As I was investigating a "package require" issue, I came across the
following surprising, to me, behavior. Needless to say, I don't
understand it.

I expected the following code to instantiate the variable "myGlobal" in
the global namespace:

namespace eval Wrapper {
global myGlobal
if { ![info exists myGlobal] } {
set myGlobal [namespace current]
}
puts "myGlobal: $myGlobal"
foreach varName {myGlobal ::myGlobal ::Wrapper::myGlobal} {
puts "Exists? $varName: [info exists $varName]"
}
}

but the output:

myGlobal: ::Wrapper
Exists? myGlobal: 1
Exists? ::myGlobal: 0
Exists? ::Wrapper::myGlobal: 1

says that "myGlobal" instead lives in the "Wrapper" namespace.

If the above code is preceded by this statement:

set myGlobal XXX

we get this output:

myGlobal: XXX
Exists? myGlobal: 1
Exists? ::myGlobal: 1
Exists? ::Wrapper::myGlobal: 0

which shows "myGlobal" is defined in the global namespace.

Finally, if we change the namespace command in the first example to a proc:

proc Wrapper {} {
global myGlobal
set myGlobal XXX
}
Wrapper
puts "Exists? myGlobal [info exists myGlobal]"

"myGlobal" is defined in the global namespace which is expected.

I get that in examples 1 and 3 "myGlobal" does not exist initially but
the "namespace" and "proc" commands yield different results. So, the
question is why the difference?

TIA for the coming education!

-mike

the first line of the global man page says: This command has no effect unless executed in the context of a proc body.

If you remove the "global myGlobal: line from your example, there should be no change in the results.
The variable is created (or not) in the normal way by the following set command.

Dave B

On 03/17/2023 10:44 AM, dave bruchie wrote:
> the first line of the global man page says: This command has no effect unless executed in the context of a proc body.
>
> If you remove the "global myGlobal: line from your example, there should be no change in the results.
> The variable is created (or not) in the normal way by the following set command.
>
> Dave B
Hi Dave,

Thanks for the reply. I had seen the clear statement in that manpage
but I guess it didn't register.

So, it looks like you don't need to use a "variable" command in a
namespace in order to create that variable in the namespace. The reason
for using one, outside of a proc, is to ensure the variable is created
in the namespace just in case that variable also exists in the global
namespace.

The method used to resolve variable names is explained in the "Name
Resolution" section of the "namespace" manpage.

-mike

On Friday, March 17, 2023 at 11:32:48 AM UTC-4, Michael Soyka wrote:
> just in case that variable also exists in the global

I think globals become much more manageable and readable if you just enforce a style of always referring to them with the fully qualified name, such as ::VAR and also use all caps. The global command is unnecessary and hinders readability in my opinion.