A Slot Map is a high-performance associative container with persistent unique keys to access stored values. Upon insertion, a key is returned that can be used to later access or remove the values. Insertion, removal, and access are all guaranteed to take O(1)
time (best, worst, and average case)
Great for storing collections of objects that need stable, safe references but have no clear ownership.
The difference between a std::unordered_map
and a dod::slot_map
is that the slot map generates and returns the key when inserting a value. A key is always unique and will only refer to the value that was inserted.
Usage example:
slot_map<std::string> strings;
auto red = strings.emplace("Red");
auto green = strings.emplace("Green");
auto blue = strings.emplace("Blue");
const std::string* val1 = strings.get(red);
if (val1)
{
printf("red = '%s'\n", val1->c_str());
}
strings.erase(green);
printf("%d\n", strings.has(green));
printf("%d\n", strings.has(blue));
Output:
red = 'Red'
0
1
The slot map container will allocate memory in pages (default page size = 4096 elements) to avoid memory spikes during growth and be able to deallocate pages that are no longer needed. Also, the page-based memory allocator is very important since it guarantees "pointers stability"; hence, we never move values in memory.
Keys are always uses uint64_t/uint32_t
(configurable) and technically typless, but we "artificially" make them typed to get a few extra compile-time checks.
i.e., the following code will produce a compiler error
slot_map<std::string> strings;
slot_map<int> numbers;
slot_map<int>::key numKey = numbers.emplace(3);
const std::string* value = strings.get(numKey); // <---- can not use slot_map<int>::key to index slot_map<std::string> !
The keys can be converted to/from their numeric types if you do not need additional type checks.
slot_map<int> numbers;
slot_map<int>::key numKey = numbers.emplace(3);
uint64_t rawKey = numKey; // convert to numeric type (like cast pointer to void*)
...
slot_map<int>::key numKey2{rawKey}; // create key from numeric type
When a slot is reused, its version is automatically incremented (to invalidate all existing keys that refers to the same slot). But since we only use 20-bits (10-bits for 32 bit keys) for version counter, there is a possibility that the version counter will wrap around, and a new item will get the same key as a removed item.
To mitigate this potential issue, once the version counter overflows, we disable that slot so that no new keys are returned for this slot (this gives us a guarantee that there are no key collisions)
To prevent version overflow from happening too often, we need to ensure that we don't reuse the same slot too often.
So we do not reuse recently freed slot-indices as long as their number is below a certain threshold (kMinFreeIndices = 64
).
Keys also can carry a few extra bits of information provided by a user that we called tag
.
That might be handy to add application-specific data to keys.
For example:
slot_map<std::string> strings;
auto red = strings.emplace("Red");
red.set_tag(13);
auto tag = red.get_tag();
assert(tag == 13);
Here is how a key structure looks like internally
64-bit key type
Component | Number of bits |
---|---|
tag | 12 |
version | 20 (0..1,048,575 |
index | 32 (0..4,294,967,295) |
32-bit key type
Component | Number of bits |
---|---|
tag | 2 |
version | 10 (0..1023) |
index | 20 (0..1,048,575) |
Note: To use your custom memory allocator define SLOT_MAP_ALLOC
/SLOT_MAP_FREE
before including "slot_map.h"
#define SLOT_MAP_ALLOC(sizeInBytes, alignment) aligned_alloc(alignment, sizeInBytes)
#define SLOT_MAP_FREE(ptr) free(ptr)
#include "slot_map.h"
bool has_key(key k) const noexcept
Returns true if the slot map contains a specific key
void reset()
Clears the slot map and releases any allocated memory.
Note: By calling this function, you must guarantee that no handles are in use!
Otherwise calling this function might be dangerous and lead to key "collisions".
You might consider using "clear()" instead.
void clear()
Clears the slot map but keeps the allocated memory for reuse.
Automatically increases version for all the removed elements (the same as calling "erase()" for all existing elements)
const T* get(key k) const noexcept
If key exists returns a const pointer to the value corresponding to the given key or returns null elsewere.
T* get(key k)
If key exists returns a pointer to the value corresponding to the given key or returns null elsewere.
key emplace(Args&&... args)
Constructs element in-place and returns a unique key that can be used to access this value.
void erase(key k)
Removes element (if such key exists) from the slot map.
std::optional<T> pop(key k)
Removes element (if such key exists) from the slot map, returning the value at the key if the key was not previously removed.
bool empty() const noexcept
Returns true if the slot map is empty.
size_type size() const noexcept
Returns the number of elements in the slot map.
void swap(slot_map& other) noexcept
Exchanges the content of the slot map by the content of another slot map object of the same type.
slot_map(const slot_map& other)
Copy constructor
slot_map& operator=(const slot_map& other)
Copy assignment
slot_map(slot_map&& other) noexcept
Move constructor
slot_map& operator=(slot_map&& other) noexcept
Move asignment
const_values_iterator begin() const noexcept
const_values_iterator end() const noexcept
Const values iterator
for (const auto& value : slotMap)
{
...
}
Items items() const noexcept
Const key/value iterator
for (const auto& [key, value] : slotMap.items())
{
...
}
Sean Middleditch
Data Structures for Game Developers: The Slot Map, 2013
https://web.archive.org/web/20180121142549/http://seanmiddleditch.com/data-structures-for-game-developers-the-slot-map/
Niklas Gray
Building a Data-Oriented Entity System (part 1), 2014
http://bitsquid.blogspot.com/2014/08/building-data-oriented-entity-system.html
Noel Llopis
Managing Data Relationships, 2010
https://gamesfromwithin.com/managing-data-relationships
Stefan Reinalter
Adventures in data-oriented design - Part 3c: External References, 2013
https://blog.molecular-matters.com/2013/07/24/adventures-in-data-oriented-design-part-3c-external-references/
Niklas Gray
Managing Decoupling Part 4 - The ID Lookup Table, 2011
https://bitsquid.blogspot.com/2011/09/managing-decoupling-part-4-id-lookup.html
Sander Mertens
Making the most of ECS identifiers, 2020
https://ajmmertens.medium.com/doing-a-lot-with-a-little-ecs-identifiers-25a72bd2647
Michele Caini
ECS back and forth. Part 9 - Sparse sets and EnTT, 2020
https://skypjack.github.io/2020-08-02-ecs-baf-part-9/
Andre Weissflog
Handles are the better pointers, 2018
https://floooh.github.io/2018/06/17/handles-vs-pointers.html
Allan Deutsch
C Now 2017: "The Slot Map Data Structure", 2017
https://www.youtube.com/watch?v=SHaAR7XPtNU
Jeff Gates
Init, Update, Draw - Data Arrays, 2012
https://greysphere.tumblr.com/post/31601463396/data-arrays
Niklas Gray
Data Structures Part 1: Bulk Data, 2019
https://ourmachinery.com/post/data-structures-part-1-bulk-data/