Comments on: The New Old Iterator Concepts http://conceptgcc.wordpress.com/2007/05/03/the-new-old-iterator-concepts/ The Design and Implementation of Concepts for C++ Thu, 06 Aug 2009 18:19:18 +0000 http://wordpress.com/ hourly 1 By: Kostas Skolarikis http://conceptgcc.wordpress.com/2007/05/03/the-new-old-iterator-concepts/#comment-699 Kostas Skolarikis Wed, 16 May 2007 10:07:29 +0000 http://conceptgcc.wordpress.com/2007/05/03/the-new-old-iterator-concepts/#comment-699 Something went wrong with the conversion of my previous message and template brackets and parameters disappeared. So I repeat it hopefully correct this time I was able to solve with this version of ConceptGCC the (infamous also) problem of sorting two “parallel” vectors, in an eventually conforming way! If you have two vectors (<code>vector<key_type> , vector<data_type></code>) and you like to treat them as a unified structure(as an alternative to <code>map<key_type, data_type></code> for example) the most natural way is to make up an iterator with value type something like <code>pair<key_type, data_type></code> and reference_type <code>pair<key_type& , data_type& ></code> (that is not value_type&). Unfortunately this solution was not conforming. I was able to compile such an iterator with this version of ConceptGCC and it worked just fine. It had been working also with gcc 4.1.1 (and older i think) but it was not a portable solution. It was my fist time testing ConceptGCC and I thing the compilation speed is going to be a real problem. Something went wrong with the conversion of my previous message and template brackets and parameters disappeared. So I repeat it hopefully correct this time

I was able to solve with this version of ConceptGCC the (infamous also) problem of sorting two “parallel” vectors, in an eventually conforming way!
If you have two vectors (vector<key_type> , vector<data_type>) and you like to treat them as a unified structure(as an alternative to map<key_type, data_type> for example) the most natural way is to make up an iterator with value type something like pair<key_type, data_type> and reference_type pair<key_type& , data_type& > (that is not value_type&). Unfortunately this solution was not conforming. I was able to compile such an iterator with this version of ConceptGCC and it worked just fine.

It had been working also with gcc 4.1.1 (and older i think) but it was not a portable solution.

It was my fist time testing ConceptGCC and I thing the compilation speed is going to be a real problem.

]]> By: Kostas Skolarikis http://conceptgcc.wordpress.com/2007/05/03/the-new-old-iterator-concepts/#comment-684 Kostas Skolarikis Tue, 15 May 2007 13:09:07 +0000 http://conceptgcc.wordpress.com/2007/05/03/the-new-old-iterator-concepts/#comment-684 I was able to solve with this version of ConceptGCC the (infamous also) problem of sorting two “parallel” vectors, in an eventually conforming way! If you have two vectors (vector, vector) and you like to treat them as a unified structure(as an alternative to map for example) the most natural way is to make up an iterator with value type something like pair and reference_type pair (that is not value_type&). Unfortunately this solution was not conforming. I was able to compile such an iterator with this version of ConceptGCC and it worked just fine. It had been working also with gcc 4.1.1 (and older i think) but it was not a portable solution. It was my fist time testing ConceptGCC and I thing the compilation speed is going to be a real problem. I was able to solve with this version of ConceptGCC the (infamous also) problem of sorting two “parallel” vectors, in an eventually conforming way!
If you have two vectors (vector, vector) and you like to treat them as a unified structure(as an alternative to map for example) the most natural way is to make up an iterator with value type something like pair and reference_type pair (that is not value_type&). Unfortunately this solution was not conforming. I was able to compile such an iterator with this version of ConceptGCC and it worked just fine.

It had been working also with gcc 4.1.1 (and older i think) but it was not a portable solution.

It was my fist time testing ConceptGCC and I thing the compilation speed is going to be a real problem.

]]>