Hi,
I spent a few hours to fix a bug where I was using an integer in place of a range
for i in N
do_something_with_i(i)
end
instead of
for i in 1:N
do_something_with_i(i)
end
I wish i in 3 wasn’t a valid expression, but it is…
Maybe some iDE hint could save me some time ? VSCode extension , other ?
VS Code should already warn you in that case:
It’s obviously quite basic, but at least addresses the literal case. Shouldn’t be too hard to support the case mentioned here too, probably.
I do not get any warning for this function:
function toto(N::Int)
for i ∈ N
println("i=$i")
end
end
Maybe, I forgot to activate specific options in vscode ?
Edit : You already said that it only catches the litteral case. Thanks.
The non litteral case would be useful 
Yeah, this only works for literal values in the loop iterator. So i in 1 will show the warning, but i in N won’t even if we could know that N is an Int. That case is much harder to do reliably, although your case should be ok.
This is why my preferred style (which I seem to be alone in, but c’est la vie) is to always use for i = 1:n when the inferrable is a literal range object and always use for x in itr when the intervals is anything besides a literal range. This uses the syntactic choice between = and in as an extra bit communicating user intent. If you accidentally write for i = n you could get a warning that you should either change it to for i = 1:n if you wanted to iterate a range or for i in n if you wanted to iterate n as a collection.
I am afraid I do not get your point 
function titi(N::Int)
for i = N
println("i=$i")
end
end
does not trigger any warning. And in the for i = case I cannot even wish that the in operator was undefined for scalar RHS.
Super Cool !
Will these hints be in the next release ?
I think Stefan’s point was
(emphasis added)
Naturally
Yes, it would be a linter warning, but I haven’t convinced anyone. But it’s simple enough to check for that you can even do it with a regex.
If you would like to have a warning, you could do this:
function Base.iterate(x::Integer)
@warn "Iterating over an integer (value: $x) - did you mean to use a range like 1:$x?"
# Actually, integers iterate once, yielding themselves
return (x, nothing)
end
function Base.iterate(x::Integer, state)
@warn "Iterating over an integer (value: $x) - did you mean to use a range like 1:$x?"
return nothing # Original behavior: iterate once
end
So that is the orgiginal implementation of iterators for numbers and now it just throws a warning. That has the benefit over error that it doesnt break packages which use them.
Neat.
“gentle” kind if type piracy ?
This breaks parts of the REPL in Julia 1.12.3:
randn(5) (or sin(5) , or sin(5) without the trailing space), don’t press Enter yet... The result should be randn(5) .julia> randn(5) .┌ Warning: Iterating over an integer (value: 1) - did you mean to use a range like 1:1?
└ @ Main REPL[1]:2
┌ Warning: Iterating over an integer (value: 1) - did you mean to use a range like 1:1?
└ @ Main REPL[2]:2
julia> sin(5) .┌ Warning: Iterating over an integer (value: 1) - did you mean to use a range like 1:1?
└ @ Main REPL[1]:2
┌ Warning: Iterating over an integer (value: 1) - did you mean to use a range like 1:1?
└ @ Main REPL[2]:2
Ahh I get what you mean.
Something like this randn(5).? But what is the . supposed to do?
I meant “input sin(5) . in the same line”, see edit to my post.
I tried to trigger some weird behavior with matrix multiplication, but everything worked fine. Then I tried elementwise multiplication and wanted to input randn(5) .* randn(5), but got the warnings in the middle of writing randn(5) ..
Ahh I get what you mean.
Well I guess that is to be expected when you do this kind of type piracy 
You will get the warnings everywhere, also in other packages. Even the REPL I guess 
