No isnull for arrays?

[tlnagy]

Base doesn't have any way of checking which positions in an array have valid values.

a = Nullable{Float64}[]
for i in 1:10
    push!(a, Nullable{Float64}())
end
a[1] = 2.0
isnull(a)

while the following works

map(x -> isnull(x), a)

[johnmyleswhite]

What do you envision isnull(a) meaning?

[tlnagy]

Returning a boolean array with the values that are null. Something
similar to how isna already operates.

On Thu, Apr 28, 2016, at 17:56, John Myles White wrote:

What do you envision isnull(a) meaning?
—
You are receiving this because you authored the thread.
Reply to this email directly or view it on GitHub[1]

Links:

1. No isnull for arrays? #16112 (comment)

[johnmyleswhite]

So it would be vectorized. I think that's not the best path forward for two reasons:

• We want to remove vectorized functions, not add more of them.
• It's actually very reasonable to imagine that isnull(a) should return a Boolean.

[andreasnoack]

map(isnull, a)

[tlnagy]

Thanks everyone for weighing in. I guess with #13412 merged, using an anonymous function will be plenty fast.

[nalimilan]

@tlnagy As @andreasnoack said, you don't need an anonymous function at all, so it's already fast in 0.4. Also, with the syntax proposed at #8450, you might be able to write it as isnull.(a) at some point.

[tlnagy]

@nalimilan Good catch. I thought that used an anonymous function internally, but now I realize that doesn't make too much sense. Thanks!