[cfe-dev] [PATCH]: Parsing for C++ classes

[Chris Lattner]

On Jun 24, 2008, at 11:07 AM, Argiris Kirtzidis wrote:
> Chris Lattner wrote:
>>
>> +  /// LexedMethodStacks - During parsing of a top (non-nested) C++  
>> class, its
>> +  /// inline method definitions and the inline method definitions  
>> of its nested
>> +  /// classes are lexed and stored here.
>> +  /// A new lexed methods stack is pushed for local classes in  
>> inline methods.
>> +  std::stack< std::stack<LexedMethod> > LexedMethodStacks;
>>
>> Out of curiousity, why use a std::stack instead of std::deque or  
>> std::vector?
>
> About the outer stack, it's the last "MethodTokensForOneClass"  
> pushed that is going to be used for adding lexed methods, so a stack  
> seems appropriate.

Ok

> About the inner stack, there won't be a difference whether it's a  
> stack, list or deque, although std::vector might not be as  
> appropriate since there's no need to have LexedMethods in contiguous  
> memory.
> Since a std::stack was already brought in for the outer stack, I  
> figured I could use it again to avoid bringing in an additional  
> container.

std::stack is actually not a container, it is an adaptor that wraps  
std::deque (by default).

>> These really push one 'nested class' worth of methods, right?  It  
>> might be better to name these "PushClassStack" or something.
>
> Ok to both suggestions.
> Just a sidenote, I think these containers and methods will need  
> refactoring to accomodate other types of "delayed parsing", like  
> default parameters. So they are subject to change; probably a new  
> class will be added for all this.

Makes sense, using a new struct (even with one member) is even better  
than a typedef!

>> In Parser::ConsumeAndStoreUntil, do you think it would make sense  
>> to specialize it for T=r_brace, or do you expect other clients?
>
> This method calls itself recursively for braces and parentheses.

Ok!

-Chris