[cfe-commits] r74802 - in /cfe/trunk: include/clang/Index/ include/clang/Index/Entity.h include/clang/Index/EntityHandler.h include/clang/Index/IndexProvider.h include/clang/Index/Program.h include/clang/Index/TranslationUnit.h lib/CMakeLists.txt lib/Index/ lib/Index/CMakeLists.txt lib/Index/Entity.cpp lib/Index/IndexProvider.cpp lib/Index/Makefile lib/Index/Program.cpp lib/Index/ProgramImpl.h lib/Makefile

[Chris Lattner]

On Jul 12, 2009, at 3:49 PM, Argyrios Kyrtzidis wrote:

> Hi Chris,
>
> On Jul 12, 2009, at 3:35 PM, Chris Lattner wrote:
>
>>
>>> +class Entity : public llvm::FoldingSetNode {
>>> +public:
>>
>> Why should Entity's be FoldingSetNodes?  There are two reasons to use
>> a folding set (instead of a DenseMap<pair<Entity*,void*>, Entity*>):
>>
>> 1. If you want to derive from Entity, so you may have heterogenous
>> keys indexing the map.
>> 2. If the key data is large.
>>
>> I don't think either of these applies here, should Entity's be  
>> uniqued
>> with a DenseMap instead of a FoldingSet?
>
> The key data is simple now but eventually it will grow, especially  
> for supporting C++.
> e.g., overloaded functions will be associated with different  
> entities, so the key will include info about the parameter types.
> And templates will introduce more complexity as well.

Ok, fair enough.

-Chris