if g(x, y)]

In fact, this is nothing new. Haskell's list-comprehensions are just syntactic sugar for the list monad. With this perspective, Haskell's let assignments can be implemented using the return function of the list-monad – which means putting a value in a minimal list-context, i.e. return v = [v].

Know your options

Of course, there are more ways of writing the above function.

Imperative style (ugh…):

l = []
for x in X:
    y = f(x)
    if g(x, y):
        l.append(h(x, y))


[h(x, y)
 for x, y in [(x, f(x)) for x in X]
 if g(x, y)]

If X does not get consumed when iterating over, you can write:

[h(x, y)
 for x, y in zip(X, map(f, X))
 if g(x, y)]

# or even:
map(h, filter(g, zip(X, map(f, X))))

If g and h are independent of x, this becomes simpler:

 for y in map(f, X)
 if g(y)]

# or even:
map(h, filter(g, map(f, X)))

This entry was tagged functional, idiom, programming and python