Re: alternative tuple implementation (was: [boost] Re: Any interest for a parser class?)

From: Roland (roland_at_[hidden])
Date: 2002-01-18 08:41:57

> > > I never understood what the benefits of doing it differently might be, so I
> > > had nothing to say. If it is just an implementation detail, and it avoids
[...]
> Agreed, that's crucially important! Why will the code be faster?

 Performance was just one reason for me trying an alternative tuple implementation;
 the main reason, however, was my interest in tuples of containers, and
 tuples of iterators.

 What I'm thinking of is to write something like the following:

 vector<int> a; vector<double> b; vector<string> c; 

 tuple< vector<int>, vector<double>, vector<string> > t( a,b,c );
   // Is it the same as vector< tuple<int, double, string> >?
   // Close to? Or completely different?

 t[0] = tuple<int, double, string>( 42, 3.14, "foo" );

 tuple< vector<int>, vector<double>, vector<string> >::iterator it;
   // Is the same as tuple< vector<int>::iterator, 
   //                       vector<double>::iterator, vector<string>::iterator >

 for( it = t.begin(); it != t.end(); ++it ) // requires begin(), end(), ++

  
 etc.

 Of course there are still a couple of unsolved issues, for instance,
 how to deal with different container types / sizes etc.
 And perhaps it's a bad idea to stuff all this functionality 
 into one tuple class. 

 Another way would be to use zip_view of the View Template Library,
 but I'd like to tie together more than two containers.

 - Roland