hash_map

[Jon Cosby]

I need help in hashmaps. Why doesn't this work:

#include <hash_map>

hash_map <int, string> hm1;
typedef pair <int, string> pr;
hm1.insert(str_pair(1, "Hello"));

It compiles, but crashes at runtime, pointing to something in xhash.

 

[P.J. Plauger]

I need help in hashmaps. Why doesn't this work:

#include <hash_map>

hash_map <int, string> hm1;
typedef pair <int, string> pr;
hm1.insert(str_pair(1, "Hello"));

It compiles, but crashes at runtime, pointing to something in xhash.

Depends a lot on what your undocumented str_pair does. It would
be a really good idea for it to return something like:

hash_map<int, string>::value_type(1, string("Hello"));

P.J. Plauger
Dinkumware, Ltd.
http://www.dinkumware.com

 

[Gianni Mariani]

I need help in hashmaps. Why doesn't this work:

#include <hash_map>

hash_map <int, string> hm1;
typedef pair <int, string> pr;
hm1.insert(str_pair(1, "Hello"));

.... str_pair is not defined here.

It compiles, but crashes at runtime, pointing to something in xhash.

hash_map is not part of the standard, however it is popular, I have no
idea what the "right(TM)" way to get access to it under gcc so I'd check
up on that. I used gcc 3.3.1.

The code below works find for me. Maybe your str_pair() returns a
reference local variable ?

#include <ext/hash_map>
#include <string>
#include <iostream>

using namespace std;
using namespace __gnu_cxx;

int main()
{
hash_map<int, string> hm1;

typedef hash_map<int, string>::value_type pr;

hm1.insert(pr(1, "Hello"));

cout << hm1[1] << endl;

}

 

[Jon Cosby]

Depends a lot on what your undocumented str_pair does.

That was a typo. Should be just "pr".


Jon

 

[Mike Wahler]

That was a typo. Should be just "pr".

Then P.J.'s reply is still essentially the same, but changing
'str_pair' to 'pr'.

BTW 'hash_map' is not part of standard C++ (although provided
as an extension by many implementations).

-Mike

 

[Mike Wahler]

Then P.J.'s reply is still essentially the same, but changing
'str_pair' to 'pr'.

Ack! Disregard that, I see you have defined the object 'pr'
above. I didn't read carefully enough. Sorry!

BTW 'hash_map' is not part of standard C++ (although provided
as an extension by many implementations).

This does remain true.

-Mike

 

[Jon Cosby]

BTW 'hash_map' is not part of standard C++ (although provided
as an extension by many implementations).

-Mike

Are you sure? VC lists it in the Standard C++ Library.

Jon

 

[Pete Becker]

Are you sure? VC lists it in the Standard C++ Library.

He's sure. And he's right.

 

[Mike Wahler]

Are you sure?
Yes.

VC lists it in the Standard C++ Library.

VC does not define the C++ language.

Also, I find no reference at all to 'hash_map' in
my VC++6.0 documentation. Perhaps version 7 has
such an animal, but it's still not standard.

-Mike

 

[David Fisher]

Are you sure? VC lists it in the Standard C++ Library.

The STL Tutorial and Reference Guide (2nd ed) says on page 161:

"Ideally, both sorted and hashed associated containers should be in the C++
Standard Library, but only sorted associated containers are included
(unofficial STL-based hash table specifications and implementations are
available)"

BTW a reason given for this (on p.161) is that hash tables can be order (N)
in the worst case, but that implies the use of linked lists at each hash
node ... surely a binary tree, etc. could be used instead. Anyone know if
std::maps and sets have been permitted to be implemented as hash tables
since this book was written (2001) ?

David F

 

[David Fisher]

"Ideally, both sorted and hashed associated containers should be in the
C++

Standard Library ..."

Oops, I meant "associative containers" ...

David F