Boost::multi_index allows storing and accessing data in multiple ways simultaneously by maintaining multiple indices. It can index the same data by different fields, like id and name, without needing to duplicate or synchronize data. Users define the indices upfront, such as ordered by id and name. This allows efficient lookup and iteration by either field without additional containers or synchronization. Boost::multi_index provides common index types like ordered and hashed indices as well as tools to extract keys from objects. It offers powerful and flexible indexing of data.

1. multi_index
unleashing the beast
Sławomir Zborowski @

2. AGENDA
What is it?
Why is it there?
How to use it?

3. WHAT IS BOOST::MULTI_INDEX

4. SIMPLE USE CASE
1 struct Employee {
2 Employee(int const id, string const& name)
3 : id(id)
4 , name(name)
5 { }
6
7 bool operator<(Employee const& other) const {
8 return id < other.id;
9 }
10
11 int id;
12 string name;
13 };

5. SIMPLE USE CASE
Sorting by id → easy
Sorting by name → not really …
copy to vector & sort with other functor
maintain secondary container with pointers (like view
in databases)
Synchronization needed
Exceptions = real trouble

6. STD

7. MULTI_INDEX - HOW TO USE
1 typedef multi_index_container<
2 Employee,
3 indexed_by<
4 ordered_unique<identity<Employee>>,
5 ordered_non_unique<member<Employee, string,
&Employee::name>>
6 >
7 > Employees;
8
9 Employees employees = {/* ... */};
10 Employees::nth_index<1>::type const& nameIndex =
employees.get<1>();
11
12 // begin(nameIndex), end(nameIndex), ...

8. ALICE'S ADVENTURES IN WONDERLAND
ALICE was beginning to get very tired of
sitting by her sister on the bank and of
having nothing to do: once or twice she
had peeped into the book her sister was
reading, but it had no pictures or
conversations in it, "and what is the use
of a book," thought Alice, "without
pictures or conversations?'

9. FOLLY
TimeoutQueue.hpp
Identifier
Expiration

10. INDICES
Few indices:
sequenced: sequenced<>
ordered: ordered_unique<>,
ordered_non_unique<>
hashed: hashed_unique<>,
hashed_non_unique<>
random: random_access<>
Extractors:
base object: identity<>
object member: member<>
functions: mem_fun<>, global_fun<> , …
user-defined …

11. TAGGING

12. TAGGING
1 struct name {};
2
3 indexed_by<
4 // ...
5 ordered_non_unique<tag<name>, member<Employee,
string, &Employee::name>>
6 >
7
8 // ...
9 employees.get<1>() == employees.get<name>();

13. MORE POWER
iterators
custom comparison predicates
special lookup operations
ranges …

14. GO GET IT!
boost::multi_index

15. Q & A