Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Iterating over unique elements of `std::multiset`

All I need is to know if something exists and how many times it exist. I will iterate over the existent things and query how much of that exists.

My implementation so far uses multiset, I do as follow:

std::multiset<thing> a;
auto previous = a.end();
for( auto each = a.begin(); each != a.end(); ++each ) {
    if( previous == a.end() || *previous != *each ) {
        a.count(*each);
    }
    previous = each;
}

Clarifications

I have a vector of things. But they repeat the value sometimes, I want to iterate over unique things and for each unique do something. This "something" needs to know the amount of time this thing appears on the vector.

The code I posted above is how I am resolving my problem right now, it does not seems to be the most elegant way to do what I want.

I am just following the Stackoverflow guidelines: I tell what is my problem, and I tell my (tried) solution.

If a sentence with a question mark is really needed, there you go: Is there a way to iterate over unique elements over a multiset?

like image 894
André Puel Avatar asked Feb 07 '13 13:02

André Puel


1 Answers

Three possible approaches:

  • Use std::unique to create a temporary collection of unique values. This might make the code a little more readable, but less efficient.
  • Advance your iterator by using std::multiset::upper_bound rather than increments: for( auto each = a.begin(); each != a.end(); each=a.upper_bound(*each)) - that way you don't need the if check insider your loop, plus it is guaranteed to be logarithmic in size. Pretty cool (didn't know that before I looked it up). For the following suggestion, all credit goes to @MarkRansom: Using std::upper_bound from <algorithm>, you can specify a range in which to look for the upper bound. In your case, you already have a good candidate for the start of that range, so this method is likely to be more efficient, depending on the implementation in your standard library.
  • If this is a real performance problem for you and the previous solution still isn't good enough, consider switching to map<thing, unsigned> or even unordered_map<thing,unsigned> where the unsigned just keeps track of the number of equivalent things you have. That implies rewriting your insertion/deletion code though.
like image 80
us2012 Avatar answered Nov 15 '22 02:11

us2012