Boxed types for CommonC++

[Marc Boris Dürner]

>        It sounds interesting. I suppose this relates the boxed types 
>and MemoryChunk template you comment on another post? 
 
Yes, but not neccessarily. For the boxed types we also would need to
implement float, 
char etc... 
It just seemed a good analogy to low level C-programming where you use
uint8_t for a 
byte and the larger int types for short series of bytes. The union is a fast
way of extracting 
bytes. 
 
   union chunk_u {  
      TYPE chunk;  
      uint8_t bytes[sizeof(TYPE)/sizeof(uint8_t)];  
   } _chunk; 
 
The union doesnt make much sense though if TYPE is a float. So would it work
to have a 
template class BoxedType that boxes everything and then use derivatives of
BoxedType 
for MemoryChunk? So BoxedType would be identical to MemoryChunk, just
without the 
union. 
 
BoxedType <- UInt8 
                  <- UInt16 
                  <- UInt32 
                  <- UInt64 
 
typedef MemoryChunk<UInt8> Byte; 
typedef MemoryChunk<UInt16> Word; 
etc... 
 
Marc 
 
On Mon, Dec 29, 2003 at 03:20:19AM +0100, "Marc Boris Dürner" wrote: 
> Hi,  
> For the implementation of classes for numeric IP addresses I am trying to 
> treat an IP  
> address as a series of bytes. I have a base class NetAddress that has a 
> vector<uint8_t>  
> as a member and is so usable for inet4, inet6, ipx etc... (and maybe mac 
> address as well)  
> The derived class Inet4Address for instance has a validator that can 
> validate a string as  
> input and put 4 bytes into the vector. It provides compatibility 
> getter/setter to in_addr.  
>   
> Is this a good idea?  
>   
> regards,  
> Marc  

-- 
+++ GMX - die erste Adresse für Mail, Message, More +++
Neu: Preissenkung für MMS und FreeMMS! http://www.gmx.net