CS100 Lecture 23

More on STL: Sequence containers and associative containers

Overview of STL

Standard Template Library

Added into C++ in 1994.

Containers
Iterators (In Lecture 20)
Algorithms (In Lecture 20)
Function objects
Some other adaptors, like container adaptors and iterator adaptors
Allocators

The next generation: C++20 Ranges

Containers

Sequence containers
vector, list, deque, array (since C++11), forward_list (since C++11)
Associative containers
set, map, multiset, multimap (often implemented with binary search trees)
Unordered associative containers (since C++11)
unordered_set, unordered_map, unordered_multiset, unordered_multimap (implemented with hash tables)
Container adaptors: provide a different interface for sequential containers, but they are not containers themselves.
stack, queue, priority_queue
(since C++23) flat_set, flat_map, flat_multiset, flat_multimap

Iterators

Without iterators:

Traverse an array cpp for (int i = 0; i != sizeof(a) / sizeof(a[0]); ++i) do_something(a[i]);
Traverse a vector cpp for (std::size_t i = 0; i != v.size(); ++i) do_something(v[i]);
Traverse a linked-list? cpp for (ListNode *p = l.head(); p; p = p->next) do_something(p->data);

Iterators

A generalization of pointers, used to access elements in different containers in a uniform manner.

With iterators:

The following works no matter whether c is an array, a std::string, or any container.

for (auto it = std::begin(c); it != std::end(c); ++it)
  do_something(*it);

Equivalent way: range-based for loops

for (auto &x : c) do_something(x);

Algorithms

The algorithms library defines functions for a variety of purposes: - searching, sorting, counting, manipulating, ...

Examples:

// assign every element in `a` with the value `x`.
std::fill(a.begin(), a.end(), x);
// sort the elements in `b` in ascending order.
std::sort(b.begin(), b.end());
// find the first element in `b` that is equal to `x`.
auto pos = std::find(b.begin(), b.end(), x);
// reverse the elements in `c`.
std::reverse(c.begin(), c.end());

Algorithms

Example: Obtain the rank of each number in a sequence.

auto getRank(const std::vector<int> &data) {
  auto tmp = data;
  std::sort(tmp.begin(), tmp.end()); // sort
  auto pos = std::unique(tmp.begin(), tmp.end()); // drop duplicates
  auto ret = data;
  for (auto &x : ret)
    x = std::lower_bound(tmp.begin(), pos, x) - tmp.begin(); // binary search
  return ret;
}

Function objects

Things that look like "functions": Callable - functions, and also function pointers - objects of a class type that has an overloaded operator() (the function-call operator) - lambda expressions

The standard library has defined some common function objects: std::less, std::greater, ...

std::sort(a.begin(), a.end(), std::greater<>{}); // Sort in descending order.

Adaptors

Container adaptors: std::stack, std::queue, std::priority_queue - Represent the stack, queue and the priority-queue data structures respectively. - They are not containers themselves. They are based on some underlying container, and provide the interfaces of the corresponding data structures.

std::stack<int> stk; // By default, uses `std::deque<int>` as
                     // the underlying container.
std::stack<int, std::vector<int>> stk2; // Uses `std::vector<int>`.

Iterator adaptors: To be discussed in recitations.

Sequence containers

Note: std::string is not treated as a container but behaves much like one.

Sequence containers

std::vector<T>: dynamic contiguous array (we are quite familiar with)

std::deque<T>: double-ended queue (often pronounced as "deck")
std::deque<T> supports fast insertion and deletion at both its beginning and its end. (push_front, pop_front, push_back, pop_back)

std::array<T, N>: same as T[N],but it is a container
It will never decay to T *.
Container interfaces are provided: .at(i), .front(), .back(), .size(), ..., as well as iterators.

Sequence containers

std::list<T>: doubly-linked list
std::list<T> supports fast insertion and deletion anywhere in the container,
but fast random access is not supported (i.e. no operator[]).
Bidirectional traversal is supported.

std::forward_list<T>: singly-linked list
Intended to save time and space (compared to std::list).
Only forward traversal is supported.

Interfaces

STL containers have consistent interfaces. See here for a full list.

Element access:

c.at(i), c[i]: access the element indexed i. at performs bounds checking, and throws std::out_of_range if i exceeds the valid range.
c.front(), c.back(): access the front/back element.

Interfaces

Size and capacity: c.size() and c.empty() are what we already know.

c.resize(n), c.resize(n, x): adjust the container to be with exactly n elements. If n > c.size(), n - c.size() elements will be appended.
c.resize(n): Appended elements are value-initialized.
c.resize(n, x): Appended elements are copies of x.
c.capacity(), c.reserve(n), c.shrink_to_fit(): only for string and vector.
c.capacity() returns the capacity (number of elements that can be stored in the current storage)
c.reserve(n): reserves space for at least n elements.
c.shrink_to_fit(): requests to remove the unused capacity, so that c.capacity() == c.size().

Interfaces

Modifiers:

c.push_back(x), c.emplace_back(args...), c.pop_back(): insert/delete elements at the end of the container.
c.push_front(x), c.emplace_front(args...), c.pop_front(): insert/delete elements at the beginning of the container.
c.clear() removes all the elements in c.

Interfaces

Modifiers:

c.insert(...), c.emplace(...), c.erase(...): insert/delete elements at a specified location.
Warning: For containers that need to maintain contiguous storage (string, vector, deque), insertion and deletion somewhere in the middle can be very slow (\(O(n)\)).
These functions have a lot of overloads. Remember a few common ones, and STFW (Search The Friendly Web) when you need to use them.

Interfaces

Some of these member functions are not supported on some containers, depending on the underlying data structure. For example: - Any operation that modifies the length of the container is not allowed for array. - push_front, emplace_front and pop_front are not supported on string, vector and array. - size is not supported on forward_list in order to save time and space. - operator[] and at are not supported on linked-lists.

This table tells you everything.

Iterators

Every container has its iterator: Container::iterator. e.g. std::vector<int>::iterator, std::forward_list<std::string>::iterator

auto comes to our rescue!

c.begin() returns the iterator to the first element of c.

c.end() returns the iterator to the element following the last element of c.

Iterator categories

ForwardIterators: supports *it, it->mem, ++it, it++, it1 == it2, it1 != it2

BidirectionalIterator: a ForwardIterator that can be moved in both directions - supports --it and it--.

RandomAccessIterator: a BidirectionalIterator that can be moved to point to any element in constant time. - supports it + n, n + it, it - n, it += n, it -= n for an integer n. - supports it[n], equivalent to *(it + n). - supports it1 - it2, returns the distance of two iterators. - supports <, <=, >, >=.

Iterator categories

ForwardIterators: an iterator that can be moved forward. - forward_list<T>::iterator

BidirectionalIterator: a ForwardIterator that can be moved in both directions - list<T>::iterator

RandomAccessIterator: a BidirectionalIterator that can be moved to point to any element in constant time. - string::iterator, vector<T>::iterator, deque<T>::iterator, array<T,N>::iterator

Iterator categories

To know the category of an iterator of a container, consult its type alias member iterator_category.

using vec_iter = std::vector<int>::iterator;
using category = vec_iter::iterator_category;

Put your mouse on category, and the IDE will tell you what it is.

It is one of the following tags: std::forward_iterator_tag, std::bidirectional_iterator_tag, std::random_access_iterator_tag.

Note: Two other categories InputIterator and OutputIterator will be discussed in recitations.

Constructors of containers

All sequence containers can be constructed in the following ways:

Container c(b, e), where [b, e) is an iterator range.
Copies elements from the iterator range [b, e).
Container c(n, x), where n is a nonnegative integer and x is a value.
Initializes the container with n copies of x.
Container c(n), where n is a nonnegative integer.
Initializes the container with n elements. All elements are value-initialized.
This is not supported by string. (Why?)

Constructors of containers

All sequence containers can be constructed in the following ways:

Container c(b, e), where [b, e) is an iterator range.
Copies elements from the iterator range [b, e).
Container c(n, x), where n is a nonnegative integer and x is a value.
Initializes the container with n copies of x.
Container c(n), where n is a nonnegative integer.
Initializes the container with n elements. All elements are value-initialized.
This is not supported by string, because it is meaningless to have n value-initializes chars (all of them will be '\0')!

Associative containers

Motivation: set

Represent a "set": - Quick insertion, lookup and deletion of elements. - Order does not matter.

Sequence containers do not suffice: - Lookup of elements is \(O(n)\). - Quick insertion/deletion only happens at certain positions for some containers. - e.g. vector only supports quick insertion/deletion at the end. - The order of elements is preserved, which is not important.

You will learn the appropriate data structures in CS101.

`std::set`

Defined in <set>.

std::set<T> is a set whose elements are of type T. operator<(const T, const T) should be supported, because it is usually implemented as Red-black trees.
std::set<T, Cmp> is also available. x < y will be replaced with cmp(x, y), where cmp is a function object of type Cmp.

std::set<int> s1; // An empty set of ints
std::set<std::string> s2{"hello", "world"}; // A set of strings,
                                            // initialized with two elements
struct Student { std::string name; int id; };
std::set<Student> s3; // No operator< for Student is available.
                      // This line alone does not cause error, but you cannot
                      // insert elements into it.
s3.insert(Student{"Alice", 42}); // Error: No operator< available.

`std::set`

Defined in <set>.

std::set<T> is a set whose elements are of type T. operator<(const T, const T) should be supported, because set is usually implemented as Red-black trees.
std::set<T, Cmp> is also available. x < y will be replaced with cmp(x, y), where cmp is a function object of type Cmp.

struct Student { std::string name; int id; };
struct CmpStudentByName {
  bool operator()(const Student &a, const Student &b) const {
    return a.name < b.name;
  }
};
std::set<Student, CmpStudentByName> students; // OK
students.insert(Student{"Alice", 42}); // OK

`std::set`

Constructors

std::set<Type> s1{a, b, c, ...};
std::set<Type> s2(begin, end); // An iterator range [begin, end)

C++17 CTAD (Class Template Argument Deduction) also applies:

std::set s1{a, b, c, ...}; // Element type is deduced according to the list
std::set s2(begin, end); // Element type is deduced according to
                         // the type of elements pointed by `begin` and `end`.

Besides, std::set is copy-constructible, copy-assignable, move-constructible and move-assignable, just as the sequence containers we have learned.

std::set does not contain duplicate elements. These constructors will ignore duplicate elements.

`std::set`: operations

Common operations: s.empty(), s.size(), s.clear().

Insertion: insert and emplace. Duplicate elements will not be inserted. - s.insert(x), s.insert({a, b, ...}), s.insert(begin, end).

std::set s{3, 2, 5, 5, 1}; // {1, 2, 3, 5}. The duplicate 5 is removed.
std::cout << s.size() << std::endl; // 4
s.insert(42); // {1, 2, 3, 5, 42}
s.insert(42); // Nothing is inserted. (No errors.)
int a[]{10, 20, 30};
s.insert(a, a + 3); // An iterator range.
                    // s now contains {1, 2, 3, 5, 10, 20, 30, 42}.
s.insert({11, 12}); // {1, 2, 3, 5, 10, 11, 12, 20, 30, 42}.

`std::set`: insertion

Insertion: insert and emplace. Duplicate elements will not be inserted. - s.emplace(args...). Forwards the arguments args... to the constructor of the element type, and constructs the element in-place.

std::set<std::string> s;
s.emplace(10, 'c'); // inserts a string "cccccccccc"

s.insert(x) and s.emplace(args...) returns std::pair<iterator, bool>: - On success, .first is an iterator pointing to the inserted element, and .second is true. - On failure, .first is an iterator pointing to the element that prevented the insertion, and .second is false.

`std::set`: iterators

s.begin(), s.end(): Begin and off-the-end iterators.

The iterator of std::set is BidirectionalIterator: - Supports *it, it->mem, ++it, it++, --it, it--, it1 == it2, it1 != it2.

The elements are in ascending order: The following assertion always succeeds (if both tmp and iter are dereferenceable).

auto tmp = iter;
++iter;
assert(*tmp < *iter);

`std::set`: iterators

Elements in a set cannot be modified directly: *iter returns a reference-to-const. - The elements are stored in specific positions in the red-black tree, according to their values. - You cannot change their values arbitrarily.

`std::set`: traversal

Range-for still works!

std::set<int> s{5, 5, 7, 3, 20, 12, 42};
for (auto x : s)
  std::cout << x << ' ';
std::cout << std::endl;

Output: 3, 5, 7, 12, 20, 42. The elements are in ascending order.

Equivalent way: Use iterators

for (auto it = s.begin(); it != s.end(); ++it)
  std::cout << *it << ' ';
std::cout << std::endl;

`std::set`: deletion

Delete elements: erase - s.erase(x), s.erase(pos), s.erase(begin, end), where pos is an iterator pointing to some element in s, and [begin, end) is an iterator range in s. - s.erase(x) removes the element that is equivalent to x, if any. - returns 0 or 1, indicating the number of elements removed.

std::set<int> s{5, 5, 7, 3, 20, 12, 42};
std::cout << s.erase(42) << std::endl; // 42 is removed. output: 1
// s is now {3, 5, 7, 12, 20}.
s.erase(++++s.begin()); // 7 is removed.

`std::set`: element lookup

s.find(x), s.count(x), and some other functions.

s.find(x) returns an iterator pointing to the element equivalent to x (if found), or s.end() (if not found).

std::set<int> s = someValues();
if (s.find(x) != s.end()) // x is found
  // ...

`std::set`: pros and cons

The time complexity of insertion, deletion, and lookup of elements in a std::set: logarithmic in the size of the container. (\(O(\log n)\)) - Compared to sequence containers, this is (almost) a huge improvement.

Elements are sorted automatically.

Fast random access like v[i] is not supported.

Other kinds of sets:

Sets based on red-black trees: - std::set - std::multiset: allows duplicate elements

Sets based on hash-tables: (since C++11) - std::unordered_set: hash-table version of std::set - std::unordered_multiset: allows duplicate elements

Sets based on hash-tables provides (average-case) \(O(1)\) time operations, but requires the data to be hashable.

Motivation: map

Represent a map: \(f:S\to T\).

For sequence containers Container<Type>: \(S=\{0,1,2,\cdots,N-1\}\) (index), \(T\) is the set of values of type Type.
For std::set<Type>: \(T=\{\text{exist}, \text{not-exist}\}\), \(S\) is the set of values of type Type.

std::map<Key, Value>: defined in <map> - Key is the type of elements in \(S\), and Value is the type of elements in \(T\). - Stores "key-value" pairs.

Motivation: map

Example: Count the occurrences of strings.

std::map<std::string, int> counter; // maps every string to an integer
std::string word;
while (std::cin >> word)
  ++counter[word]; // !!

Now for any string str, counter[str] is an integer indicating how many times str has occurred.

`std::map`: comparison with `std::set`

std::map<Key, Value> has two template parameters: Key and Value. - If we ignore Value, it is a std::set<Key>. - Duplicate keys are not allowed. - operator<(const Key, const Key) is required. - Elements are stored in ascending order of keys. - Keys cannot be modified directly. - The element type of std::map<Key, Value> is std::pair<const Key, Value>. - *iter returns std::pair<const Key, Value> &.

`std::map`: comparison with `std::set`

Constructors: - std::map<Key, Value> m{{key1, value1}, {key2, value2}, ...}; - std::map<Key, Value> m(begin, end), but the elements should be pairs:

cpp std::vector<std::pair<int, int>> v{{1, 2}, {3, 4}}; std::map<int, int> m(v.begin(), v.end());

Insertion: - m.insert({key, value}) - m.insert({{key1, value1}, {key2, value2}, ...}) - m.insert(begin, end)

`std::map`: comparison with `std::set`

Deletion: - m.erase(pos), m.erase(begin, end): same as std::set<T>::erase. - m.erase(key): Removes the element whose key is key.

Iterators: BidirectionalIterator, pointing to std::pair<const Key, Value>.

std::map<std::string, int> counter = someValues();
for (auto it = counter.begin(); it != counter.end(); ++it)
  std::cout << it->first << " occurred " << it->second << " times.\n";

`std::map`: traversal

Use range-for:

for (const auto &kvpair : counter)
  std::cout << kvpair.first << " occurred " << kvpair.second << " times.\n";

It's so annoying to deal with the pair stuff...

`std::map`: traversal

Use range-for:

for (const auto &kvpair : counter)
  std::cout << kvpair.first << " occurred " << kvpair.second << " times.\n";

It's so annoying to deal with the pair stuff...

C++17 structured binding kills the game!

for (const auto &[word, occ] : counter)
  std::cout << word << " occurred " << occ << " times.\n";

(Looks very much like Python unpacking.)

`std::map`-specific: `operator[]`

m[key] finds the key-value pair whose key is equivalent to key. - If such key does not exist, inserts {key, Value{}} - the value is value-initialized. - Then, returns reference to the value.

std::map<std::string, int> counter;
std::string word;
while (std::cin >> word)
  ++counter[word]; // If `word` does not exist in `counter`,
                   // a pair {word, 0} is inserted first.

`std::map`: element lookup

m.find(key), m.count(key), and some other member functions.

Note: m.find(key) does not insert elements. m[key] will insert an element if that key does not exist.

Other kinds of maps:

Maps based on red-black trees: - std::map - std::multimap: allows duplicate keys

Maps based on hash-tables: (since C++11) - std::unordered_map: hash-table version of std::map - std::unordered_multimap: allows duplicate keys

Maps based on hash-tables provides (average-case) \(O(1)\) time operations, but requires the key to be hashable.

Summary

Sequence containers

std::vector<T>: dynamic contiguous array (we are quite familiar with)

std::deque<T>: double-ended queue (often pronounced as "deck")

std::array<T, N>: same as T[N], but it is a container
std::list<T>: doubly-linked list

std::forward_list<T>: singly-linked list

Summary

Associative containers

std::set<T>: A finite set \(\{e_1,e_2,\cdots,e_n\}\) where elements are of type T.
std::map<Key, Value>: A map \(f:S\mapsto T\), where \(S\) and \(T\) are the sets of values of type Key and Value respectively.
std::set and std::map are ordered: T and Key need to have an ordering, either in the form of operator< or some user-supplied comparator.
std::unordered_set and std::unordered_map are unordered and hash-based.

CS100 Lecture 23

Contents

Overview of STL

Standard Template Library

Containers

Iterators

Without iterators:

Iterators

With iterators:

Algorithms

Algorithms

Function objects

Adaptors

Sequence containers

Sequence containers

Sequence containers

Interfaces

Interfaces

Interfaces

Interfaces

Interfaces

Iterators

Iterator categories

Iterator categories

Iterator categories

Constructors of containers

Constructors of containers

Associative containers

Motivation: set

std::set

std::set

std::set

std::set: operations

std::set: insertion

std::set: iterators

std::set: iterators

std::set: traversal

std::set: deletion

std::set: element lookup

std::set: pros and cons

Other kinds of sets:

Motivation: map

Motivation: map

std::map: comparison with std::set

std::map: comparison with std::set

std::map: comparison with std::set

std::map: traversal

std::map: traversal

std::map-specific: operator[]

std::map: element lookup

Other kinds of maps:

Summary

Summary

`std::set`

`std::set`

`std::set`

`std::set`: operations

`std::set`: insertion

`std::set`: iterators

`std::set`: iterators

`std::set`: traversal

`std::set`: deletion

`std::set`: element lookup

`std::set`: pros and cons

`std::map`: comparison with `std::set`

`std::map`: comparison with `std::set`

`std::map`: comparison with `std::set`

`std::map`: traversal

`std::map`: traversal

`std::map`-specific: `operator[]`

`std::map`: element lookup