Selecting the begin() and end() for C++

Question

I have a project that required me to read and sort values read from a .csv files by a certain column. I'm using the algorithm library in C++ i.e #include <algorithm> The library has the function sort() which I plan on utilizing. It required me to specify the beginning and the end e.g for a 2d vector called vect it would require vect.begin() as well as vect.end(). My question is, is it possible to specify the exact start position and end position for the sort() function. That is like start at vect[1][0] and end at vect[9][9].

Thanks for your help

Answer 1

I'm assuming for sake of discussion you are working with a std::vector<std::vector<int> >. The same discussion applies for a 2D vector of other types.

If you want to sort the individual ints so they are ordered within the std::vector<std::vector<int> > then it's not possible to do directly. There is no iterator that can be obtained directly from a std::vector<std::vector<int>> which runs over all of the nested ints.

One way might be to set up a temporary copy into a std::vector<int> (i.e. create a flattened 1D vector), sort that, and then copy the elements back. For example;

 std::vector<std:vector<int> > vec;

   // populate vec somehow

 std::vector<int> elements(0);

 // create a single std::vector<int> from the vector<vector<int>> by
 //    appending the vector<int>s end to end

 for (const auto &row : vec)
 {
      elements.insert(elements.end(), row.begin(), row.end());
 }

 std::sort(elements.begin(), elements.end());   // sort in ascending order

 //   now copy the sorted elements back

 auto start = elements.begin();

 for (auto &row : vec)   //  non-const here since we seek to change the vector<int>s within vec
 {
      auto end = start + row.size();
      std::copy(start, end, row.begin());
      start = end;
 }

The shenanigans in the last loop with row.size() and row.begin() deals with the possibility that the vector<int> within vec are of different sizes, so will change

 {{5,6,7}, {1,2}, {3,4,8}}

to be

 {{1,2,3}, {4, 5}, {6,7,8}}

rather than to something else such as to

 {{1,2}, {3,4,5}, {6, 7, 8}};     //   vector<int>s resized

Things can be simplified a little if you assume all the inner vectors are of the same size.

Alternatively, you might try hand-rolling a struct/class type that has all the properties of a random access iterator (which is what std::sort() requires). That structure (or its member functions/operators) will need to both track which std::vector<int> (within the 2D vector) AND the particular int within that vector that it refers to. That will be moderately tricky (e.g. if the custom iterator refers to the last element of a particular std::vector<int>, incrementing it must give a result that references the first element of the next vector<int>). A std::vector<std::vector<int> > simply does not have any built in capability to give you such an iterator directly. I'll leave rolling such a custom iterator as an exercise.

Answer 2

Given that you know the order of iteration you can specify elements relative to the start.

vect.begin()     // first element
vect.begin() + 2 // third element

so if you want to sort only the first let's say 10 elements use the following:

std::sort(vect.begin(), vect.begin() + 10);

More here.

As others already mentioned you can not really sort a 2D vector. So either you sort each vector individually or flatten it to a 1D vector and use index calculations for 2D interpretation.

Answer 3

1. Load your column you sort by into std::vector< std::pair > index; and set
   • the first element as row index in vector (0,1,2,...)
   • the second element will be value in the row and column
2. Sort it by second item of the pair using this question: How do I sort a vector of pairs based on the second element of the pair?
3. Now you have row index sorted by the column values in the vector < pair > and you can create new 2d vector and populate its rows using rows from the original 2d vector. In each iteration select row with index in original 2D vector using your first item in index vector.