comprehension without allocating · helpdesk (published)

A comprehension by default creates an vector.
Sometimes though I use comprehension as an one-line mutating procedure.
e.g.:

a = collect(1:100)

# my one-liner mutating comprehension
deletespec!(a,v) = [deleteat!(a, x) for x in v];

Clearly here, I am not interested in the result.
I just want the mutating operation to happen.

Calling it will return a vector, which means something got allocated, that I didn't really need

julia> deletespec!(a, [1,2,3])
[[2, 4, 6, 7, 8, 9, 10, 11, 12, 13  …  91, 92, 93, 94, 95, 96, 97, 98, 99, 100], [2, 4, 6, 7, 8, 9, 10, 11, 12, 13  …  91, 92, 93, 94, 95, 96, 97, 98, 99, 100], [2, 4, 6, 7, 8, 9, 10, 11, 12, 13  …  91, 92, 93, 94, 95, 96, 97, 98, 99, 100]]

Is there a way to produce comprehensions that return nothing, i.e. allocate nothing ?
Should I even worry about this performance-wise ?

PS I know what I am doing to delete multiple indices is wrong and there are better ways to do that, but please don't focus on this.

Filippos Christou (Jan 12 2023 at 15:16):

A comprehension by default creates an vector.
Sometimes though I use comprehension as an one-line mutating procedure.
e.g.:

a = collect(1:100)

# my one-liner mutating comprehension
deletespec!(a,v) = [deleteat!(a, x) for x in v];

Clearly here, I am not interested in the result.
I just want the mutating operation to happen.

Calling it will return a vector, which means something got allocated, that I didn't really need

julia> deletespec!(a, [1,2,3])
[[2, 4, 6, 7, 8, 9, 10, 11, 12, 13  …  91, 92, 93, 94, 95, 96, 97, 98, 99, 100], [2, 4, 6, 7, 8, 9, 10, 11, 12, 13  …  91, 92, 93, 94, 95, 96, 97, 98, 99, 100], [2, 4, 6, 7, 8, 9, 10, 11, 12, 13  …  91, 92, 93, 94, 95, 96, 97, 98, 99, 100]]

Is there a way to produce comprehensions that return nothing, i.e. allocate nothing ?
Should I even worry about this performance-wise ?

PS I know what I am doing to delete multiple indices is wrong and there are better ways to do that, but please don't focus on this.

mbaz (Jan 12 2023 at 15:18):

Filippos Christou (Jan 12 2023 at 15:31):

you mean writing (deleteat!(a, x) for x in v) instead of [deleteat!(a, x) for x in v]. It would work but again it feels like there is some effort going somewhere (e.g. constructing the generator) which is unneeded.

Filippos Christou (Jan 12 2023 at 15:32):

Filippos Christou (Jan 12 2023 at 15:33):

the mutation will happen only when I start iterating the elements of the generator.

Brian Chen (Jan 12 2023 at 15:59):

Filippos Christou (Jan 12 2023 at 16:09):

Yes, this is a good suggestion.. However nested foreach will get ugly. Comprehensions are easier to follow in that regard..

Gunnar Farnebäck (Jan 12 2023 at 18:58):

Nested foreach using the do construction shouldn't be that ugly, or am I missing something?