Problems with local scope

[krychu]

Hello,

Consider two examples:

x = 1
for i = 1:10
    x = 10
end
println( x )

out: 10

x = 1
function a()
    x = 10
end
a()
println( x )

out: 1

According to the docs both the for loop and the function definition should introduce a new scope with an inner variable x. But this doesn't seem to be the case for the for loop. I'm sure I'm missing something (?)

Here's the relevant snippet from the docs:

"As a consequence, the only way to assign to a global variable inside a non-top-level scope is to explicitly declare the variable as global within some scope, since otherwise the assignment would introduce a new local rather than assigning to the global."

Mac OS, Julia 0.3.9

[yuyichao]

This is the documented correct behavior.

See http://julia.readthedocs.org/en/release-0.3/manual/variables-and-scoping/

In particular, the following text and the examples below it.

An assignment x = y introduces a new local variable x only if x is neither declared global nor introduced as local by any enclosing scope before or after the current line of code.

[yuyichao]

Actually when I read the document on the variable scope just now, it isn't very clear on what's the difference between a let block and a function in terms of scope. Maybe it can be improved?

@krychu also note that you didn't call the function in your second example (although it doesn't make a difference)

[krychu]

Sorry I forgot to add a() above (I've added it now). I've read the docs multiple times and it is not clear. Both for and function are listed in the docs as constructs introducing new scopes. So all further scope rules described in the docs should apply to them equally.

The docs say:

An assignment x = y introduces a new local variable x only if x is neither declared global nor introduced as local by any enclosing scope before or after the current line of code.

which would suggest for and function will inherit the outer x.

Then the docs say:

When a variable is introduced into a scope, it is also inherited by all inner scopes unless one of those inner scopes explicitly overrides it.

and this:

... the only way to assign to a global variable inside a non-top-level scope is to explicitly declare the variable as global within some scope, since otherwise the assignment would introduce a new local rather than assigning to the global.

which would suggest that for and function will introduce a new local variable x.

[krychu]

Also this part is confusing

[...] Due to the special identification of the prompt’s scope block with the global scope, it is not necessary to declare global y inside the loop. However, in code not entered into the interactive prompt this declaration would be necessary in order to modify a global variable.

which doesn't seem to be true as this script (run from outside the interactive prompt):

y = 0
for i = 1:1; y = 10; end
println( y )

modifies the global y without using global declaration within the loop.

I'm pointing these out so that the docs can be made clearer for newcomers (like me).

[mauro3]

This is soft vs hard scope of functions which is not described in the docs, its issue is #9955. And also have a look at #10559. And if anyone improves the scope docs, the meaning of global should be defined: i.e. that a global is a module-wide global. There are no truly global globals (which is different from Matlab).

[krychu]

Thanks!

[JeffBezanson]

Closing as dup of #9955.