By StackHeapCollision


2012-12-10 10:58:50 8 Comments

Suppose I wanted to map data with a string as the key. What container should I have chosen, map or unordered_map? unordered_map takes up more memory so let's suppose memory isn't an issue, and the concern is speed.

unordered_map should generally give average complexity of O(1) with the worst case of O(n). In what cases would it get to O(n)? When does a map get more time efficient than unordered_map? Does it happen when n is small?

Assuming I would use STL unordered_map with the default haser Vs. map. string is the key.

If I'm going to iterate over the elements rather than access an individual element each time, should I prefer map?

4 comments

@ypnos 2012-12-10 11:07:14

In practice, if memory is no issue, unordered_map is always faster if you want single element access.

The worst case is theoretical and bound to a single hash accounting for all of the elements. This is not of practical relevance. The unordered_map gets slower as soon as you have at least log N elements belonging to the same hash. This is also not of practical relevance. In some special scenarios you could use a specific hashing algorithm that ensures a more uniform distribution. For ordinary strings that don't share a specific pattern, the generic hash functions coming with unordered_map are just as good.

If you want to traverse the map (using iterators) in a sorted fashion, you cannot use unordered_map. On the contrary, map not only allows that, but also can provide you with the next element in a map based on an approximation of the key (see lower_bound and upper_bound methods).

@Don Hatch 2016-10-07 00:22:36

This answer is misleading at best. It's not true that "unordered_map is always faster for single element access" -- the only thing I can think of that's always true is that it's always faster amortized and asymptotically. The "amortized" is an important caveat in practce: assuming it's implemented as a hash table of some sort, if I remember my hash tables correctly, as you grow it by inserting elements, it will "hiccup" with an Ω(n) operation every so often. That may or may not be something any particular app can tolerate.

@user1773602 2012-12-10 11:16:33

                       | map              | unordered_map
---------------------------------------------------------
element ordering       | strict weak      | n/a 
                       |                  |
common implementation  | balanced tree    | hash table
                       | or red-black tree|  
                       |                  |
search time            | log(n)           | O(1) if there are no hash collisions
                       |                  | Up to O(n) if there are hash collisions 
                       |                  | O(n) when hash is the same for any key
                       |                  |     
Insertion time         | log(n)+rebalance | Same as search
                       |                  | 
Deletion time          | log(n)+rebalance | Same as search
                       |                  | 
needs comparators      | only operator <  | only operator ==
                       |                  |
needs hash function    | no               | yes
                       |                  |
common use case        | when good hash is| In most other cases. 
                       | not possible or  | 
                       | too slow. Or when|
                       | order is required| 

@HelloGoodbye 2015-10-18 14:44:18

Comment about common implementation: A red–black tree is a kind of balanced tree (or more specifically, a kind of self-balancing binary search tree).

@mtk 2016-11-04 20:53:38

rebalance would take no more than log(n)

@Shashwat 2018-07-05 19:46:49

What about iterating over all the elements?

@Pubby 2012-12-10 11:09:10

What container should I have chosen, map or unordered_map? unordered_map takes up more memory so let's suppose memory isn't an issue, and the concern is speed.

Profile and then decide. unordered_map is generally faster, but it varies per case.

In what cases would it get to O(n)?

When the hashing isn't good and a bunch of elements are being assigned to the same bins.

When does a map get more time efficient than unordered_map? Does it happaen when n is small?

Probably not, but profile it if you really care. Having a container with a small size be the bottleneck of your program seems extremely unlikely. Anyway, a simple vector with linear search may be faster for such cases.


The most important thing when deciding is the requirements of ordering and lack of iterator invalidation. If you need either, you pretty much have to use map. Otherwise, unordered_map.

@zaufi 2012-12-10 11:07:04

In what cases would it get to O(n)?

if you have such a bad hash function which produces the same hash value for all input stirngs (i.e. produce collisions)...

What container should I have chosen, map or unordered_map?

It is always the questions of requirements and kind/amount of data do you have.

When does a map get more time efficient than unordered_map?

It is just different structures. You'd better to make a chiose to use one of them depending on your typical use cases (takeing in account what kind of data do you have and its amount)

Does it hppaen when n is small?

In case of small data amount everything depends on particular STL implementation... So sometimes even a plain vector/array could be faster than associative containers...

Related Questions

Sponsored Content

39 Answered Questions

11 Answered Questions

[SOLVED] How to remove a key from a Python dictionary?

56 Answered Questions

[SOLVED] Sort a Map<Key, Value> by values

34 Answered Questions

[SOLVED] How do I sort a dictionary by value?

42 Answered Questions

[SOLVED] How do I efficiently iterate over each entry in a Java Map?

42 Answered Questions

[SOLVED] How do I merge two dictionaries in a single expression?

18 Answered Questions

12 Answered Questions

[SOLVED] Is there any advantage of using map over unordered_map in case of trivial keys?

21 Answered Questions

[SOLVED] Why should I use a pointer rather than the object itself?

  • 2014-03-03 11:54:16
  • gEdringer
  • 307158 View
  • 1567 Score
  • 21 Answer
  • Tags:   c++ pointers c++11

20 Answered Questions

Sponsored Content