Hello, everyone,

I am wondering what is the hash function that is used in
hash_map/hash_set.  Can I replace it with my own
hash function?  Any comments on this? 

Thanks!

-Andy

yuyang08@gmail.com wrote:

> 
> I am wondering what is the hash function that is used in
> hash_map/hash_set.  Can I replace it with my own
> hash function?  Any comments on this? 
> 

By default, hash_map and hash_set use

namespace std {
namespace tr1 {
template <class Ty> struct hash
	: unary_function<Ty, size_t>
{
size_t operator()(Ty val) const;
};
} }

This template is specialized for bool, all integer types, all 
floating-point types, pointers, std::string, and std::wstring.

You can specialize it for any of your own types.

You can also use your own type as an argument to hash_map or hash_set.

For details, see section 5.6 of my book, "The C++ Standard Library 
Extensions: a Tutorial and Reference."

Pete Becker wrote:
> yuyang08@gmail.com wrote:
> >
> > I am wondering what is the hash function that is used in
> > hash_map/hash_set.  Can I replace it with my own
> > hash function?  Any comments on this?
> >
> By default, hash_map and hash_set use
>
> namespace std {
> namespace tr1 {
> template <class Ty> struct hash
> 	: unary_function<Ty, size_t>
> {
> size_t operator()(Ty val) const;
> };
> } }
>
> This template is specialized for bool, all integer types, all
> floating-point types, pointers, std::string, and std::wstring.
>
> You can specialize it for any of your own types.
>
> You can also use your own type as an argument to hash_map or hash_set.
>
> For details, see section 5.6 of my book, "The C++ Standard Library
> Extensions: a Tutorial and Reference."

If I'm not mistaken, section 5.6 of your book deals with the hash
function object for std::tr1::unordered_map and
std::tr1::unordered_set.  hash_map and hash_set, which the OP refers
to, are implementation-specific extensions (from SGI, perhaps?), not
the new std::tr1 containers.

Best regards,

Tom

Thomas Tutone wrote:
> 
> If I'm not mistaken, section 5.6 of your book deals with the hash
> function object for std::tr1::unordered_map and
> std::tr1::unordered_set.  hash_map and hash_set, which the OP refers
> to, are implementation-specific extensions (from SGI, perhaps?), not
> the new std::tr1 containers.
> 

You're right. Sorry for any confusion this caused.

Pete Becker wrote:
> Thomas Tutone wrote:
> 
>>
>> If I'm not mistaken, section 5.6 of your book deals with the hash
>> function object for std::tr1::unordered_map and
>> std::tr1::unordered_set.  hash_map and hash_set, which the OP refers
>> to, are implementation-specific extensions (from SGI, perhaps?), not
>> the new std::tr1 containers.
>>
> 
> You're right. Sorry for any confusion this caused.

I should add, though, that the specifications in TR1 for unordered_set 
and unordered_map were designed to be compatible with existing 
implementations of hash_set and hash_map. The Dinkumware implementations 
of hash_set, hash_map, unordered_set, and unordered_map all share the 
same underlying code.