add partially-applied versions of ∋, ∉, ∌, occursin

[JeffBezanson]

in(x) is really useful, but sometimes I want this for the other argument. E.g.

julia> findall(∋(0), [-1:1, 1:10, -2:20])
2-element Vector{Int64}:
 1
 3

for finding ranges that contain 0. It's not great that this requires unicode --- is there another name for this already? If not, should we add one?

[KristofferC]

ni!

[Keno]

inleft?

[StefanKarpinski]

I actually like calling it ni 😬

[JeffBezanson]

Other languages may try to reference it, but WE ARE THE DEVS WHO SAY "NI"!

[mcabbott]

There is findall(contains('a'), ["cat", "dog", "zebra"]) but only for strings. For which findall(in("cat"), 'a':'e') works but occursin("cat") does not.

[JeffBezanson]

Yeah we should also add occursin(x).

[nickrobinson251]

we should also add occursin(x).

I think this criteria is the reason we don't have occursin(x) (and one of the motivations for adding the argument-flipped contains function was that it fit the criteria for having a partially-applied one-arg method).

[StefanKarpinski]

These definitions would both fit the criteria:�

• occursin(haystack) would be needle -> occursin(needle, haystack)
• contains(needle) would be haystack -> contains(haystack, needle).

[jw3126]

ni!

We should add more pythonic then python to the greed of julia.

[stevengj]

Needs docs and/or NEWS?

[JeffBezanson]

Added all the things.

[nickrobinson251]

oh... it's really gonna be called ni 😆 I don't want to bikeshed this too much, but it does seem a bit funny... since it's not a word. Maybe that's fine, because in practice ∋ will probably be used instead, and this is just for when an ascii name is required for some reason?

I didn't want to raise a naming question without an alternative suggestion, so... How about has? e.g. has([1, 2, 3], 2), any(has(0), (-1:1, 1:3))? it's still very short, basically another word for contains, probably fine to claim as name (i've never seen a has function before), and i think consistent enough with functions names with a has* prefix, like haskey, hasvalue etc.

(I don't feel strongly, so don't hold up the PR over this if everyone else is in favour of ni ...as i say, i doubt ni will appear in too many codebases, even if ∋ gets wider use)

[stevengj]

ni is cute, but it's not a great antonym for in — it would be pretty confusing to most people if they saw that appearing in code, I think. And since it's not infix, it's not really symmetrical with in (and we can't make it infix because that would involve introducing a breaking new keyword).

[oxinabox]

We could always say the ASCII form is xs->in(x,xs)
Like we do for not in, and used to do for compose

[JeffBezanson]

Ok, I don't insist on ni. has is ok, or I could just delete ni for now to get the rest of this in.

[StefanKarpinski]

or I could just delete ni for now to get the rest of this in.

Let's just do that for now and open a separate issue for whether we need an ASCII name for ∋.