PLF C++ Library - plf::colony

Assign

• void assign (const size_type n, const value_type &val)

  Assigns n copies of val to the colony. Colony will erase all current contents and attempt to preserve existing memory blocks. Example:

  plf::colony<unsigned int> i_colony;
i_colony.assign(10, 3);
• template <class InputIterator> void assign (const InputIterator &first, const InputIterator &last)

  Assigns a series of value_type elements from an external source to the colony. Example:

  // Assign all contents of colony2 into colony1:
colony1.assign(colony2.begin(), colony2.end());
• void assign (const std::initializer_list<value_type> &il)

  Assigns elements from an initializer list to the colony. Example:

  std::initializer_list<int> some_ints = {4, 3, 2, 5};

plf::colony<int> i_colony;
i_colony.assign(some_ints);

Erase

• iterator erase(const const_iterator it)

  Removes the element pointed to by the supplied iterator, from the colony. Returns an iterator pointing to the next non-erased element in the colony (or to end() if no more elements are available). This must return an iterator because if a colony group becomes entirely empty, it will be removed from the colony, invalidating the existing iterator. Attempting to erase a previously-erased element results in undefined behaviour (this is checked for via an assert in debug mode). Example:

  plf::colony<unsigned int> data_colony(50);
plf::colony<unsigned int>::iterator an_iterator;
an_iterator = data_colony.insert(23);
an_iterator = data_colony.erase(an_iterator);
• iterator erase(const const_iterator first, const const_iterator last)

  Erases all elements of a given colony from first to the element before the last iterator. This function is optimized for multiple consecutive erasures and will always be faster than sequential single-element erase calls in that scenario. Example:

  plf::colony<int> iterator1 = colony1.begin();
colony1.advance(iterator1, 10);
plf::colony<int> iterator2 = colony1.begin();
colony1.advance(iterator2, 20);
colony1.erase(iterator1, iterator2);

Other functions

• [[nodiscard]] bool empty() const noexcept

  Returns a boolean indicating whether the colony is currently empty of elements.
  Example: if (object_colony.empty()) return;

• size_type size() const noexcept

  Returns total number of elements currently stored in container.
  Example: std::cout << i_colony.size() << std::endl;

• size_type max_size() const noexcept

  Returns the maximum number of elements that the allocator can store in the container.
  Example: std::cout << i_colony.max_size() << std::endl;

• size_type capacity() const noexcept

  Returns total number of elements currently able to be stored in container without expansion.
  Example: std::cout << i_colony.capacity() << std::endl;

• size_type memory() const noexcept

  Returns the memory use of the container plus it's elements. Does not include memory dynamically-allocated by the elements themselves.
  Example: std::cout << i_colony.memory() << std::endl;

• void shrink_to_fit()

  Reduces container capacity to the amount necessary to store all currently stored elements, consolidates elements and removes any erased locations. If the total number of elements is larger than the maximum group size, the resultant capacity will be equal to ((total_elements / max_group_size) + 1) * max_group_size (rounding down at division). Invalidates all pointers, iterators and references to elements within the container.
  Example: i_colony.shrink_to_fit();

• void trim_capacity() noexcept
void trim_capacity(size_type n) noexcept

  Deallocates any unused groups retained during erase() or unused since the last reserve(). Unlike shrink_to_fit(), this function cannot invalidate iterators or pointers to elements, nor will it shrink the capacity to match size(). If n parameter is specified, will not remove a group if doing so will reduce the total capacity of the colony below n elements.
  Example: i_colony.trim_capacity();

• void reserve(const size_type reserve_amount)

  Preallocates memory space sufficient to store the number of elements indicated by reserve_amount. This function is useful from a performance perspective when the user is inserting elements singly, but the overall number of insertions is known in advance. By reserving, colony can forgo creating many smaller memory block allocations (due to colony's growth factor) and reserve larger memory blocks instead.
  Example: i_colony.reserve(15);

• void clear() noexcept

  Destroys all elements and sets size() to zero. Currently existing groups are put in the 'reserved groups' list for future use. To clear these groups as well, follow up clear() with shrink_to_fit(), or just call reset() instead (faster).
  Example: object_colony.clear();

• void reset() noexcept

  Destroys all elements, sets size() to zero, and deallocates all groups. Unlike clear(), existing groups are not retained. Faster than clear() + shrink_to_fit().
  Example: object_colony.reset();

• void reshape(plf::colony_limits minmax)

  Changes the minimum and maximum internal group sizes, in terms of number of elements stored per group. If the colony is not empty and either minmax.min is larger than the smallest group in the colony, or minmax.max is smaller than the largest group in the colony, the colony will be internally copy-constructed into a new colony which uses the new group sizes, invalidating all pointers/iterators/references. If trying to change group sizes with a colony storing a non-copyable/movable type, an exception is thrown. If minmax.min or minmax.max are outside of the colony's hard limits, an exception is thrown.
  Example: object_colony.reshape(plf::colony_limits(100, 1000));

• constexpr plf::colony_limits block_capacity_limits() const noexcept

  Returns the current minimum and maximum block limits within the .min and .max fields of a limits struct.
  Example: plf::colony_limits temp = object_colony.block_capacity_limits();

• static constexpr plf::colony_limits block_capacity_hard_limits() noexcept

  Returns the current minimum and maximum hard block limits within the .min and .max fields of a limits struct. In this context 'hard' means, if the user supplies a block capacity limit to a constructor or reshape(), which is above the hard maximum or below the hard minimum, an exception is thrown. As a static member function, this is designed to be able to be called before instatiating a colony using user-defined limits, or before calling reshape().

• static constexpr plf::colony_limits block_capacity_default_limits() noexcept

  Returns the current minimum and maximum default block limits for a given colony type within the .min and .max fields of a limits struct.

• static constexpr size_type block_capacity_default_min() noexcept

  Returns the default minimum block limit for a given colony type.

• static constexpr size_type block_capacity_default_max() noexcept

  Returns the default maximum block limit for a given colony type.

• void swap(colony &source) noexcept(std::allocator_traits<allocator_type>::propagate_on_container_swap::value || std::allocator_traits<allocator_type>::is_always_equal::value)

  Swaps the colony's contents with that of source.
  Example: object_colony.swap(other_colony);

• void sort();

template <class comparison_function>
void sort(comparison_function compare);

  Sort the content of the colony. By default this compares the colony content using a less-than operator, unless the user supplies a comparison function (i.e. same conditions as std::list's sort function). Uses std::sort internally but will use another sort function if it's name (including namespace) is defined as the macro PLF_COLONY_SORT_FUNCTION before including the colony .h. Supplied sort function must take same parameters as std::sort.
  Example: // Sort a colony of integers in ascending order:
int_colony.sort();
// Sort a colony of doubles in descending order:
double_colony.sort(std::greater<double>());

• void unique()

template <class comparison_function>
void unique(comparison_function compare);

  Removes duplicates of values which are consecutive in the iterative sequence. It is assumed that sort() will have been called before this function, or that elements have been inserted in a specific order. By default this compares the colony content using a == operator, unless the user supplies a comparison function (i.e. same conditions as std::list's unique function).
  Example: // Remove duplicates in a colony of integers in ascending order:
int_colony.unique();

• void splice(colony &source)

  Transfer all elements from source colony into destination colony without invalidating pointers/iterators to either colony's elements (in other words the destination takes ownership of the source's memory blocks). After the splice, the source colony is empty. Splicing is much faster than range-moving or copying all elements from one colony to another. Colony does not guarantee a particular order of elements after splicing, for performance reasons; the insertion location of source elements in the destination colony is chosen based on the most positive performance outcome for subsequent iterations/insertions. For example if the destination colony is {1, 2, 3, 4} and the source colony is {5, 6, 7, 8} the destination colony post-splice could be {1, 2, 3, 4, 5, 6, 7, 8} or {5, 6, 7, 8, 1, 2, 3, 4}, depending on internal state of both colonies and prior insertions/erasures.

  Note: If the minimum group size of the source is smaller than the destination, the destination will change it's minimum group size to match the source. The same applies for maximum group sizes (if source's is larger, the destination will adjust its).
  Example: // Splice two colonies of integers together:
colony<int> colony1 = {1, 2, 3, 4}, colony2 = {5, 6, 7, 8};
colony1.splice(colony2);

• colony_data * data();
  

  Returns a pointer to a dynamically-allocated struct with the following members:

  colony::aligned_pointer_type * const block_pointers; // dynamically-allocated array of pointers to element memory blocks
unsigned char * * const bitfield_pointers; // dynamically-allocated array of pointers to bitfields in the form of unsigned char arrays
// these represent whether an element is erased or not (0 for erased).
// they can be reinterpret_cast'd to whatever SIMD mask bitdepth is needed.
size_t * const block_capacities; // dynamically-allocated array of the number of elements in each memory block
const size_t number_of_blocks; // size of each of the arrays above

  This can be used with SIMD intrinsics to perform gather/scatter operations on colony non-erased elements. Currently AVX512 is the only intel instruction set with performance-efficient gather/scatter ops. The structure must be freed with delete when finished with. See test suite for a more thorough inspection of the structure.

  Example: colony<int> i_colony(10000, 5);
colony<int>::colony_data *data = i_colony.data();
// do stuff
delete data;

• colony & operator = (const colony &source)

  Copy the elements from another colony to this colony, clearing this colony of existing elements first.
  Example: // Manually swap data_colony1 and data_colony2 in C++03
data_colony3 = data_colony1;
data_colony1 = data_colony2;
data_colony2 = data_colony3;

• colony & operator = (colony &&source) noexcept(std::allocator_traits<allocator_type>::propagate_on_container_move_assignment::value || std::allocator_traits<allocator_type>::is_always_equal::value)C++11 and upwards

  Move the elements from another colony to this colony, clearing this colony of existing elements first. Source colony is now empty and in a valid state (same as a new colony without any insertions), can be safely destructed or used in any regular way without problems.
  Example: // Manually swap data_colony1 and data_colony2 in C++11
data_colony3 = std::move(data_colony1);
data_colony1 = std::move(data_colony2);
data_colony2 = std::move(data_colony3);

• colony & operator = (const std::initializer_list<value_type> il))

  Copy the elements from an initializer list to this colony, clearing this colony of existing elements first.

• friend bool operator == (const colony &lh, const colony &rh) noexcept

  Compare contents of one colony to this colony. Returns a boolean as to whether their iterative sequences are equal.
  Example: if (object_colony == object_colony2) return;

• friend bool operator != (const colony &lh, const colony &rh) noexcept

  Compare contents of another colony to this colony. Returns a boolean as to whether their iterative sequences are not equal.
  Example: if (object_colony != object_colony2) return;

• iterator begin(), const_iterator begin(), iterator end(), const_iterator end(), const_iterator cbegin(), const_iterator cend()

  Return iterators pointing to, respectively, the first element of the colony and the element one-past the end of the colony.

• reverse_iterator rbegin(), const_reverse_iterator rbegin(), reverse_iterator rend(), const_reverse_iterator rend(), const_reverse_iterator crbegin(), const_reverse_iterator crend()

  Return reverse iterators pointing to, respectively, the last element of the colony and the element one-before the first element of the colony.

• iterator get_iterator(const_pointer element_pointer) noexcept
const_iterator get_const_iterator(const_pointer element_pointer) const noexcept (non-O(1))

  Getting a pointer from an iterator is simple - simply dereference it then grab the address i.e. "&(*the_iterator);". Getting an iterator from a pointer is typically not so simple. This function enables the user to do exactly that. This is expected to be useful in the use-case where external containers are storing pointers to colony elements instead of iterators (as iterators for colonies have 3 times the size of an element pointer) and the program wants to erase the element being pointed to or possibly change the element being pointed to. Converting a pointer to an iterator using this method and then erasing, is about 20% slower on average than erasing when you already have the iterator.

  This is less dramatic than it sounds, as it is still faster than all std:: container erasure times which it is roughly equal to. However this is generally a slower, lookup-based operation. If the lookup doesn't find a non-erased element based on that pointer, it returns end(). Otherwise it returns an iterator pointing to the element in question. For const pointers, use a const_cast when supplying the pointer to the function, and for const_iterator, simply supply a const_iterator for the return value. Example:

  plf::colony<a_struct> data_colony;
plf::colony<a_struct>::iterator an_iterator;
a_struct struct_instance;
an_iterator = data_colony.insert(struct_instance);
a_struct *struct_pointer = &(*an_iterator);
iterator another_iterator = data_colony.get_iterator(struct_pointer);
if (an_iterator == another_iterator) std::cout << "Iterator is correct" << std::endl;

• bool is_active(const const_iterator it) const

  This is similar to get_iterator(pointer), but for iterators. For a given iterator, returns whether or not the location the iterator points to is still valid ie. that it points to an active non-erased element in a memory block which is still allocated. It does not differentiate between memory blocks which have been re-used - the only thing it reports, is whether the memory block pointed to by the iterator exists in an active state in this particular colony, and if there is an element being pointed to by the iterator, in that memory block, which is also in an active, non-erased state.

• static constexpr size_type block_metadata_memory(const size_type block_capacity) noexcept

  Given a block element capacity, tells you, for this particular colony type, what the resultant amount of metadata (including skipfield) allocated for that block will be, in bytes.

• static constexpr size_type block_allocation_amount(const size_type block_capacity) noexcept

  Given a block element capacity, tells you, for this particular colony type, what the resultant allocation for that block will be, in bytes. In terms of this version of colony, the block and skipfield are allocated at the same time (other block metadata is allocated separately). So this function is expected to be primarily useful for developers working in environments where an allocator or system only allocates a specific capacity of memory per call, eg. some embedded platforms.

• static constexpr size_type max_elements_per_allocation(const size_type allocation_amount) noexcept

  Given a fixed allocation amount, tells you, for this particular colony type, exactly what the maximum element capacity per-block can be. In other words, if an allocator can only supply a fixed or low amount of memory per allocation, this function tells the developer the upper limit of what their max block capacity limit should be set to in the constructor (or in reshape()). Like the above this function is expected to be primarily useful for developers working in embedded environments. The return value is bounded by the colony's maximum hard block capacity limit (255 or 65535 depending on the type and priority_type used) and will return zero if the return value is below the colony's minimum hard block capacity limit (currently 3).

Non-member function overloads

• template <class iterator_type> void advance(iterator_type iterator, const distance_type distance)

  Increments/decrements the iterator supplied by the positive or negative amount indicated by distance. Speed of incrementation will almost always be faster than using the ++ operator on the iterator for increments greater than 1. In some cases it may approximate O(1). The iterator_type can be an iterator, const_iterator, reverse_iterator or const_reverse_iterator.
  Example: colony<int>::iterator it = i_colony.begin();
i_colony.advance(it, 20);

• template <class iterator_type> iterator_type next(iterator_type &iterator, const distance_type distance)

  Creates a copy of the iterator supplied, then increments/decrements this iterator by the positive or negative amount indicated by distance.
  Example: colony<int>::iterator it = i_colony.next(i_colony.begin(), 20);

• template <class iterator_type> iterator_type prev(iterator_type &iterator, const distance_type distance)

  Creates a copy of the iterator supplied, then decrements/increments this iterator by the positive or negative amount indicated by distance.
  Example: colony<int>::iterator it2 = i_colony.prev(i_colony.end(), 20);

• template <class iterator_type> difference_type distance(const iterator_type &first, const iterator_type &last)

  Measures the distance between two iterators, returning the result, which will be negative if the second iterator supplied is before the first iterator supplied in terms of it's location in the colony.
  Example: colony<int>::iterator it = i_colony.next(i_colony.begin(), 20);
colony<int>::iterator it2 = i_colony.prev(i_colony.end(), 20);
std::cout "Distance: " i_colony.distance(it, it2) std::endl;

• template <class container_type> void std::swap(container_type &A, container_type &B)

  Swaps colony A's contents with that of colony B (assumes both colonies have same element type, allocator type, etc).
  Example: swap(object_colony, other_colony);

• template <class container_type, class predicate> size_type std::erase_if(container_type A, predicate pred)

  Erases all elements in the colony which match predicate pred. Returns the number of erased elements.

• template <class container_type> size_type std::erase(container_type A, container_type::value_type v)

  Erases all elements in the colony which == v. Returns the number of erased elements.

• template <plf::colony_iterator_concept it_type>
class reverse_iterator<it_type> : public it_type::reverse_type

  Overload for std::reverse_iterator, for returning a reverse_iterator from a supplied colony iterator/const_iterator. Primarily intended for use with ranges v3.

Benchmarks

Benchmark results for colony v7.39 under GCC 11.2 x64 on an Intel 2018 i5 are here.
Benchmark results for colony v6.11 under GCC 9.2 x64 on an Intel Xeon E3-1241 (Haswell) are here.

Old benchmark results for colony v3.87 under GCC 5.1 x64 on an Intel E8500 (Core2) are here,
and under MSVC 2015 update 3 on an Intel Xeon E3-1241 (Haswell) are here. There is no commentary for the MSVC results.

Frequently Asked Questions

1. TLDR, what is a colony?

   A combination of a linked-list of increasingly-large memory blocks with metadata. Part of this metadata is the head of a free-list of erased elements within that group, and also a 'next' pointer to the next group which has erased element locations which can be re-used. Another part of the metadata is a low complexity jump-counting skipfield, which is used to skip over erased elements during iteration in O(1) amortised time. The end result's a bidirectional, unordered C++ template data container which maintains positive performance characteristics while ensuring pointer stability to non-erased container elements.

2. Where is it worth using a colony in place of other std:: containers?

   As mentioned, it is worthwhile for performance reasons in situations where the order of container elements is not important and:

   1. Insertion order is unimportant
   2. Insertions and erasures to the container occur frequently in performance-critical code, and
   3. Links to non-erased container elements may not be invalidated by insertion or erasure.

   Under these circumstances a colony will generally out-perform other std:: containers. In addition, because it never invalidates pointer references to container elements (except when the element being pointed to has been previously erased) it may make many programming tasks involving inter-relating structures in an object-oriented or modular environment much faster, and could be considered in those circumstances.

3. What are some examples of situations where a colony might improve performance?

   Some ideal situations to use a colony: cellular/atomic simulation, persistent octtrees/quadtrees, game entities or destructible-objects in a video game, particle physics, anywhere where objects are being created and destroyed continuously. Also, anywhere where a vector of pointers to dynamically-allocated objects or a std::list would typically end up being used in order to preserve object references but where order is unimportant.

4. What are the time complexities for general operations?

   In most cases the time complexity of operations results from the fundamental requirements of the container itself, not the implementation. In some cases, such as erase complexity, this can change a little based on the implementation, but the change is minor.

   Insert (single): O(1) amortised

   One of the requirements of colony is that pointers to non-erased elements stay valid regardless of insertion/erasure within the container. For this reason the container must use multiple memory blocks. If a single memory block were used, like in a std::vector, reallocation of elements would occur when the container expanded (and the elements were copied to a larger memory block). Hence, colony will insert into existing memory blocks when able, and create a new memory block when all existing memory blocks are full.

   Because colony (by default) has a growth pattern for new memory blocks, the number of insertions before a new memory block is required increases as more insertions take place. Hence the insertion time complexity is O(1) amortised.

   Insert (multiple): O(N) amortised

   This follows the same principle - there will occasionally be a need to create a new memory block, but since this is infrequent it is O(N) amortised.

   Erase (single): O(1) amortised (as of v5.00)

   Generally speaking erasure is a simple matter of destructing the element in question and updating the skipfield. Since we started using the low complexity jump-counting pattern (v5.00) instead of the high complexity jump-counting skipfield pattern, that skipfield update is always O(1). However, when a memory block becomes empty of non-erased elements it must be freed to the OS (or stored for future insertions, depending on implementation) and removed from the colony's sequence of memory blocks. It it was not, we would end up with non-O(1) amortised iteration, since there would be no way to predict how many empty memory blocks there would be between the current memory block being iterated over, and the next memory block with non-erased (active) elements in it. Since removal of memory blocks is infrequent the time complexity becomes O(1) amortised.

   Erase (multiple): O(N) amortised for non-trivially-destructible types, for trivially-destructible types between O(1) and O(N) amortised depending on range start/end (approximating O(log n) average)

   In this case, where the element is non-trivially destructible, the time complexity is O(N) amortised, with infrequent deallocation necessary from the removal of an empty memory block as noted above. However where the elements are trivially-destructible, if the range spans an entire memory block at any point, that block and it's skipfield can simply be removed without doing any individual writes to it's skipfield or individual destruction of elements, potentially making this a O(1) operation.

   In addition (when dealing with trivially-destructible types) for those memory blocks where only a portion of elements are erased by the range, if no prior erasures have occurred in that memory block you can erase that range in O(1) time, as there will be no need to check within the range for previously erased elements, and the low complexity jump-counting pattern only requires two skipfield writes to indicate a range of skipped nodes. The reason you would need to check for previously erased elements within that portion's range is so you can update the metadata for that memory block to accurately reflect how many non-erased elements remain within the block. If that metadata is not present, there is no way to ascertain when a memory block is empty of non-erased elements and hence needs to be removed from the colony. The reasoning for why empty memory blocks must be removed is included in the Erase(single) section, above.

   However in most cases the erase range will not perfectly match the size of all memory blocks, and with typical usage of a colony there is usually some prior erasures in most memory blocks. So for example when dealing with a colony of a trivially-destructible type, you might end up with a tail portion of the first memory block in the erasure range being erased in O(N) time, the second and intermediary memory block being completely erased and freed in O(1) time, and only a small front portion of the third and final memory block in the range being erased in O(N) time. Hence the time complexity for trivially-destructible elements approximates O(log n) on average, being between O(1) and O(N) depending on the start and end of the erasure range.

   std::find: O(n)

   This relies on basic iteration so is O(N).

   splice: O(1)

   Colony only does full-container splicing, not partial-container splicing (use range-insert with std::move to achieve the latter, albiet with the loss of pointer validity to the moved range). When splicing the memory blocks from the source colony are transferred to the destination colony without processing the individual elements or skipfields, inserted after the destination's final block. However if there are unused element memory spaces at the back of the destination container (ie. the final memory block is not full), the skipfield nodes corresponding to those empty spaces must be altered to indicate that these are skipped elements. Luckily as of v5.00 this is also a O(1) operation, hence the overall operation is O(1).

   Iterator operators ++ and --: O(1) amortised

   Generally the time complexity is O(1), as the skipfield pattern used (current) must allow for skipping of multiple erased elements. However every so often iteration will involve a transistion to the next/previous memory block in the colony's sequence of blocks, depending on whether we are doing ++ or --. At this point a read of the next/previous memory block's corresponding skipfield is necessary, in case the first/last element(s) in that memory block are erased and hence skipped. So for every block transition, 2 reads of the skipfield are necessary instead of 1. Hence the time complexity is O(1) amortised.

   Skipfields must be per-block and independent between memory blocks, as otherwise you would end up with a vector for a skipfield, which would need a range erased every time a memory block was removed from the colony (see notes under Erase above), and reallocation to larger skipfield memory block when the colony expanded. Both of these procedures carry reallocation costs, meaning you could have thousands of skipfield nodes needing to be reallocated based on a single erasure (from within a memory block which only had one non-erased element left and hence would need to be removed from the colony). This is unacceptable latency for any fields involving high timing sensitivity (all of SG14).

   begin()/end(): O(1)

   For any implementation these should generally be stored as member variables and so returning them is O(1).

   advance/next/prev: between O(1) and O(n), depending on current iterator location and distance. Average approximates O(log N).

   The reasoning for this is similar to that of Erase(multiple), above. Memory blocks which are completely traversed by the increment/decrement distance can be skipped in O(1) time by reading the metadata for the block which details the number of non-erased elements in the block. If the current iterator location is not at the very beginning of a memory block, then the distance within that block will need to be processed in O(N) time, unless no erasures have occurred within that block, in which case the skipfield does not need to be checked and traversal is simply a matter of pointer arithmetic and checking the distance between the current iterator position and the end of the memory block, which is a O(1) operation.

   Similarly for the final block involved in the traversal (if the traversal spans multiple memory blocks!) if no erasures have occurred in that memory block pointer arithmetic can be used and the procedure is O(1), otherwise it is O(N) (++ iterating over the skipfield). We can therefore say that the overall time complexity for these operations might approximate O(log N), while being between O(1) and O(N) amortised respectively.

5. Is it similar to a deque?

   A deque is reasonably dissimilar to a colony - being a double-ended queue, it requires a different internal framework. It typically uses a vector of memory blocks, whereas the colony implementation uses a linked list of memory blocks, essentially. A deque can't technically use a linked list of memory blocks because it will make some random_access iterator operations (e.g. + operator) non-O(1). In addition, being random-access container, having a growth factor for memory blocks in a deque is problematic (not impossible though).

   A deque and colony have no comparable performance characteristics except for insertion (assuming a good deque implementation). Deque erasure performance varies wildly depending on the implementation compared to std::vector, but is generally similar to vector erasure performance. A deque invalidates pointers to subsequent container elements when erasing elements, which a colony does not.

6. What are the thread-safe guarantees?

   Unlike a std::vector, a colony can be read from and inserted into at the same time (assuming different locations for read and write), however it cannot be iterated over and written to at the same time. If we look at a (non-concurrent implementation of) std::vector's thread-safe matrix to see which basic operations can occur at the same time, it reads as follows (please note push_back() is the same as insertion in this regard):

   std::vector	Insertion	Erasure	Iteration	Read
   Insertion	No	No	No	No
   Erasure	No	No	No	No
   Iteration	No	No	Yes	Yes
   Read	No	No	Yes	Yes

   In other words, multiple reads and iterations over iterators can happen simultaneously, but the potential reallocation and pointer/iterator invalidation caused by insertion/push_back and erasure means those operations cannot occur at the same time as anything else.

   Colony on the other hand does not invalidate pointers/iterators to non-erased elements during insertion and erasure, resulting in the following matrix:

   plf::colony	Insertion	Erasure	Iteration	Read
   Insertion	No	No	No	Yes
   Erasure	No	No	No	Mostly*
   Iteration	No	No	Yes	Yes
   Read	Yes	Mostly*	Yes	Yes

   * Erasures will not invalidate iterators unless the iterator points to the erased element.

   In other words, reads may occur at the same time as insertions and erasures (provided that the element being erased is not the element being read), multiple reads and iterations may occur at the same time, but iterations may not occur at the same time as an erasure or insertion, as either of these may change the state of the skipfield which's being iterated over. Note that iterators pointing to end() may be invalidated by insertion.

   So, colony could be considered more inherently thread-safe than a (non-concurrent implementation of) std::vector, but still has some areas which would require mutexes or atomics to navigate in a multithreaded environment.

7. Any pitfalls to watch out for?

   1. Because erased-element memory locations will be reused by insert() and emplace(), insertion position is essentially random unless no erasures have been made, or an equal number of erasures and insertions have been made.
   2. Due to implementation reasons it is possible to assign (or construct) a iterator from a const_iterator, which is not really good const behaviour (only assigning/constructing a const_iterator from an iterator is good const practice). However preventing this results in errors in all versions of MSVC, so it has been ignored.

8. Am I better off storing iterators or pointers to colony elements?

   Testing so far indicates that storing pointers and then using get_iterator() when or if you need to do an erase operation on the element being pointed to, yields better performance than storing iterators and performing erase directly on the iterator. This is simply due to the size of iterators (3 pointers in width).

9. Any special-case uses?

   In the special case where many, many elements are being continually erased/inserted in realtime, you might want to experiment with limiting the size of your internal memory groups in the constructor. The form of this is as follows:
   plf::colony<object> a_colony(plf::colony_limits(500, 5000));
   where the first number is the minimum size of the internal memory groups and the second is the maximum size. Note these can be the same size, resulting in an unchanging group size for the lifetime of the colony (unless reshape is called or operator = is called).

   One reason to do this is that it is slightly slower to obtain an element location from the list of groups-with-erasures (subsequently utilising that group's free list of erased locations), than it is to insert a new element to the end of the colony (the default behaviour when there are no previously-erased elements). If there are any erased elements in the colony, the colony will recycle those memory locations, unless the entire group is empty, at which point it is freed to memory. So if a group size is large and many, many erasures occur but the group is not completely emptied, iterative performance may suffer due to large memory gaps between any two non-erased elements. In that scenario you may want to experiment with benchmarking and limiting the minimum/maximum sizes of the groups, and find the optimal size for a particular use case.

   Please note that the the fill, range and initializer-list constructors can also take group size parameters, making it possible to construct filled colonies using custom group sizes.

10. What is colony's Abstract Data Type (ADT)?

    Though I am happy to be proven wrong I suspect colony is it's own abstract data type. While it is similar to a multiset or bag, those utilize key values and are not sortable (by means other than automatically by key value). Colony does not utilize key values, is sortable, and does not provide the sort of functionality one would find in a bag (e.g. counting the number of times a specific value occurs). Some have suggested similarities to deque - but as described earlier the three core aspects of colony are:

    1. A multiple-memory-block based allocation pattern which allows for the removal of memory blocks when they become empty of elements.
    2. A skipfield to indicate erasures instead of reallocating elements, the iteration of which should typically not necessitate the use of branching code.
    3. A mechanism for recording erased element locations to allow for reuse of erased element memory space upon subsequent insertion.

    The only aspect out of these which deque also shares is a multiple-memory-block allocation pattern - not a strong association. As a result, deques do not guarantee pointer validity to non-erased elements post insertion or erasure, as colony does. Similarly if we look at a multiset, an unordered one could be implemented on top of a colony by utilising a hash table (and would in fact be more efficient than most non-flat implementations) but the fact that there is a necessity to add something to make it a multiset (to take and store key values) means colony is not an multiset.

11. Exception Guarantees?

    All operations which allocate memory have strong exception guarantees and will roll back if an exception occurs, except for operator = which has a basic exception guarantee (see below). For colony, iterators are bounded by asserts in debug mode, but unbounded in release mode, so it is possible for an incorrect use of an iterator (iterating past end(), for example) to trigger an out-of-bounds memory exception.

    The reason why operator = only has a basic guarantee is they do not utilize the copy-swap idiom, as the copy-swap idiom significantly increases the chance of any exception occurring - this is because the most common way an exception tends to occur during a copy operation is due to a lack of memory space, and the copy-swap idiom doubles the memory requirement for a copy operation by constructing the copy before destructing the original data.

    This is doubly inappropriate in game development, which colony has been initially for, where memory constraints are often critical and the runtime lag from memory spikes can cause detrimental game performance. So in the case of colony if a non-memory-allocation-based exception occurs during copy, the = operators will have already destructed their data, so the containers will be empty and cannot roll back - hence they have a basic guarantee, not a strong guarantee.

12. Iterators not following rule of zero?

    It was found under GCC and clang that the high-modification scenario benchmarks suffered a 11% overall performance loss when iterator copy and move constructors/operators where not explicitly defined, possibly because GCC did not know to vectorize the operations without explicit code or because the compiler implemented copy/swap idiom instead of simply copying directly. No performance gain was found under any tests when removing the copy and move constructors/operators.

13. Why must groups be removed when empty?

    Two reasons:

    1. Standards compliance: if groups aren't removed then ++ and -- iterator operations become O(random) in terms of time complexity, making them non-compliant with the C++ standard. At the moment they are O(1) amortised, typically one update for both skipfield and element pointers, but two if a skipfield jump takes the iterator beyond the bounds of the current group and into the next group. But if empty groups are allowed, there could be anywhere between 1 and size_type empty groups between the current element and the next. Essentially you get the same scenario as you do when iterating over a boolean skipfield. It would be possible to move these to the back of the colony as trailing groups, but this may create performance issues if any of the groups are not at their maximum size (see below).
    2. Performance: iterating over empty groups is slower than them not being present, cache wise - but also if you have to allow for empty groups while iterating, then you have to include a while loop in every iteration operation, which increases cache misses and code size. The strategy of removing groups when they become empty also removes (assuming randomised erasure patterns) smaller groups from the colony before larger groups, which has a net result of improving iteration, because with a larger group, more iterations within the group can occur before the end-of-group condition is reached and a jump to the next group (and subsequent cache miss) occurs. Lastly, pushing to the back of a colony is faster than recycling memory locations as each insertion occurs in subsequent similar memory location (which is cache-friendlier) and also less computational work is necessary. If a group is removed it's recyclable memory locations are also removed from re-use, hence subsequent insertions are more likely to be pushed to the back of the colony.

14. Why use three pointers for iterators rather than a group pointer, single index and then dereferencing?

    It's faster. In attempting on two separate occasions to switch to an index-based iterator approach, utilising three separate pointers was consistently faster.

15. Group sizes - what are they based on, how do they expand, etc

    Group sizes start from the either the default minimum size (8 elements, larger if the type stored is small) or an amount defined by the programmer (with a minimum of 3 elements). Subsequent group sizes then increase the total capacity of the colony by a factor of 2 (so, 1st group 8 elements, 2nd 8 elements, 3rd 16 elements, 4th 32 elements etcetera) until the maximum group size is reached. The default maximum group size is the maximum possible number that the skipfield bitdepth is capable of representing (std::numeric_limits<skipfield_type>::max()). By default the skipfield bitdepth is 16 so the maximum size of a group is 65535 elements. However the skipfield bitdepth is also a template parameter which can be set to any unsigned integer - unsigned char, unsigned int, Uint_64, etc. Unsigned short (guaranteed to be at least 16 bit, equivalent to C++11's uint_least16_t type) was found to have the best performance in real-world testing due to the balance between memory contiguousness, memory waste and the restriction on skipfield update time complexity. Initially the design also fitted the use-case of gaming better (as games tend to utilize lower numbers of elements than some other fields), and that was the primary development field at the time.

16. Why store a size member for each group when this can be obtained via reinterpret_cast<element_pointer_type>(skipfield) - elements?

    Because it's faster. While it can be obtained that way, having it precalculated gives a small but significant benefit. And it's only an additional 16 bits per group under 32-bit code (under 64-bit code there is typically no additional size increase, due to struct padding and the size of pointers).

17. Can a colony be used with SIMD instructions?

    No and yes. Yes if you're careful, no if you're not.
    On platforms which support scatter and gather operations you can use colony with SIMD as much as you want, using gather to load elements from disparate or sequential locations, into a SIMD register. Then use scatter to push the post-SIMD-process values elsewhere after. The data() function is useful for this.
    On platforms which require elements to be contiguous in memory for SIMD processing, this is more complicated. When you have a bunch of erasures in a colony, there's no guarantee that your objects will be contiguous in memory, even though they are sequential during iteration. Some of them may also be in different memory blocks to each other. In these situations if you want to use SIMD with colony, you must do the following:

    • Set your minimum and maximum group sizes to multiples of the width of your SIMD instruction. If it supports 8 elements at once, set the group sizes to multiples of 8.
    • Either never erase from the colony, or:
      
      1. Shrink-to-fit after you erase (will invalidate all pointers to elements within the colony).
      2. Only erase from the back or front of the colony, and only erase elements in multiples of the width of your SIMD instruction e.g. 8 consecutive elements at once. This will ensure that the end-of-memory-block boundaries line up with the width of the SIMD instruction, provided you've set your min/max block sizes as above.

    Generally if you want to use SIMD without gather/scatter, it's probably preferable to use a vector or an array.

Version history

• 2024-12-14: v7.5.5 - Corrections to move constructor, copy assignment, move assignment, fill insert/assign/construct. Minor performance improvement to copy assignment under C++20 and above.
• 2024-12-03: v7.5.3 - Added hint-supporting versions of insert, and emplace_hint (the hint is ignored in both cases) to support std::insert_iterator().
• 2024-11-04: v7.5.2 - Corrections to const use on move assignment/construction in iterators.
• 2024-10-30: v7.5.1 - Optimization for erase(begin, end) when there are no erasures in a block and the element type is non-trivially-destructible (or type traits are unavailable).
• 2024-10-19: v7.5.0 - Corrected block_limits functions so that they take into account the max_size() provided by the allocator. Made all default constructors constexpr. default_block_capacity_limits(), default_min_block_capacity() and default_max_block_capacity() changed to block_capacity_default_limits(), block_capacity_default_min() and block_capacity_default_max() (to match hive and be consistent with naming of block_capacity_hard_limits()).
• 2024-10-14: v7.4.8 - West const 4 lyfe - there was inconsistent use of const throughout due to my misunderstanding of how typedefs change const expession in C++. Added [[unlikely]] to a couple of places which will only trigger once in U32::max() group allocations.
• 2024-09-07: v7.4.7 - minor optimisation to get_iterator()/is_active() + bugfix for when they're used on a colony with no blocks. Update to test suite.
• 2024-09-04: v7.4.6 - correction to get_iterator() and is_active() when pointer/iterator do not point to elements currently existing in the colony instance. Thanks to Mishura4 for spotting the bug.
• 2024-07-17: v7.4.4 - Fix to support under GCC5 with -std=c++11, courtesy of Martin Valgur.
• 2024-04-21: v7.4.3 - Minor optimisations to iterator operator -- and reverse_iterator operator ++. Minor code reduction.
• 2024-04-10: v7.4.2 - Change to vcpkg-compatible versioning. plf::make_move_iterator changed to be static. plf::equal_to uses references now and is entirely noexcept.
• 2024-01-03: v7.41 - Macro code reduction.
• 2023-12-26: v7.40 - Minor optimisation for shrink_to_fit()/reshape(). Correction & optimisation to unique(). Optimisation to erase_if()/erase() overloads.
• 2023-12-09: v7.39 - Fix for insert + building with/without exceptions.
• 2023-12-06: v7.38 - Bugfix for reserved blocks not being correctly added back into source in splice(). Replaced some ternary instances with std::min/max for clarity. Correction to erase() assert. Needless comment removal.
• 2023-12-04: v7.37 - sort optimisation with memory, minor optimisation to unique() and std::erase_if overloads, semantic correction to get_allocator.
• 2023-11-30: v7.36 - Fix bug where if splice was used and someone tried inserting into the empty source hive post-splice it could've caused an issue.
• 2023-11-25: v7.35 - Optimisation to reshape(). Removed code in reshape to check for non-copyable/movable status of type as program will not compile with the type in question if reshape may be called (and copying/moving is thereby mandated). Minor optimisation to splice() block checking. Correction to splice() when order of source/destination blocks swapped. Swapped some manual swapping with std::swap where appropriate.
• 2023-11-14: v7.34 - Correction to splice() when destination is empty (no longer transfers allocator or block limits).
• 2023-10-12: v7.33 - shrink_to_fit and reshape now correctly support moveable-but-not-copyable types such as unique_ptr. Correction to compiler support checks for sort.
• 2023-10-10: v7.32 - Remove needless MSVC macro definitions as aligned_storage is no longer used (deprecated in C++23).
• 2023-08-24: v7.31 - Update to sort algorithm, uses less memory for trivial/movable types smaller than sizeof(value_type *) * 2, no performance difference.
• 2023-08-10: v7.30 - Breaking changes: plf::colony_limits changed back to plf::limits, advance/next/prev no longer bound to end()/cend()/rbegin/crbegin(). If you think a use of these functions may go beyond those locations, check using the < or <= operators and do the bound yourself. Iterators do not have access to the end/begin iterators of a colony, so neither do the advance/next/prev functions. The unneeded metadata removal below forces this change but the reduction in memory usage is deemed more important. Largely-unneeded group metadata variable removed, code fixed to suit. Overall performance benefit, particularly for small scalar types. noexcept removed on colony(block_limits) and colony(block_limits, allocator) due to lakos rule. Various code fluff and minor corrections. Fix to is_active() when a group is deallocated from the hive, but then the same pointer address is re-supplied to the hive via the allocator for a new group, and is_active is called with an iterator pointing to an element in the old (deallocated) group.
• 2023-07-05: v7.20 - Bypass buggy MSVC /W4 warning, comment corrections.
• 2023-06-23: v7.19 - Improvement for support on older compilers. Correction to logic and optimisations for unique() and std::erase_if/std::erase overloads.
• 2023-06-13: v7.18 - Less codegen when exceptions are disabled but types aren't nothrow. Corrections to some noexcept qualifiers on fill patterns.
• 2023-06-08: v7.17 - Technical correction to casting for pointers and casting pointers to void *. Minor optimisations to unique() and erase_if overload.
• 2023-05-22: v7.16 - correction to group constructor for c++03/98.
• 2023-04-30: v7.15 - max_block_capacity_per_allocation renamed to the more intuitive max_elements_per_allocation. max_elements_per_allocation now returns 0 if return value is below the colony hard minimum block capacity limit.
• 2023-04-22: v7.13 - Performance improvement for overaligned small types. Addition of max_block_capacity_per_allocation function, to allow people using allocators which can only supply a fixed or low maximum amount of memory for any one allocation (eg. embedded) to correctly assess what their maximum block capacity should be, if they don't want to go over those limits. Correction to block_allocation_amount function. Both of these functions truncate to the hard max block capacity limit of colony ie. numeric_limits<skipfield_type>::max(), so, 255 for types with sizeof < 10, 65535 for others.
• 2023-02-15: v7.11 - Correction of internal namespacing for iterator subtypes to avoid external namespace mis-matches. Correction to test suite under c++03/98. Minor performance improvement to range erase.
• 2023-02-12: v7.10 - Correction to distance overload.
• 2023-02-03: v7.08 - More accurate compiler/library detection for C++20 support, C++20 support disabled for apple clang as it is not as up to date as non-apple clang.
• 2023-01-26: v7.07 - Fix to reverse_iterator(iterator) constructors and '= iterator' operators.
• 2023-01-11: v7.06 - Correction to C++03 support. Correction to colony_data() struct specification. Support for disabling exceptions under GCC, Clang and MSVC. Will (a) now compile under gcc/clang when -fno-exceptions is specified and (b) will terminate in cases where exceptions would normally be thrown.
• 2022-11-30: v7.04 - Workaround redefine macro warning.
• 2022-10-05: v7.03 - Correction to C++03/98 support when using 8-bit types.
• 2022-10-02: v7.02 - Correction to move constructor. Removed std::aligned_storage (this is deprecated in C++23) and corrected logic in aligned storage types. Correction for older compilers which partially support C++11 like MSVC2010.
• 2022-09-03: v7.01 - More tooling sharing between plf:: headers. Some reinterpret_cast's corrected to convert_pointer ie. corrections to data(), range-erase and C++03 group constructor when non-trivial pointer types used. convert_pointer now uses std::bit_cast instead of reinterpret_cast, to allow for potential constexpr use in future.
• 2022-08-27: v7.00 - Bring colony in line with bugfixes, correctness changes and additional features for plf::hive:
  Minor optimisation to sort. Explicit-ness constructor corrections. Change 'block_limits()' to 'block_capacity_limits()'. Addition of 'block_capacity_hard_limits()'. Added std::type_identity_t to allocator copy constructor overloads. Correction to next and prev ADL overloads. Improvement to erase_if algorithm. Addition of unique(). Removal of range-based sentinel overloads for constructor, insert, assign (these are covered by the C++20 ranges v3 overloads). trim() changed to trim_capacity (trim has a specific meaning in some parts of the standard library). Addition of trim_capacity(n). skipfield_type switching changed to be partially based on element_type sizeof. Addition of is_active(const_iterator). Corrections to distance, advance, reshape, trim_capacity, unique, block_capacity_hard_limits, splice. Corrections to erase to allow get_iterator to work correctly. Corrections to exception recovery in range assign, range insert, range constructor. Splice now retains original reserved memory blocks in both source and destination post-operation. Const_iterator and const_reverse_iterator can no longer be converted into iterator and reverse_iterator respectively (under C++11 and above - C++03 and below do not support this functionality). Return type of const_reverse_iterator::base() is now a const_iterator instead of an iterator. rbegin/rend/crbegin/crend no longer segfault when calling .base() on an empty hive. Changes to allocator rebinds to allow use with pmr allocators. Default maximum capacity changed to 8192 as per benchmark results. default_min_block_capacity, default_max_block_capacity and default_block_capacity_limits functions added. Correction to splice to allow get_iterator and is_active to work. Corrections to get_iterator in situations where the iterator/pointer points to a block which has been removed then added back into the iterative sequence (via either unused_groups or an allocator), but now points to an element after end(). Optimizations to splice, get_iterator. Conformance changes for iterator/reverse_iterator. Compatibility with std::reverse_iterator added (for use with rangesv3). Correction/optimisation to range erase. Change to group numbers strategy - never update unless overflow possible- which is a 1 in numeric_limits<size_type>::max() event. This removes a O(n) factor in erase() when there is a large number of groups. Singly-linked groups-with-erasures intrusive list changed to doubly-linked to avoid O(n) performance factors with large numbers of small blocks. Corrections to allocator rebinds. Correction to pointer conversions for non-trivial pointer types. Iterators moved to bottom of file to make human parsing easier. Added plf::ranges::from_range_t for use in rangesv3 constructors until such point as std::ranges::from_range_t becomes more widely available in libraries. Correction to free list writes to make compliant with C++ standards rules (though will have no noticable effect for bulk of situations). Correction to exception safety for reshape(). Fix for situations with 128-bit platform when using char as element_type. Minor optimization to group allocations under some compilers. Minor correction to fill-insert functions/constructors. range erase now returns iterator in case of situation where end() is supplied as last in a loop and return value is necessary to make sure iterator == end() end condition is reached. Comment updating. Some tooling functions made public for access across multiple plf:: libraries. Other minor code reductions. Correction to copy constructor.
• 2021-12-11: v6.33 - Added spaceship operator for three-way comparison of colonies. block_metadata_memory() added as a convenience function, for now.
• 2021-10-31: v6.32 - Range constructors/inserts/assigns updated so that the slower form that is required to match sentinels, will not match iterator types which can be converted to one another eg. const_iterator and iterator. This means that, for example, a call to range-assign using colony's const_iterator and iterator will be able to use colony's faster ADL distance() overload rather than just doing a straight for-loop to determine length of range (necessary with sentinels, as there is no overload for distance() for sentinels).
• 2021-10-26: v6.31 - Comment updating.
• 2021-09-18: v6.30 - Correction to range constructor/insert/assign with differing iterator types. Correction to compiler feature detections.
• 2021-09-08: v6.29 - Removal of hive typedef (hive now has it's own separate codebase at https://github.com/mattreecebentley/plf_hive. Corrections to test suite.
• 2021-06-28: v6.28 - Clear() behaviour changed - now retains all memory blocks in reserved pool, rather than deallocating them. reset() introduced, which replicates old clear() functionality (destroy elements + deallocate blocks). Alternatively to deallocate blocks you can follow clear() with shrink_to_fit() or trim(). Bug in -> operator for reverse_iterator fixed.
• 2021-06-21: v6.26 - Added support for differing iterator types (eg. using sentinels or const_iterator and iterator) in range constructor.
• 2021-06-02: v6.25 - Comment updating.
• 2021-05-26: v6.24 - Added sentinel support for assign. Correction to get_iterator under C++98/03.
• 2021-05-25: v6.23 - Correction to compiler detection routines for clang/gcc running in MSVC environment.
• 2021-05-20: v6.22 - Correction to compiler detection routines for clang - only libc++ 13 and above support the concept library.
• 2021-05-09: v6.21 - Minor code reduction, corrections to splice (now throws if source memory blocks are outside destination's memory block capacity limits).
• 2021-04-30: v6.20 - Changed optimization strategy for 3/4-way decision patterns in erase/insert/emplace. Previous strategy worked better on earlier compilers, however seems to be better with straight if statements for modern clang/gcc. Correction to move-insert.
• 2021-04-02: v6.17 - get_iterator_from_pointer changed to get_iterator - the pointer part is implied by the fact that it's the only argument. Added get_const_iterator - both take a const_pointer, the latter returns a const_iterator. Correction to rbegin() and rend() const status.
• 2021-03-22: v6.16 - Update to container based on LEWG feedback. plf::limits becomes plf::colony_limits. Correction to MSVC compiler CPP20 support.
  Instead of taking an unsigned integer type as a parameter, the colony template now takes a (basically boolean) plf::priority enum parameter, allowing one to choose between plf::performance (equivalent to unsigned short - default - in previous versions of colony) and plf::memory_use (equivalent to unsigned char in previous versions of colony). There was never any point to choosing a higher bit-depth unsigned integer type anyway (see Group sizes in FAQ above).
  Publicly-accessible skipfield_type parameter removed.
  Correction to noexcept statuses on some functions when variadics unsupported by compiler.
  Removal of needless function call in constructors when alignment is supported by compiler.
  Addition of range insert when using different types of iterators which are equality_comparable, under CPP20. ie. support sentinels and comparisons between const_iterators and regular iterators.
  plf::hive, plf::hive_limits and plf::hive_priority are alternate typedefs for plf::colony, plf::colony_limits and plf::colony_priority respectively (anticipating a possible name change for the std:: version of colony). These are only available in C++11 and above, as C++03 and below do not support typedefing of templates.
  data() function reintroduced, now provides bitfields for SIMD usage.
• 2021-01-13: v6.11 - Reduction of compiler feature macros for reasons of readability, code size and convenience of updating between the different containers. Hopefully, PLF_ by itself will be enough to disambiguate these from similarly-named macros. If not, will revert back. Improvement of a few compiler rules. Correction to shrink_to_fit/reshape. Consolidation of catch recovery code for range_fill/fill and range_fill_skipblock/fill_skipblock.
• 2021-01-11: v6.10 - Corrections to compiler feature detection.
• 2021-01-09: v6.09 - Correction to group allocations under C++98/03. Removal of unneeded static_casts.
• 2021-01-08: v6.08 - Corrections to distance() under C++98/03 (undefined behaviour).
• 2021-01-07: v6.07 - Corrections to noexcept specifiers, courtesy of cppcheck.
• 2021-01-06: v6.06 - Increase of 6%-10% speed under GCC 9/10 in high-modification general-use benchmarks (mitigation of gcc optimisation bug). Correction to exception handling in reserve. Removal of code redundancies & code reduction. Under GCC 8, 1.5% faster insertion and 2.5% faster erasure in small struct benchmarks.
• 2021-01-03: v6.05 - Better support for type alignment courtesy of Tomas Maly (https://github.com/malytomas). Resultant performance improvement of 2% in high-modification benchmarks. plf::pcg_rand renamed to plf::rand, modified to enable better support for C++03 and lower compilers. Correction to support under c++03/98.
• 2020-12-26: v6.04 - Corrections to compiler feature detection code. Test suite now uses plf::pcg_rand instead of xor_rand. Minor correction to overalignment code (more to come in future).
• 2020-11-29: v6.03 - Optimization to destroy_all_data (destructor/clear/move).
• 2020-11-28: v6.02 - Default value fill constructor added. Reserve now throws instead of auto-correcting capacity value if it is greater than max_size().
• 2020-11-06: v6.01 - Correction to noexcept guarantees for splice. Addition of std::erase() and std::erase_if() non-member function overloads for colony container. Range-erase now returns iterator to the element after the erased range, or if the end of the erased range is end(), returns end(). Typedef and macro tidy up. Correction to overaligned element support. Removal of some unneeded asserts.
• 2020-10-26: v6.00 - Overall re-write to be more in line with proposed STL addition. Correction to edge case optimization in distance() + additional minor optimizations. Fill-constructor changed to be explicit.
  Change of crbegin(), rbegin(), crend() and rend() to be noexcept (adds minor performance cost to rbegin() and crbegin()). Additionally, now (rbegin() == rend() && crbegin() == crend()) when container is empty - important for loops.
  Distance, next/prev and advance are now non-member functions, MSVC2013 support is removed (MSVC2013 has an unfixed bug to do with templates which prevented making these functions non-members originally - previous and subsequent versions of MSVC versions do not have the bug). get_index_from_iterator is removed, please use distance(colony.begin(), iterator) instead. get_iterator_from_index is removed, please use "iterator = begin(); colony.advance(iterator, index);" instead. get_iterator_from_pointer is now noexcept.
  reserve() is now a fully-functioning ie. multiple-memory-block-reserving reserve function, and insert will take advantage of this. Erase will now reserve some but not all memory blocks which have all elements erased. Trim() function introduced, which removes unused reserved memory blocks without causing reallocation.
  assign() - both fill, range and initializer-list versions - are added.
  Performance improvements in multiple areas, particularly fill-insert, range/initializer-list insert, fill/range constructors and operator =.
  Comment corrections. Some asserts changed to static asserts.
  set_block_capacity_limits() renamed to reshape(). get_block_capacity_limits() renamed to block_limits(). set_minimum_block_capacity() and set_maximum_block_capacity() removed, use a combination of block_limits() and reshape() instead.
• 2020-10-25: v5.43 - Removal of problematic asserts in operators -- and ++. Change of name from approximate_memory_usage() to memory().
• 2020-09-25: v5.42 - Correction to technically undefined behaviour under C++98/03.
• 2020-09-21: v5.41 - Correction to splice.
• 2020-09-19: v5.40 - Correction to non-member swap, should've been in namespace std.
• 2020-09-17: v5.39 - Removal of destructor calls within sort() since all element locations inevitably get written to in the container/array so any necessary destructor calls for non-trivial objects will be handled by the move/copy operators of that element type.
• 2020-09-16: v5.38 - Removal of unnecessary branch checks in sort(), thanks to sentinel reduction, courtesy of github user morwenn.
• 2020-08-16: v5.37 - Re-included sort(), using the indiesort variant (higher performance for larger types and significantly-reduced memory usage). Uses std::sort internally but will use another sort function if it's name (including namespace) is defined as the macro PLF_COLONY_SORT_FUNCTION before including the colony .h. Supplied sort function must take same parameters as std::sort.
• 2020-08-15: v5.36 - Correction for edge case within range erase. Correction to range insert in compilers without type traits.
• 2020-07-07: v5.35 - Range-insert/constructor optimization now available in C++11-and-above modes.
• 2020-06-27: v5.34 - Fixes to distance, set_block_capacity_limits. Reduction to compiler switches.
• 2020-05-03: v5.33 - Minor optimisation to operator ==. Some internal iterator usage changed to const_iterator.
• 2020-04-26: v5.32 - Correction to <=> iterator operator, some static_cast's to avoid additional MSVC warnings. Fix for minor regression in prev version.
• 2020-04-24: v5.31 - Shifted some code out of the insert functions into a separate function, which improves performance for mixed-code scenarios like the test suite, while providing no performance detriment in the higher complexity tests. Fix to small problem with conversion warnings in fill-insert.
• 2020-04-16: v5.29 - Correction to operator = && when pointers as supplied by an allocator are non-trivial. Update some comments. Optimization for fill-insert with non-trivial types.
• 2020-04-13: v5.28 - Fixed an issue with prior range insert optimisation which would've caused reallocation of elements and iterator invalidation under some scenarios. Minor optimisation to reinitialize() based on most common use-case.
• 2020-04-10: v5.27 - Corrections to test suite. Added [[nodiscard]] to empty().
• 2020-04-08: v5.26 - Added default-value fill constructor (eg. 'colony<int>(14)' to create an int colony with 14 default-constructed ints). Changed names of change_group_sizes, get_group_sizes, change_minimum_group_size and change_maximum_group_size to: set_block_capacity_limits, get_block_capacity_limits, set_minimum_block_capacity and set_maximum_block_capacity. This is to bring the implementation more in line with the C++ standards proposal and to make the naming more independent of implementation, and more correct. Removed 'const_iterator get_iterator_from_pointer(const const_pointer)', as users can manually cast the pointers themselves to 'iterator get_iterator_from_pointer(const pointer)', and the return type will automatically cast from iterator to const_iterator if a const_iterator is used to receive the return value. This also brings the implementation more in line with current C++ standards. Added get_reverse_iterator_from_pointer() and get_reverse_iterator_from_index(). Corrected missing macro undef in previous version. Changed get_raw_memory_block_pointers() to data(), to be more in line with std library.
• 2020-03-22: v5.25 - Added colony operator = for std::initializer_list. Updated noexcept conditions for swap and move operations. Added const_pointer overload for get_iterator_from_pointer(). Added code to automatically reserve necessary capacity in range/initializer_list insert/construct/operator =, if the range in question uses random_access iterators. Added spaceship operator (why) for compilers which support it.
• 2019-11-27: v5.24 - Removed unneeded assert for iterator operator = when assigning between const and non-const iterators. Added operator = and copy constructor for assigning/constructing between const and non-const reverse iterators. Increased ability for reverse_iterators (const and non-const) to take assignment from regular const/non-const iterators.
• 2019-11-26: v5.23 - Correction to iterator operator = when assigning between const and non-const iterators. Correction to begin() and end() to support pre-C++11 semantics ie. can return const_iterator as an overload. Which in turn allows colony to function with eric neibler's ranges implementation.
• 2019-11-18: v5.22 - Correction to edge-case in advance(), slight speed increases to iterator operator ++, insert, emplace. Changes to remove warnings in test suite. Overall 1% performance improvement in general use benchmarking, 4-5% performance improvement in general use benchmarking when dealing with low numbers of elements and low amounts of modification. Improvements partially based on advice about pointer aliasing from Elliot Goodrich.
• 2019-08-13: v5.21 - Fix for MSVC2019 in permissive mode incorrectly reporting a warning of use of std::allocator instead of std::allocator_traits
• 2019-08-01: v5.20 - Edited skipfield pattern to enable parallel conversion to SIMD-gather mask (erased skipfield nodes in middle of skipblocks will always be non-zero now). Added function get_raw_memory_block_pointers() to enable users to supply elements directly to SIMD-gather/scatter operations using the skipfield as a mask (skipfield will require conversion to specific SIMD instruction set format, which can also be accomplished in SIMD, in parallel). Tests added to test suite for both non-trivial types, and the get_raw_memory_block_pointers() function. Corrections to fill-insert for aligned types. Minor correction to avoid type safety warning in GCC under C++03 and using "-Werror=sign-conversion".
• 2019-06-17: v5.11 - Explicitly specified some implicit type promotion conversions in order to deal with errors when using GCC with "-Werror=conversion". Added non-copyable type test to test suite.
• 2019-06-11: v5.10 - Correction to reserve() semantics. Reserve() also will no longer assert that reserve amount is within bounds of maximum and minimum group sizes, and will silently round the value down or up accordingly. Please consult the documentation to understand the current limitations of reserve() for colony.
• 2019-04-15: v5.09 - Constexpr added to type traits 'if' blocks and some other scenarios, for compilers which support it - including MSVC2017 with std:c++17 switch. This fixes some issues for movable-but-not-copyable elements, and creates performance benefits in some scenarios. May also speed up compile times slightly.
• 2019-04-12: v5.08 - shrink_to_fit(), reserve(), change_group_sizes(), change_minimum_group_size() and change_maximum_group_size() are now valid operations for element types which are move-constructible but not copy-constructible, when type traits are supported by compiler. These operations will also be faster for non-trivially-copyable-but-trivially-movable types, if reallocation of existing elements is triggered.
• 2019-03-19: v5.07 - Removed unnecessary const casts (to avoid GCC warning) and added newline at end of file (to avoid clang warning).
• 2019-02-28: v5.06 - Added static_cast's to work around GCC's -Werror=sign-conversion warnings.
• 2019-01-30: v5.05 - Bizarre minor redundancy removal. Add post-move assignment test to test suite.
• 2019-01-16: v5.04 - Correction to library detection features to include type traits for libc++. Removed support for plf::timsort, added support for gfx::timsort (differences between plf and gfx version were not substantive enough to justify maintaining a fork of the original code).
• 2018-11-16: v5.03 - Correction to max_size() for use with allocator_traits and to be in line with C++20.
• 2018-10-28: v5.02 - Optimization to move assignment - this also improves regular assignment, sort, splice, swap, change_group_sizes, shrink_to_fit & reserve.
• 2018-10-08: v5.01 - Comment corrections, blank() no longer differentiates based on compiler (re-testing revealed this is best strategy across compilers). Correction: included header for offsetof.
• 2018-10-02: v5.00 - Change of skipfield type from high-complexity jump-counting skipfield (previously known as 'advanced jump-counting') to low complexity skipfield (previously known under working title of 'Bentley pattern'). Change of erased-element-memory-location storage mechanism from free-list of erased element locations, to free-list of blocks of consecutive erased element locations. Resultant procedural change means that all singular insert/emplace and erase operations now have O(1) time complexity, while multi-insert is O(n) and multi-erase is O(n) for non-trivially-destructible elements, ~O(log n) for trivially-destructible elements (see time complexity in the faq). In addition iteration will generally be faster in situations where elements are being inserted and erased a lot, as skipblocks can no longer get split up into multiple skipblocks by insert, mean there are less overall 'jumps' during subsequent iteration. Also since insertion and erasure can no longer result an an unpredictable skipfield update, the container is much more suitable for extremely-low-latency environments. <, >, <= & >= iterator operators and operations which use them e.g. range-erase in debug mode, no longer throw exceptions if supplied with begin() and end() iterators of an empty colony. Other corrections to range-erase. Optimization to fill-insert. More extensive testing for range-erase and fill-insert added to test suite. Platform-specific code removed for iterator operator ++ (default now best across all platforms). Range-erase and fill-insert 10% faster overall in tests. General-use-case testing up to 10% faster on core2 (4.5% faster on average).
• 2018-07-12: v4.62 - Optimization to swap for when the allocator supplies non-trivial-but-movable pointers.
• 2018-07-07: v4.61 - Optimization to swap. Updating of some CPU-specific tests. Comment correction/updates.
• 2018-07-06: v4.60 - Multiple corrections to nothrow_constructible checks under insert/emplace. Addition of second emplace test to test suite. Added void * casts to deal with GCC 8.1's garbage warning messages around memset. Update/addition of comments. Minor update to benchmark suite.
• 2018-06-27: v4.59 - Change of sort dereferencer function to allow for non-const sort functions (with internal state for example).
• 2018-06-19: v4.58 - Corrections to test suite compiler feature detection macros, change reverse_iterator constructor type to non-explicit, minor optimization to insert for non-trivially-constructible types, minor code-comment/code cleanup.
• 2018-05-09: v4.57 - Mitigated NULL == 0 warning for clang, fixed compiler feature definitions for clang + libstdc++, thanks to bluebear94.
• 2018-05-02: v4.56 - Same optimization for GCC & clear/splice/etc now applied to other compilers generically.
• 2018-04-28: v4.55 - Postcondition of source colony for move-assignment/construction is the same as an empty colony now (i.e. you can use it normally post-move without undefined behaviours). Minor optimisation to move-assignment/construction and several other functions (splice, clear, range-erase) as a result under C++11 and above (when allocator does not supply smart pointers). Minor optimisation to swap; the original C++03 code was faster than the C++11 move-swap code under C++11. Minor optimisations to distance, advance. Code comment cleanups. It is now valid behaviour to call range-erase with two equal iterators i.e. an empty range.
• 2018-04-13: v4.51 - Mistake in NOEXCEPT_MOVE_ASSIGNMENT compiler feature macro, fixed.
• 2018-04-02: v4.50 - Stack-based erasure-recording system replaced with per-memory-block index-based free list + erasures-groups-list structure. Resultant speed up of 10% in high-modification general-usage benchmarks, 17% reduction in code size, and reduced memory usage. This also makes erase exception-safe (no allocations due to no stack) unless destructor of object throws exception (i.e. fails to close a file lock, or whatever) or an invalid/uninitialized iterator is supplied to function. Reversion of insert from function call to emplace, back to a separate function - as this was causing copy-construction on non-trivial types, leading to additional temporaries and severe performance drop on large types. General code cleanup and renaming. Corrections to compiler feature support detection. Corrections to some noexcept function statuses. Minor optimisations for C++03/98 and in general. Minor optimisation to operator = under c++11 and upwards. Support for over-aligned types.
• 2018-01-05: v4.14 - Minor optimization to insert.
• 2017-11-15: v4.13 - One noexcept change, fill-insert correction/optimization, minor optimizations and code comment cleanup.
• 2017-11-1: v4.12 - Correction to no-exception compile-time checking for insert/etc. get_iterator_from_index() now templated, supports multiple index types. This also reduces codegen when both this function and advance() are unused. Change to some inline hinting.
• 2017-10-11: v4.11 - Codegen reduction for insert, move-insert for compilers with variadic parameter support.
• 2017-10-03: v4.10 - Rollback of a previous insert optimization due to potential for overflow with extremely large numbers of elements and extremely large numbers of erasures. Improvement to range-erase performance.
• 2017-09-26: v4.09 - Large single-insertion performance improvement for small scalar types. Other minor optimizations to splice, range-erase and advance. False-negative removal in test suite for 32-bit compilers.
• 2017-09-16: v4.07 - Improved fill-insert performance for trivially_copy_constructible types under C++11/C++14/etc. Avoid generating exception try-catch code for nothrow-constructible types. Correction to reserve() - people were probably calling it with size_type's, causing it to truncate/overflow. Minor correction to type traits use in sort(). Minor optimization to operator --.
• 2017-09-11: v4.06 - Corrections to fill-insert. Optimization for reserve() followed by fill-insert. Removal of return values from fill-insert, range-insert and initializer-list insert (no point, unordered insertion, always).
• 2017-08-31: v4.05 - void splice(colony &source) added to colony. This means you can move all elements from one colony into another colony of the same type without invalidating pointers or iterators to the source elements. Splicing is also much faster than range-moving the elements. Splice does not guarantee appending to the destination colony - the insert location of source groups is chosen based on the most positive overall performance outcome for subsequent iteration and insertion.
• 2017-08-26: v4.01 - Same optimization as v4 applied for skylake/kabylake CPUs under GCC and -march=native. Unlike haswell, this is associated with a small decrease (~1%) to unordered use-case performance, but the same 15-26% performance improvement for iteration tests. Further feedback from external testing welcome. Redundant iterator parameter removed (allocator). C++14-specific const/non iterator comparisons (e.g. if (reverse_iterator != const_reverse_iterator)) now available for all compilers and C++ versions. Correction to benchmark suite.
• 2017-07-26: v4.00 - Up to 7% iteration performance improvement on Core2 processors and/or under MSVC, 15%-26% iteration performance improvement on pre-"kaby lake", post-"Westmere" Core i processors (haswell, broadwell etc) processors under GCC (with -march=native enabled). Amount of improvement depends on number of elements. Overall 2% improvement in modification use-case testing on all platforms. Correction to fill-constructor and initialization-list constructor. Thanks to Phil Williams and Jason Turner for testing for me on alt processors!
• 2017-06-21: v3.95 - Corrected support for GCC v5.x. Correction to test suite under GCC versions > v5. Correction to distance() for reverse iterators.
• 2017-06-01: v3.94 - Minor optimizations and test suite cleanup.
• 2017-05-21: v3.93 - Added sort(). Colonies can now be sorted as you would with a list, with or without a supplied comparison function (less-than if without). Removed MSVC warnings at level 4, ignoring those which're obviously compiler bugs. More support for non-gcc compilers using gcc libraries or libc++.
• 2017-05-11: v3.91 - Additional type safety check on stack consolidate. Move assignment now follows C++17 noexcept rule. Correction to missing undef's at end of header.
• 2017-05-11: v3.90 - Minor optimizations to initializer-list insert and copy constructors. Emplace availability correction for some compilers. Corrections to shrink_to_fit() and reserve() (will not longer go below minimum group size). Minor optimizations to operator = and c++03 swap. Other minor code improvements.
• 2017-05-06: v3.89 - Minor correction to noexcept statuses. Correction to test suite under C++03.
• 2017-03-28: v3.88 - Correction to emplace. Corrections to default constructor to allow for empty initialization lists. Some standards compliance updates and corrections to noexcept statuses. skipfield_type is now a publicly-accessable member typedef. get_group_sizes() is now available. Allocator-extended copy and move constructors added. get_allocator() added. swap(colony &A, colony &B) is now non-member and non-friend.
• 2017-02-08: Update and corrections to benchmark code and test results.
• 2016-12-10: v3.87 - Optimized and corrected range erase. Documentation updated.
• 2016-11-19: v3.86 - Performance improvements for low numbers of scalar types. Minor optimizations for fill-constructor and initializer-list constructor. Corrections to advance.
• 2016-11-12: v3.83 - Minor optimizations. Edge-case corrections for advance functions and copy constructor. Optimizations to get_index_from_iterator, get_iterator_from_pointer, distance, advance. Advance now correctly bounds to rbegin and rend for reverse iterators. Warning removal for test suite.
• 2016-11-1: v3.81 - Optimizations to insert functions. Corrections to fill-insert/fill-construct functions. Minor correction to internal stack copy constructor. Some function and comment tidyup. Moved referencer test code out from plf_bench.h to individual test files due to MSVC's forward declaration bug (referencer tests not possible with MSVC for this reason). Possible edge-case out-of-bounds memory access for insert removed. Benchmarks updated.
• 2016-10-27: v3.80 - Further optimizations. Removed a potential memory leak which would only occur upon exception-based state rollback within insert functions. Update to benchmark suite, addition of packed_deque container and associated tests.
• 2016-10-15: v3.76 - Removed potential memory leak in internal stack mechanism.
• 2016-10-14: v3.75 - Added get_index_from_reverse_iterator(). Further optimization (and one minor correction) of the reduced_stack storage mechanism , leading to an 8% performance increase in general_use benchmarks for small structs. Additional minor optimization for C++03. Some code comment tidyup.
• 2016-10-10: v3.73 - change_group_size functions no longer copy-construct into a new colony unless the stack is not empty, and either min_group_size is larger than the smallest group in the stack, or max_group_size is smaller than the largest group in the stack. Correction to change_group_sizes code (was clearing erased locations unnecessarily). Reinitialize function added, which supports the changing of group sizes for colonies of non-copy-constructible element types. Emplace changed to no longer require move semantics, which enables the use of non-move-constructible/non-copy-constructible types with colony (example: anything containing a std::mutex). Minor correction to colony test suite under C++03.
• 2016-10-04: v3.72 - minor fix to erase.
• 2016-10-03: v3.71 - Replaced all but one vectorizing skipfield update with memmove ops for greater performance. Colony is no longer dependent on plf::stack, now has it's own reduced-functionality internal structure. TLDR, slightly faster, slightly faster build time & slightly less memory use. Colony and stack are no longer friends :( (aw). Updated documentation. Stack separated out of colony package and test suite and vice-versa. Added functions to convert element pointer to iterator, and to convert iterator to index and vice-versa. Removed pointer-based erase function (use get_iterator_from_pointer() instead and call erase() with the resultant iterator). Minor update to distance, advance, prev and next functions. Benchmark suite updated.
• 2016-09-02: v3.61 - Slight performance improvement for general-use-cases for colony, small performance/usability improvements in benchmark suite. Tweak for MSVC2010 performance. Correction to stack emplace.
• 2016-08-24: v3.6 - 8% speedup on iteration, vs v3.55, for colony with larger-than-scalar types. Performance increases and minor corrections in benchmark suite.
• 2016-08-20: v3.55 - Updates to benchmark suite - now uses a faster xorshift-type rand generator and a fast_mod function, test differences between containers more obvious now. Memory approximation also now outside of main loops, test cases updated for general_use test. Entirely new test-case function. Minor performance enhancements to insert/erase/operator++ for colony + code tidy-up. Minor warning removal for test suite.
• 2016-08-04: v3.51 - Correction to move-insert courtesy of Peet Nick. Removal of end-block edge-case scenario code, as it was found to be of such low statistical significance as to be useless. The fix increased memory use in the scenario of the edge case which might be it's own problem in most scenarios. To get around the edge case, if it occurs in a given context, an allocator approach would be most appropriate. Update to benchmark suite.
• 2016-07-19: v3.5 - Colony now accepts a template parameter for the skipfield type. This enables the use of, for example, unsigned char as the skipfield type. The bitdepth of the unsigned integer also defines the default maximum size of colony groups. The benchmark suite has been updated with memory measurements for the general_use tests. General_use tests changed to percentage-based tests. Internal erased location stack now stores only element pointers, not full iterators. Storing just the pointer and having to reconstruct the iterator slows down reinsertion after an erase, but speeds up the erase itself and results in an overall performance improvement, as well as less memory consumption. Slight improvement for fill-insert. Sanity-checks for range-insert and initializer-list-insert implemented. Minor fix to C++03 support in benchmark suite.
• 2016-07-4: v3.35 - Another compliance fix for benchmark suite in clang, courtesy of Arthur O'Dwyer.
• 2016-07-3: v3.34 - Minor warning removal for test suite. Performance improvement for colony edge-case scenario where a single element is inserted and then erased repeatedly, and where the insertion point happens to be at the end of a memory block.
• 2016-07-1: v3.33 - Empty destructors/constructors removed in iterators. Small compliancy fixes in both colony and benchmark suite. Thanks to Arthur O'Dwyer for this update.
• 2016-06-25: v3.31 - Colony iterators no longer inherit from std::iterator and implement the required typedefs themselves. Code tidyup and minor changes. Default of 1 implemented for prev/next.
• 2016-06-17: v3.30 - Correction to colony range constructor. Corrections to advance() asserts.
• 2016-06-15: v3.29 - Support for stateful allocators added, to both colony and stack. More generic solution applied for preventing calls for fill-insert to be matched by range insert template.
• 2016-06-14: v3.25 - Edge-case corrections to colony's clear(), range-insert and fill-insert functions. Minor performance improvement in erase edge-case. Removed older (colony-specific) range insert. Simplified constructors.
• 2016-06-12: v3.22 - Range, fill and initializer-list based insert()'s and constructors now implemented in colony. Bugfix for Microsoft compilers. Warnings fix for clang. Reinitialize functions removed, replaced with change_group_sizes(), change_minimum_group_size() and change_maximum_group_size().
• 2016-06-11: v3.19 - Minor bugfix to plf::stack.
• 2016-06-9: v3.18 - Corrections to colony and stack C++11 typedefs. Some code tidyup. Minor code corrections. Simplified reserve() and shrink_to_fit() functions. Added trim_trailing_groups() function to stack. Minor performance improvements.
• 2016-05-31: v3.16 - Corrections to colony and stack reserve(). Some code tidyup. Correction to test suite.
• 2016-05-30: v3.15 - Colony and stack are now compliant with C++11 and C++14 container concepts. Vectorized some erasure and insertion operations resulting in better overall performance in some scenarios. Corrected edge case in reverse iteration operator ++. All iterators now bounded by asserts in debug mode. Correction to capacity functions.
• 2016-05-26: v3.14 - Added swap (std::swap worked fine without it, mainly required for standards compliance, also is exception-safe) and max_size() functions to stack and colony.
• 2016-05-24: v3.13 - erase(element_pointer) added to colony. This enables erasure via a colony element's pointer, rather than it's iterator. Slightly slower than iterator-based erasure as a lookup has to be done for the colony element's group. Update to test suite.
• 2016-05-19: v3.11 - Fix for colony shrink_to_fit(), added special case to stack copy constructor for when source size > max_group_size, colony and stack constructors now explicit. Update to test suite.
• 2016-05-12: v3.10 - Both stack and colony no longer preallocate upon construction, only upon first insertion. Minor assert bugfix in colony. shrink_to_fit(), reserve() and capacity() functions added to both classes. Some code cleanup, corrections for C++03 and minor optimizations.
• 2016-04-28: v3.05 - Removal of compiler-specific iteration code, resulting in 8% speedup for larger-than-scalar type iteration, 6% slowdown for scalar-type iteration, under GCC x64. MSVC unaffected by change. Addition of benchmark suite. Correction to plf_nanotimer under linux/bsd.
• 2016-04-16: v3.04 - Corrections to reinitialize asserts. Added bound-checking asserts to advance/next/prev implementations.
• 2016-03-26: v3.03 - Correction to operator = on colony. Replaced SDL2/SDL_Timer in benchmarks with plf_nanotimer, a cross-platform ~nanosecond-precision timer. Correction to end() and begin() overloads - internal begin/end iterator could previously be altered by user activity, in some cases. Correction to .next functions.
• 2016-03-24: v3.02 - Small performance improvements. Fix for demo for regular SDL2 usage. Duplication of macros from plf_stack.h to plf_colony.h to fix issue when plf_stack.h is also used within project separately from plf_colony.h.
• 2016-02-23: v3.01 - Colony now uses a jump-counting skipfield instead of a boolean field, which mitigates worst-case scenario performance by factors of ten. Iterators now bidirectional (but include > < >= and <= operators) instead of random-access, and compliant with C++-specification's time-complexity requirements (all operations O(1) amortised). +=, -=, - and + iterator operators relegated to advance, distance, next and prev functions under colony. There is greater compliance with allocators. Group size now limited to 65535 max for colony, to decrease memory wastage and improve skipfield performance. Various bugfixes and corrections. plf::stack storage limit no longer 32-bit uint max (relies on allocator to supply correct size_type). Added worst-case and best-case scenario tests to colony benchmark, and added std::stack comparison to plf::stack benchmark. Benchmarks yet to be updated in docs.
• 2015-11-20: v2.34 - Correction to clear() and reinitialize() functions.
• 2015-11-17: v2.33 - Minor bugfix for GCC versions < 5.
• 2015-11-2: v2.32 - Minor bugfix for non-trivial pointers and non-trivial stack contents.
• 2015-10-31: v2.31 - Allocators which supply non-trivial pointers are now supported. Large macros removed. Some code cleanup and minor performance gain.
• 2015-10-10: v2.26 - 'back()' in plf::stack changed to 'top()' to better reflect std:: library stack implementation function titles. Added standard container typedefs to colony and stack. Colony iterators now return element_allocator::pointer and element_allocator::reference instead of value_type * and value_type &. Some minor code cleanup.
• 2015-10-6: v2.24 - Bugfix for C++03 with x64 compilers. Minor corrections, performance improvements for C++03 compilers.
• 2015-10-4: v2.21 - Improved gcc x64 iteration. std::find now working with reverse_iterator. reverse_iterator function corrections, performance improvements. Hintless allocators under C++11 now supported. Allocation now uses allocator_traits under C++11.
• 2015-9-30: v2.13 - Fixed performance regression with small primitive types and MS VCC. Small bugfixes.
• 2015-9-28: v2.12 - Added const_iterator, const_reverse_iterator, .cbegin, .cend, .crbegin, .crend. Some small bugfixes.
• 2015-9-27: v2.00 - Larger performance improvements, particularly for small primitive type storage. Corrections to exception handling and EBCO. ".insert" replaces ".add" (realised there is precedence for non-position-based insert via unordered_map/unordered_set). Added postfix iterator increment and decrement. Zero-argument ".add()" and ".push()" function overloads removed (same thing achievable with "colony.insert(value_type());" and "stack.push(value_type());". Many bugfixes and general code cleanup.
• 2015-9-15: v1.79 - Empty base class allocator optimisations and smaller optimisation.
• 2015-9-1: v1.77 - Some small bugfixes.
• 2015-8-30: v1.76 - Now exception-safe. Also fixed C++03 regression bug.
• 2015-8-22: v1.75 - Iterators more compliant and now work with std::find and should work with other std:: algorithm functions. Native find functions removed. (iterator - iterator) and (iterator + iterator) support added (relatively useless to end-users, but required by some std:: algorithms).
• 2015-8-15: v1.73 - Correction to noexcept/nothrow use under some non-c++11 compilers. Some rvalue functions corrected.
• 2015-8-3: v1.72 - Added noexcept to more functions.
• 2015-7-20: v1.71 - Small correction to demo, changed a data member name in demo to remove conflict in MSVC 2015.
• 2015-7-13: v1.70 - Added deque comparisons to performance tests - this was necessary because many people were making comparisons between colonies and deques, without understanding what a deque is and how it performs...
• 2015-6-21: v1.69 - Code tidy-up.
• 2015-6-16: v1.67 - Correction to VS2013/VS2010 support: move semantics and other C++11 facets enabled for MS compilers. Support for VS2012, VS2015 and later added. Performance bottleneck in destructors in debug mode fixed. General cleanup. Some microoptimisations. Macros now better named to avoid conflicts.
• 2015-5-30: v1.60 - Correction to add function, microoptimizations.
• 2015-5-29: v1.53 - Performance improvement. Corrected move constructor and reverse_iterator function semantics.
• 2015-5-28: v1.52 - Small correction.
• 2015-5-26: v1.51 - Small performance improvement. In demo, including SDL.h caused some weird behaviour with timer when used, changed to SDL_timer.h. Benchmarks to be updated.
• 2015-5-25: v1.50 - Microoptimizations, small improvements, fixed crash in demo where memory would overflow when compiled to x86 without LAA.
• 2015-5-21: v1.43 - Minor performance improvements for release mode, larger improvements for debug mode.
• 2015-5-16: v1.42 - Some demo corrections.
• 2015-5-16: v1.41 - Fixed demo on pre-C++11 compilers. Minor fixes in code and corrections in tests. Performance improvements.
• 2015-5-11: v1.40 - Improved iterator ++ speed and Add speed, added real-world OO performance tests to demo. Added max group-size tuning (max number of elements per group).
• 2015-5-6: v1.34 - Improved iterator ++ speed under VC2010 and GCC x64.
• 2015-5-5: v1.33 - Improved erase, and some small performance tweaks.
• 2015-5-4: v1.32 - Minor fixes.
• 2015-5-3: v1.31 - Resolved all current known edge-case crashes. Improved stack pop performance.
• 2015-4-26: v1.20 - Found and resolved several edge-case bugs. Improved +, -, += and -= iterator operator performance tremendously.
• 2015-4-18: v1.10 - Allocators fully supported now. Improved performance. Some bugfixes. Added std::iterator trait to iterator. Added large struct tests to demo. Improved realism of array tests.
• 2015-4-13: v1.00 - Allocators now largely supported (some work to do). Erasure bug fixed. First proper release.
• 2015-4-9: v0.983 - Updated demo to be more realistic for array usage and to run all tests sequentially.
• 2015-4-6: v0.982 - Bugfixes, speedup, added find functions.
• 2015-4-2: v0.98 - First release

Contact:
plf:: library and this page Copyright (c) 2023, Matthew Bentley