Maybe I'm confused about what you mean by "using globals". I assumed you
mean just accessing a global in any way from a function. So, for example,
this would be using a global:

module M
  foo = 1
  bar() = foo
end


But this has no side effects since the global isn't modified. Are you
talking about assigning to a global from inside a function? If so, doesn't
that *require* the global declaration? Or are you talking about mutating
the content of a global – which doesn't change it's binding. E.g. something
like this:

module N
  a = [1]
  b(x::Int) = push!(a,x)
end



On Wed, Jun 11, 2014 at 8:28 AM, Tony Fong <[email protected]> wrote:

> Yes, I thought I was catching the right error, but later realized that I
> was being too militant assuming that a proper function has no "closure".
> Obviously there is, and the default closure has all the globals.
>
> However, I think using globals in a function can be surprising for a user
> that assumes no side effect, so I could "downgrade" the lint message level
> from ERROR to FYI. What do you think?
>
> Tony
>
>
> On Tuesday, June 10, 2014 10:22:42 AM UTC-4, Stefan Karpinski wrote:
>
>> This is really nice work. In the future, I'd really like to move bits of
>> TypeCheck and this sort of linting into base Julia, maybe invoked with a -w
>> flag.
>>
>> Regarding this:
>>
>> Using globals in function without declaring them (This one isn't an
>>> error, but I personally prefer explicit declaration of globals dependency
>>> inside functions for readability.)
>>
>>
>> I'm not sure how you're dealing with this, but this would, as stated,
>> warn every time you call a function the way I'm reading it. Are you
>> specifically not warning for globals that are used as call heads? I
>> certainly wouldn't want to have to declare as global every function I'm
>> going to use.
>>
>>
>> On Tue, Jun 10, 2014 at 7:18 AM, Tony Fong <[email protected]> wrote:
>>
>>> It's a great idea! I'll look into it. Thanks for the pointer.
>>>
>>>
>>> On Tuesday, June 10, 2014 5:50:27 AM UTC-4, René Donner wrote:
>>>>
>>>> I don't know how feasible it is, but a (perhaps optional) inclusion of
>>>> the functionality in https://github.com/astrieanna/TypeCheck.jl would
>>>> be great!
>>>>
>>>>
>>>>
>>

Reply via email to