In this next part of the big STL algorithm tutorial, we are going to talk about set operations on sorted ranges:
includes
set_difference
set_intersection
set_symmetric_difference
set_union
Before we start, it is worth mentioning that set operations do not mean that these operations are applied on containers of type std::set
.
The set prefix simply means that these are operations on subsets of collections.
Let's have a look.
includes
Yes, this one doesn't have the set prefix. Never mind.
std::includes
in its simplest form takes 4 parameters, 4 iterators. The first two defining one range, and the second two another range.
This algorithm returns a boolean and returns true
in particular if the second range is a subsequence of the first one.
Let's see a simple example.
#include <algorithm>
#include <iostream>
#include <vector>
int main() {
std::vector nums {1, 2, 3, 4, 5, 6, 7, 8, 9, 10};
std::vector subsequece {3, 4, 5};
std::vector subset {5, 4, 3};
std::vector otherNums {42, 51, 66};
std::cout << std::boolalpha;
std::cout << "std::includes(nums.begin(), nums.end(), subsequece.begin(), subsequece.end()): " << std::includes(nums.begin(), nums.end(), subsequece.begin(), subsequece.end()) << '\n';
std::cout << "std::includes(nums.begin(), nums.end(), subset.begin(), subset.end()): " << std::includes(nums.begin(), nums.end(), subset.begin(), subset.end()) << '\n';
std::cout << "std::includes(nums.begin(), nums.end(), otherNums.begin(), otherNums.end()): " << std::includes(nums.begin(), nums.end(), otherNums.begin(), otherNums.end()) << '\n';
}
/*
std::includes(nums.begin(), nums.end(), subsequece.begin(), subsequece.end()): true
std::includes(nums.begin(), nums.end(), subset.begin(), subset.end()): false
std::includes(nums.begin(), nums.end(), otherNums.begin(), otherNums.end()): false
*/
We can observe that in order to get a positive result from the algorithm, the second range must be a subsequence of the first one. Having the elements to be a subset of the first container is not enough.
What would happen if the first container would not be sorted?
#include <algorithm>
#include <iostream>
#include <vector>
int main() {
std::vector nums {1, 2, 5, 4, 3, 6, 7, 8, 9, 10};
std::vector subseq {5, 4, 3};
std::cout << std::boolalpha;
std::cout << "std::includes(nums.begin(), nums.end(), subseq.begin(), subseq.end()): " << std::includes(nums.begin(), nums.end(), subseq.begin(), subseq.end()) << '\n';
}
/*
std::includes(nums.begin(), nums.end(), subseq.begin(), subseq.end()): true
*/
We can see that our first range is not ordered, but std::includes
was able to find a subsequence in it. Yet, you should not rely on this. If you don't pass sorted ranges to std::includes
, the behaviour is undefined.
std::includes
can take two extra parameters, I'd say the usual ones.
Before all others, it can take an execution policy and at the last position, it can a custom comparator in the form of a function pointer, function object or lambda expression to compare items of the two passed in containers.
set_difference
This algorithm takes 2 ranges and will copy all the elements from the first range that is not in the second range to a destination range.
Just like each algorithm in this article, set_difference
is only guaranteed to work with sorted ranges.
As we could already get used to it, the two input ranges are taken by a pair of iterators and the output range is only denoted by its beginning point. As usual, it's the caller's responsibility to make sure that the destination range can accommodate enough items. You can also pass an inserter iterator.
std::set_difference
can also take the usual two extra parameters, like an execution policy before all the others or a comparator after all the parameters.
Let's have here an example:
#include <algorithm>
#include <iostream>
#include <vector>
int main() {
std::vector nums {1, 2, 3, 4, 5, 5};
std::vector otherNums {1, 2, 3, 6, 7};
std::vector<int> difference;
std::set_difference(nums.begin(), nums.end(),
otherNums.begin(), otherNums.end(),
std::back_inserter(difference));
for (auto n : difference) {
std::cout << n << " ";
}
std::cout << '\n';
}
/*
4 5 5
*/
It's worth noticing that if the same value appears multiple times in the first container but never in the second, then it will be copied multiple times into the output range.
In the above example, we had 5
twice in nums
and not at all in otherNums
, so it appears twice in difference
. But if 5
appears once in otherNums
too, it will still appear in the difference
, but then only once. After all, that's the difference. If it appears twice in the first input and only once in the second, that is the difference.
set_intersection
set_intersection
takes the same parameters as set_difference
.
Two pairs of iterators as input, an output iterator an optional execution policy and a comparator.
It will copy each element to the destination range that is both in the input and the output range.
If a value appears multiple times in both ranges, it will be copied multiple times. To be more exact, if it appears in the first range m
times and n
times in the second, it will be copied std::min(m,n)
times.
std::set_intersection
also keeps the items in their relative order, the order of the items in the input and in the output range is the same.
Here are some examples:
#include <algorithm>
#include <iostream>
#include <vector>
int main() {
std::vector nums {1, 2, 3, 4, 5};
std::vector sameNums {1, 2, 3, 4, 5};
std::vector otherNums {1, 2, 7};
std::vector<int> intersectionOfSame;
std::vector<int> otherIntersection;
std::set_intersection(nums.begin(), nums.end(),
sameNums.begin(), sameNums.end(),
std::back_inserter(intersectionOfSame));
for (auto n : intersectionOfSame) {
std::cout << n << " ";
}
std::cout << '\n';
std::set_intersection(nums.begin(), nums.end(),
otherNums.begin(), otherNums.end(),
std::back_inserter(otherIntersection));
for (auto n : otherIntersection) {
std::cout << n << " ";
}
std::cout << '\n';
}
/*
1 2 3 4 5
1 2
*/
set_symmetric_difference
Regarding the possible parameters, we don't have a difficult job today. set_symmetric_difference
still operates on the very same list of parameters as our previous two algorithms.
Two pairs of iterators as input, an output iterator an optional execution policy and a comparator.
What does computing a symmetric difference mean?
It means that in the output range you'll find all the elements that are found in either of the two input ranges, but not in both.
In a way, you can consider it that it's the combination of two std::set_difference
, with the input ranges swapped between the two calls.
#include <algorithm>
#include <iostream>
#include <vector>
int main() {
std::vector nums {1, 2, 5, 6, 8};
std::vector otherNums {3, 4, 7};
std::vector<int> difference;
std::vector<int> symmetricDifference;
std::set_symmetric_difference(nums.begin(), nums.end(),
otherNums.begin(), otherNums.end(),
std::back_inserter(symmetricDifference));
for (auto n : symmetricDifference) {
std::cout << n << " ";
}
std::cout << '\n';
std::set_difference(nums.begin(), nums.end(),
otherNums.begin(), otherNums.end(),
std::back_inserter(difference));
std::set_difference(otherNums.begin(), otherNums.end(),
nums.begin(), nums.end(),
std::back_inserter(difference));
for (auto n : difference) {
std::cout << n << " ";
}
std::cout << '\n';
}
/*
1 2 3 4 5 6 7 8
1 2 5 6 8 3 4 7
*/
The difference between calling set_symmetric_difference
and calling set_difference
- as you can see above - is that set_symmetric_difference
will output a sorted range while calling set_difference
twice will leave us with a container that has two sorted parts (the result of each call), but not sorted overall.
And anyway, the implementation of set_symmetric_difference
is optimal for its purpose, unlike calling set_difference
twice.
set_union
If you followed through the previous sections, you won't encounter many surprises while learning about set_union
. This algorithm takes two ranges and will build another out of the elements that are present in either one or the other container.
If an element can be found in both, then first all the elements will be taken from the first range and then if there were more elements with the same value in the second one, the excess will be copied from there.
Regarding the parameters, set_union
behaves like the previous ones. It takes two pairs of iterators as input, an output iterator an optional execution policy and a comparator.
Let's see an example:
#include <algorithm>
#include <iostream>
#include <vector>
int main() {
std::vector nums {1, 1, 2, 2, 5, 6, 8};
std::vector otherNums {2, 5, 5, 7};
std::vector<int> unionOfNums;
std::set_union(nums.begin(), nums.end(),
otherNums.begin(), otherNums.end(),
std::back_inserter(unionOfNums));
for (auto n : unionOfNums) {
std::cout << n << " ";
}
std::cout << '\n';
}
/*
1 1 2 2 5 5 6 7 8
*/
We can observe that those items that only appear in one of the inputs appear exactly the same times in the output. We have two values that appear in both inputs.
2
, appears twice in the first input and once in the second. So it's taken twice from the first, and there is no excess in the second, so we are done.
5
appears once in the first, so it's taken once from there and then there is one more item in the second input (2-1==1), so one more is taken there.
You might ask, why don't we say that it's just taken twice from the second range. Because that's what the specs say and there is a good reason behind it. The fact that two values are considered equal after comparison doesn't mean that they are identical. We'll have a look at this the next time based on Walter Brown's talk about the Italian C++ Conference 2021.
Conclusion
This time, we learned about set operations on sorted ranges, which work on any containers not only on sets. The term set is used in its mathematical sense, it's not referring to the type of containers. Apart from that, they are quite logical, they don't have lot's of surprises, but we have to keep in mind especially for unions and intersections that items that are equal are not necessarily identical and it does matter which equal element we take.
Next time we'll discover heap operations. Stay tuned.
Connect deeper
If you liked this article, please
- hit on the like button,
- subscribe to my newsletter
- and let's connect on Twitter!
Top comments (6)
Have you planned to write an article about execution policy?
Look at what I stumbled upon!
modernescpp.com/index.php/performa...
Perfect! Thanks :)
Maybe once I finished with the algorithms. I'll need to see how to really test these policies. Have you tried them in a professional context?
Absolutely not. I am work with "tiny" MCUs so this is not something I need. In fact, I don't handle a lot collections, and these collections are always quite small. As I know nothing about them, it would be nice it someone would come with a nice article on this topic :D
We don't use C++17 yet at work, we'll see how far I can get at a home environment - once I'm getting to this topic.