Re: [antlr-dev] @init for global scopes?

Thu, 22 Jan 2009 11:38:20 -0800

Sam Harwell wrote:

I found a solution to both the variable hoisting and scope consistency problems related to prediction with predicates. The solution is clean but non-trivial, so I’m putting together some detailed documentation on my rationale, the approach I used, and the explicit preconditions that, when met, guarantee a grammar will not run into hoisting or consistency problems in the generated code. An explicit initialization section for attribute scopes is necessary for it to work.

 

For the C# port, I’m going to implement this as @scopeinit for now and see how it turns out.

 

OK - well do not forget, within your solution, that not all targets have things like constructors, finalizers/destructors, reference counting, exceptions and garbage collection ;-)

Jim

Sam

 

From: [email protected] [mailto:[email protected]] On Behalf Of Jim Idle
Sent: Thursday, January 22, 2009 12:27 PM
Cc: [email protected]
Subject: Re: [antlr-dev] @init for global scopes?

 

I am not sure why this is of any advantage? What limitation are you trying to get around? Choosing where to initialize a global scope or something? This change will generate work for all other targets and the syntactic change potentially generates confusion, especially if it is left at @init. In C, I have some macros such as SCOPE_TOP(sname) for action code - perhaps I implemented these to get around some limitation that you are proposing a fix for? I no longer remember ;-)

Jim


_______________________________________________
antlr-dev mailing list
[email protected]
http://www.antlr.org/mailman/listinfo/antlr-dev

Reply via email to