\circ and parantheses

The main thing that causes me to not use the \circ operator in julia as much as i do the mathematica equivalent (they use @) is, that I need to wrap everything in parantheses.
I can see that this relates nicely to mathematical notation but what is the most common usecase for me, is that i wrote a function call like sqrt(-2) and then i notice “ahh… I need an abs there”.
If i were to just insert a \circ abs between sqrt and (, that would go easily, but putting additional parantheses before sqrt and after abs killing the gain.
Long story short: couldn’t sqrt \circ abs(-2) be parsed or lowered as (sqrt \circ abs)(-2)? That would also make it more similar to its brother |>.

1 Like

It could, but that would be a breaking change so it couldn’t be made until Julia 2.0.

1 Like

I’m really surprised that it doesn’t do that.

Whats the current behaviour useful for?

abs(-2) could return a function?

I don’t see \circ as that useful for cases like these, where sqrt(abs(-2)) is at least as good. It’s nice for creating anonymous functions, though. sqrt ∘ abs is much better than x->sqrt(abs(x)).


Ah, to pass to map and such. Yeah I see.

Why would that be breaking? The current syntax would be still valid just not mandatory.

Because it changes the meaning of existing code.

1 Like

@DNF already suggested this, but let me elaborate: the current behavior is useful for working with higher order functions. The docstring for mentions

julia> map(uppercase∘first, ["apple", "banana", "carrot"])
3-element Array{Char,1}:

and one can find other examples in the standard library, eg in Pkg

        for (name, uuids) in sort!(collect(unresolved), by=lowercase ∘ first)

It is emphatically not for (f∘g)(x), for that f(g(x)) is much more readable. The same applies to ! for functions, eg !f is fine but !f(x) is preferred to (!f)(x).

@BeastyBlacksmith: whenever you are wondering about a design choice or a use case for something in Julia, it is good practice to look at usage in Base and the standard library. Not everything is perfect, but many things are the result of a conscious decision and form a coherent whole.

1 Like

You can easily define your own operator which does this:

julia> (<|)(f, x) = f(x)
<| (generic function with 1 method)

julia> sqrt <| abs <| 2