FWD » Core Development » Base Language

4GL allows two vars with the same name, if this construct is encountered:

def new global shared var ch-1 like book.book-title. /* this is used */
def shared var ch-1 as int format "x(10)". /* this is ignored */

This is possible only if the first definition is from a "DEFINE NEW GLOBAL SHARED" statement (and only this). All subsequent definitions for var with the same name (using DEFINE SHARED OR DEFINED VAR or anything else), are ignored.

The code looks good. I only have 2 questions:

1. Is there any difference in behavior when the non-global define variable is inside a nested scope (e.g. an internal procedure). In such cases, we would expect the global var to be temporarily hidden but the local var and would not think of that as a duplicate.

2. What happens in the case where the global var is created after the non-global var?

1. Is there any difference in behavior when the non-global define variable is inside a nested scope (e.g. an internal procedure). In such cases, we would expect the global var to be temporarily hidden but the local var and would not think of that as a duplicate.

Yes, you are correct, when inside a procedure/function, it should hide the global var.

2. What happens in the case where the global var is created after the non-global var?

Compile-time error.

Please put a comment into the variable_definitions.rules about the order of the definitions and that we deliberately ignore that case since it fails in the 4GL with a compile error. That way, future readers will know that it was intentional.

The update has passed conversion regression testing. Note that even MAJIC had two cases of duplicated vars, in two files, which looked like this:

def new global shared var ch-1 as char. /* this is used */
def new shared var ch-1 as char. /* this must be ignored */

This didn't cause compile problems because one was an instance field for the class, and the other was emitted as an instance field for the Block instance associated with the external program call.

Check it in and distribute it.

Committed to bzr revision 10255.

I need to rewrite this at the parser side, as the duped var gets ref'ed by the parser:

def new global shared var i as int. /* 1st def */

def shared var i as int. /* 2nd def */

procedure proc0.
   message i. /* the parser links this with the 2nd def. as 2nd def is hidden, it will no longer emit (as promoted) and 1st def will not be promoted, becaused is not ref'ed*/
end.

This has regressions, one of them relaed to this, as 4GL (v9-only I think) allowed this:

def var ch as char label "adsadasd" as log init no.

if ch then message ch. /* ch is not char, but logical */

In v10, the code above does not compile.

Fixed a problem (I was using varDict.getSymbol instead of varDict.getSymbolAtScope(name, 0) to search only in the current scope). Still working on the issue at note 14.

MAJIC has the case at note 14 in only one file. I've checked the history of this file and this change was introduced after they switched to P2J server. So, the conclusion is that this is a bug in the 4GL sources, not another 4GL stupid feature.

I'll edit the file and put it in a separate task.

I'm not sure why you switched from a Set to a Map in fixups/variable_definitions.rules (did you plan to use it and then later on, find you didn't need it)?

Otherwise, I don't see a problem. Go ahead and check it in and distribute it.