Re: C++ implementation of pvData

[Marty K. <mrk...@co...>]

  On 08/02/2010 09:42 AM, jam...@di... wrote:
> Hi
> Some suggestions for discussion:
> 1) Return std::string instead of shared_ptr<std::string>, the 
> implementation of std::string is typically shared anway.


Note that the java implementation of pvData, etc relies heavily on the 
fact that:

1) A String is an immutable object.
2) A String is garbage collected.

Don't we have to provide the same semantics for the c++ implementation?

For a simple example assume that a record contains a Field object that 
is not shared with any other record. It has a fieldName, which is a 
string. If the record is removed from the database, which is possible 
with pvData, etc, then the string field must be garbage collected or 
else we leak memory.
Am I missing something about how std::string is implemented?


> 2) Don't use virtual inheritance ( Nikolay has covered that ).
>

I do not understand. Why not?

I assume that pure virtual base classes are the C++ equivalent of a Java 
interface.

Marty
> I'll wait for the mercurial repository before further comments (I've 
> asked to be added as a developer so I can do the import if required).
>