blob: 39b8c8f1a9a1d4c18510554d153cf5583dcc8c2f [file] [log] [blame]
/* Copyright (c) 2015-2017, 2019-2023 The Khronos Group Inc.
* Copyright (c) 2015-2017, 2019-2023 Valve Corporation
* Copyright (c) 2015-2017, 2019-2023 LunarG, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#pragma once
#include <cmath>
#include <cassert>
#include <limits>
#include <memory>
#include <map>
#include <unordered_map>
#include <set>
#include <algorithm>
#include <iterator>
#include <type_traits>
#include <optional>
#include <utility>
#ifdef USE_ROBIN_HOOD_HASHING
#include "robin_hood.h"
#else
#include <unordered_set>
#endif
// namespace aliases to allow map and set implementations to easily be swapped out
namespace vvl {
#ifdef USE_ROBIN_HOOD_HASHING
template <typename T>
using hash = robin_hood::hash<T>;
template <typename Key, typename Hash = robin_hood::hash<Key>, typename KeyEqual = std::equal_to<Key>>
using unordered_set = robin_hood::unordered_set<Key, Hash, KeyEqual>;
template <typename Key, typename T, typename Hash = robin_hood::hash<Key>, typename KeyEqual = std::equal_to<Key>>
using unordered_map = robin_hood::unordered_map<Key, T, Hash, KeyEqual>;
template <typename Key, typename T>
using map_entry = robin_hood::pair<Key, T>;
// robin_hood-compatible insert_iterator (std:: uses the wrong insert method)
// NOTE: std::iterator was deprecated in C++17, and newer versions of libstdc++ appear to mark this as such.
template <typename T>
struct insert_iterator {
using iterator_category = std::output_iterator_tag;
using value_type = typename T::value_type;
using iterator = typename T::iterator;
using difference_type = void;
using pointer = void;
using reference = T &;
insert_iterator(reference t, iterator i) : container(&t), iter(i) {}
insert_iterator &operator=(const value_type &value) {
auto result = container->insert(value);
iter = result.first;
++iter;
return *this;
}
insert_iterator &operator=(value_type &&value) {
auto result = container->insert(std::move(value));
iter = result.first;
++iter;
return *this;
}
insert_iterator &operator*() { return *this; }
insert_iterator &operator++() { return *this; }
insert_iterator &operator++(int) { return *this; }
private:
T *container;
iterator iter;
};
#else
template <typename T>
using hash = std::hash<T>;
template <typename Key, typename Hash = std::hash<Key>, typename KeyEqual = std::equal_to<Key>>
using unordered_set = std::unordered_set<Key, Hash, KeyEqual>;
template <typename Key, typename T, typename Hash = std::hash<Key>, typename KeyEqual = std::equal_to<Key>>
using unordered_map = std::unordered_map<Key, T, Hash, KeyEqual>;
template <typename Key, typename T>
using map_entry = std::pair<Key, T>;
template <typename T>
using insert_iterator = std::insert_iterator<T>;
#endif
} // namespace vvl
// A vector class with "small string optimization" -- meaning that the class contains a fixed working store for N elements.
// Useful in in situations where the needed size is unknown, but the typical size is known If size increases beyond the
// fixed capacity, a dynamically allocated working store is created.
//
// NOTE: Unlike std::vector which only requires T to be CopyAssignable and CopyConstructable, small_vector requires T to be
// MoveAssignable and MoveConstructable
// NOTE: Unlike std::vector, iterators are invalidated by move assignment between small_vector objects effectively the
// "small string" allocation functions as an incompatible allocator.
template <typename T, size_t N, typename SizeType = uint8_t>
class small_vector {
public:
using value_type = T;
using reference = value_type &;
using const_reference = const value_type &;
using pointer = value_type *;
using const_pointer = const value_type *;
using iterator = pointer;
using const_iterator = const_pointer;
using size_type = SizeType;
static const size_type kSmallCapacity = N;
static const size_type kMaxCapacity = std::numeric_limits<size_type>::max();
static_assert(N <= kMaxCapacity, "size must be less than size_type::max");
small_vector() : size_(0), capacity_(N), working_store_(GetSmallStore()) {}
small_vector(std::initializer_list<T> list) : size_(0), capacity_(N), working_store_(GetSmallStore()) { PushBackFrom(list); }
small_vector(const small_vector &other) : size_(0), capacity_(N), working_store_(GetSmallStore()) { PushBackFrom(other); }
small_vector(small_vector &&other) : size_(0), capacity_(N), working_store_(GetSmallStore()) {
if (other.large_store_) {
MoveLargeStore(other);
} else {
PushBackFrom(std::move(other));
}
// Per the spec, when constructing from other, other is guaranteed to be empty after the constructor runs
other.clear();
}
small_vector(size_type size, const value_type &value = value_type()) : size_(0), capacity_(N), working_store_(GetSmallStore()) {
reserve(size);
auto dest = GetWorkingStore();
for (size_type i = 0; i < size; i++) {
new (dest) value_type(value);
++dest;
}
size_ = size;
}
~small_vector() { clear(); }
bool operator==(const small_vector &rhs) const {
if (size_ != rhs.size_) return false;
auto value = begin();
for (const auto &rh_value : rhs) {
if (!(*value == rh_value)) {
return false;
}
++value;
}
return true;
}
bool operator!=(const small_vector &rhs) const { return !(*this == rhs); }
small_vector &operator=(const small_vector &other) {
if (this != &other) {
if (other.size_ > capacity_) {
// Calling reserve would move construct and destroy all current contents, so just clear them before calling
// PushBackFrom (which does a reserve vs. the now empty this)
clear();
PushBackFrom(other);
} else {
// The copy will fit into the current allocation
auto dest = GetWorkingStore();
auto source = other.GetWorkingStore();
const auto overlap = std::min(size_, other.size_);
// Copy assign anywhere we have objects in this
// Note: usually cheaper than destruct/construct
for (size_type i = 0; i < overlap; i++) {
dest[i] = source[i];
}
// Copy construct anywhere we *don't* have objects in this
for (size_type i = overlap; i < other.size_; i++) {
new (dest + i) value_type(source[i]);
}
// Any entries in this past other_size_ must be cleaned up...
for (size_type i = other.size_; i < size_; i++) {
dest[i].~value_type();
}
size_ = other.size_;
}
}
return *this;
}
small_vector &operator=(small_vector &&other) {
if (this != &other) {
// Note: move assign doesn't require other to become empty (as does move construction)
// so we'll leave other alone except in the large store case, while moving the object
// *in* the vector from other
if (other.large_store_) {
// Moving the other large store intact is probably best, even if we have to destroy everything in this.
clear();
MoveLargeStore(other);
} else if (other.size_ > capacity_) {
// If we'd have to reallocate, just clean up minimally and copy normally
clear();
PushBackFrom(std::move(other));
} else {
// The copy will fit into the current allocation
auto dest = GetWorkingStore();
auto source = other.GetWorkingStore();
const auto overlap = std::min(size_, other.size_);
// Move assign where we have objects in this
// Note: usually cheaper than destruct/construct
for (size_type i = 0; i < overlap; i++) {
dest[i] = std::move(source[i]);
}
// Move construct where we *don't* have objects in this
for (size_type i = overlap; i < other.size_; i++) {
new (dest + i) value_type(std::move(source[i]));
}
// Any entries in this past other_size_ must be cleaned up...
for (size_type i = other.size_; i < size_; i++) {
dest[i].~value_type();
}
size_ = other.size_;
}
}
return *this;
}
reference operator[](size_type pos) {
assert(pos < size_);
return GetWorkingStore()[pos];
}
const_reference operator[](size_type pos) const {
assert(pos < size_);
return GetWorkingStore()[pos];
}
// Like std::vector:: calling front or back on an empty container causes undefined behavior
reference front() {
assert(size_ > 0);
return GetWorkingStore()[0];
}
const_reference front() const {
assert(size_ > 0);
return GetWorkingStore()[0];
}
reference back() {
assert(size_ > 0);
return GetWorkingStore()[size_ - 1];
}
const_reference back() const {
assert(size_ > 0);
return GetWorkingStore()[size_ - 1];
}
bool empty() const { return size_ == 0; }
template <class... Args>
void emplace_back(Args &&...args) {
assert(size_ < kMaxCapacity);
reserve(size_ + 1);
new (GetWorkingStore() + size_) value_type(args...);
size_++;
}
// Note: probably should update this to reflect C++23 ranges
template <typename Container>
void PushBackFrom(const Container &from) {
assert(from.size() <= kMaxCapacity);
assert(size_ <= kMaxCapacity - from.size());
const size_type new_size = size_ + static_cast<size_type>(from.size());
reserve(new_size);
auto dest = GetWorkingStore() + size_;
for (const auto &element : from) {
new (dest) value_type(element);
++dest;
}
size_ = new_size;
}
template <typename Container>
void PushBackFrom(Container &&from) {
assert(from.size() < kMaxCapacity);
const size_type new_size = size_ + static_cast<size_type>(from.size());
reserve(new_size);
auto dest = GetWorkingStore() + size_;
for (auto &element : from) {
new (dest) value_type(std::move(element));
++dest;
}
size_ = new_size;
}
void reserve(size_type new_cap) {
// Since this can't shrink, if we're growing we're newing
if (new_cap > capacity_) {
assert(capacity_ >= kSmallCapacity);
auto new_store = std::unique_ptr<BackingStore[]>(new BackingStore[new_cap]);
auto working_store = GetWorkingStore();
for (size_type i = 0; i < size_; i++) {
new (new_store[i].data) value_type(std::move(working_store[i]));
working_store[i].~value_type();
}
large_store_ = std::move(new_store);
assert(new_cap > kSmallCapacity);
capacity_ = new_cap;
}
UpdateWorkingStore();
// No shrink here.
}
void clear() {
// Keep clear minimal to optimize reset functions for enduring objects
// more work is deferred until destruction (freeing of large_store for example)
// and we intentionally *aren't* shrinking. Callers that desire shrink semantics
// can call shrink_to_fit.
auto working_store = GetWorkingStore();
for (size_type i = 0; i < size_; i++) {
working_store[i].~value_type();
}
size_ = 0;
}
void resize(size_type count) {
struct ValueInitTag { // tag to request value-initialization
explicit ValueInitTag() = default;
};
Resize(count, ValueInitTag{});
}
void resize(size_type count, const value_type &value) { Resize(count, value); }
void shrink_to_fit() {
if (size_ == 0) {
// shrink resets to small when empty
capacity_ = kSmallCapacity;
large_store_.reset();
UpdateWorkingStore();
} else if ((capacity_ > kSmallCapacity) && (capacity_ > size_)) {
auto source = GetWorkingStore();
// Keep the source from disappearing until the end of the function
auto old_store = std::unique_ptr<BackingStore[]>(std::move(large_store_));
assert(!large_store_);
if (size_ < kSmallCapacity) {
capacity_ = kSmallCapacity;
} else {
large_store_ = std::unique_ptr<BackingStore[]>(new BackingStore[size_]);
capacity_ = size_;
}
UpdateWorkingStore();
auto dest = GetWorkingStore();
for (size_type i = 0; i < size_; i++) {
dest[i] = std::move(source[i]);
source[i].~value_type();
}
}
}
inline iterator begin() { return GetWorkingStore(); }
inline const_iterator cbegin() const { return GetWorkingStore(); }
inline const_iterator begin() const { return GetWorkingStore(); }
inline iterator end() { return GetWorkingStore() + size_; }
inline const_iterator cend() const { return GetWorkingStore() + size_; }
inline const_iterator end() const { return GetWorkingStore() + size_; }
inline size_type size() const { return size_; }
auto capacity() const { return capacity_; }
inline pointer data() { return GetWorkingStore(); }
inline const_pointer data() const { return GetWorkingStore(); }
protected:
inline const_pointer ComputeWorkingStore() const {
assert(large_store_ || (capacity_ == kSmallCapacity));
const BackingStore *store = large_store_ ? large_store_.get() : small_store_;
return &store->object;
}
inline pointer ComputeWorkingStore() {
assert(large_store_ || (capacity_ == kSmallCapacity));
BackingStore *store = large_store_ ? large_store_.get() : small_store_;
return &store->object;
}
void UpdateWorkingStore() { working_store_ = ComputeWorkingStore(); }
inline const_pointer GetWorkingStore() const {
DbgWorkingStoreCheck();
return working_store_;
}
inline pointer GetWorkingStore() {
DbgWorkingStoreCheck();
return working_store_;
}
inline pointer GetSmallStore() { return &small_store_->object; }
union BackingStore {
BackingStore() {}
~BackingStore() {}
uint8_t data[sizeof(value_type)];
value_type object;
};
size_type size_;
size_type capacity_;
BackingStore small_store_[N];
std::unique_ptr<BackingStore[]> large_store_;
value_type *working_store_;
#ifndef NDEBUG
void DbgWorkingStoreCheck() const { assert(ComputeWorkingStore() == working_store_); };
#else
void DbgWorkingStoreCheck() const {};
#endif
private:
void MoveLargeStore(small_vector &other) {
assert(other.large_store_);
assert(other.capacity_ > kSmallCapacity);
// In move operations, from a small vector with a large store, we can move from it
large_store_ = std::move(other.large_store_);
capacity_ = other.capacity_;
size_ = other.size_;
UpdateWorkingStore();
// We've stolen other's large store, must leave it in a valid state
other.size_ = 0;
other.capacity_ = kSmallCapacity;
other.UpdateWorkingStore();
}
template <typename T2>
void Resize(size_type new_size, const T2 &value) {
if (new_size < size_) {
auto working_store = GetWorkingStore();
for (size_type i = new_size; i < size_; i++) {
working_store[i].~value_type();
}
size_ = new_size;
} else if (new_size > size_) {
reserve(new_size);
// if T2 != T and T is not DefaultInsertable, new values will be undefined
if constexpr (std::is_same_v<T2, T> || std::is_default_constructible_v<T>) {
for (size_type i = size_; i < new_size; ++i) {
if constexpr (std::is_same_v<T2, T>) {
emplace_back(value_type(value));
} else if constexpr (std::is_default_constructible_v<T>) {
emplace_back(value_type());
}
}
assert(size() == new_size);
} else {
size_ = new_size;
}
}
}
};
// This is a wrapper around unordered_map that optimizes for the common case
// of only containing a small number of elements. The first N elements are stored
// inline in the object and don't require hashing or memory (de)allocation.
template <typename Key, typename value_type, typename inner_container_type, typename value_type_helper, int N>
class small_container {
protected:
bool small_data_allocated[N];
value_type small_data[N];
inner_container_type inner_cont;
value_type_helper helper;
public:
small_container() {
for (int i = 0; i < N; ++i) {
small_data_allocated[i] = false;
}
}
class iterator {
typedef typename inner_container_type::iterator inner_iterator;
friend class small_container<Key, value_type, inner_container_type, value_type_helper, N>;
small_container<Key, value_type, inner_container_type, value_type_helper, N> *parent;
int index;
inner_iterator it;
public:
iterator() {}
iterator operator++() {
if (index < N) {
index++;
while (index < N && !parent->small_data_allocated[index]) {
index++;
}
if (index < N) {
return *this;
}
it = parent->inner_cont.begin();
return *this;
}
++it;
return *this;
}
bool operator==(const iterator &other) const {
if ((index < N) != (other.index < N)) {
return false;
}
if (index < N) {
return (index == other.index);
}
return it == other.it;
}
bool operator!=(const iterator &other) const { return !(*this == other); }
value_type &operator*() const {
if (index < N) {
return parent->small_data[index];
}
return *it;
}
value_type *operator->() const {
if (index < N) {
return &parent->small_data[index];
}
return &*it;
}
};
class const_iterator {
typedef typename inner_container_type::const_iterator inner_iterator;
friend class small_container<Key, value_type, inner_container_type, value_type_helper, N>;
const small_container<Key, value_type, inner_container_type, value_type_helper, N> *parent;
int index;
inner_iterator it;
public:
const_iterator() {}
const_iterator operator++() {
if (index < N) {
index++;
while (index < N && !parent->small_data_allocated[index]) {
index++;
}
if (index < N) {
return *this;
}
it = parent->inner_cont.begin();
return *this;
}
++it;
return *this;
}
bool operator==(const const_iterator &other) const {
if ((index < N) != (other.index < N)) {
return false;
}
if (index < N) {
return (index == other.index);
}
return it == other.it;
}
bool operator!=(const const_iterator &other) const { return !(*this == other); }
const value_type &operator*() const {
if (index < N) {
return parent->small_data[index];
}
return *it;
}
const value_type *operator->() const {
if (index < N) {
return &parent->small_data[index];
}
return &*it;
}
};
iterator begin() {
iterator it;
it.parent = this;
// If index 0 is allocated, return it, otherwise use operator++ to find the first
// allocated element.
it.index = 0;
if (small_data_allocated[0]) {
return it;
}
++it;
return it;
}
iterator end() {
iterator it;
it.parent = this;
it.index = N;
it.it = inner_cont.end();
return it;
}
const_iterator begin() const {
const_iterator it;
it.parent = this;
// If index 0 is allocated, return it, otherwise use operator++ to find the first
// allocated element.
it.index = 0;
if (small_data_allocated[0]) {
return it;
}
++it;
return it;
}
const_iterator end() const {
const_iterator it;
it.parent = this;
it.index = N;
it.it = inner_cont.end();
return it;
}
bool contains(const Key &key) const {
for (int i = 0; i < N; ++i) {
if (small_data_allocated[i] && helper.compare_equal(small_data[i], key)) {
return true;
}
}
// check size() first to avoid hashing key unnecessarily.
if (inner_cont.size() == 0) {
return false;
}
return inner_cont.find(key) != inner_cont.end();
}
typename inner_container_type::size_type count(const Key &key) const { return contains(key) ? 1 : 0; }
std::pair<iterator, bool> insert(const value_type &value) {
for (int i = 0; i < N; ++i) {
if (small_data_allocated[i] && helper.compare_equal(small_data[i], value)) {
iterator it;
it.parent = this;
it.index = i;
return std::make_pair(it, false);
}
}
// check size() first to avoid hashing key unnecessarily.
auto iter = inner_cont.size() > 0 ? inner_cont.find(helper.get_key(value)) : inner_cont.end();
if (iter != inner_cont.end()) {
iterator it;
it.parent = this;
it.index = N;
it.it = iter;
return std::make_pair(it, false);
} else {
for (int i = 0; i < N; ++i) {
if (!small_data_allocated[i]) {
small_data_allocated[i] = true;
helper.assign(small_data[i], value);
iterator it;
it.parent = this;
it.index = i;
return std::make_pair(it, true);
}
}
iter = inner_cont.insert(value).first;
iterator it;
it.parent = this;
it.index = N;
it.it = iter;
return std::make_pair(it, true);
}
}
typename inner_container_type::size_type erase(const Key &key) {
for (int i = 0; i < N; ++i) {
if (small_data_allocated[i] && helper.compare_equal(small_data[i], key)) {
small_data_allocated[i] = false;
return 1;
}
}
return inner_cont.erase(key);
}
typename inner_container_type::size_type size() const {
auto size = inner_cont.size();
for (int i = 0; i < N; ++i) {
if (small_data_allocated[i]) {
size++;
}
}
return size;
}
bool empty() const {
for (int i = 0; i < N; ++i) {
if (small_data_allocated[i]) {
return false;
}
}
return inner_cont.size() == 0;
}
void clear() {
for (int i = 0; i < N; ++i) {
small_data_allocated[i] = false;
}
inner_cont.clear();
}
};
// Helper function objects to compare/assign/get keys in small_unordered_set/map.
// This helps to abstract away whether value_type is a Key or a pair<Key, T>.
template <typename MapType>
class value_type_helper_map {
using PairType = typename MapType::value_type;
using Key = typename std::remove_const<typename PairType::first_type>::type;
public:
bool compare_equal(const PairType &lhs, const Key &rhs) const { return lhs.first == rhs; }
bool compare_equal(const PairType &lhs, const PairType &rhs) const { return lhs.first == rhs.first; }
void assign(PairType &lhs, const PairType &rhs) const {
// While the const_cast may be unsatisfactory, we are using small_data as
// stand-in for placement new and a small-block allocator, so the const_cast
// is minimal, contained, valid, and allows operators * and -> to avoid copies
const_cast<Key &>(lhs.first) = rhs.first;
lhs.second = rhs.second;
}
Key get_key(const PairType &value) const { return value.first; }
};
template <typename Key>
class value_type_helper_set {
public:
bool compare_equal(const Key &lhs, const Key &rhs) const { return lhs == rhs; }
void assign(Key &lhs, const Key &rhs) const { lhs = rhs; }
Key get_key(const Key &value) const { return value; }
};
template <typename Key, typename T, int N = 1>
class small_unordered_map : public small_container<Key, typename vvl::unordered_map<Key, T>::value_type, vvl::unordered_map<Key, T>,
value_type_helper_map<vvl::unordered_map<Key, T>>, N> {
public:
T &operator[](const Key &key) {
for (int i = 0; i < N; ++i) {
if (this->small_data_allocated[i] && this->helper.compare_equal(this->small_data[i], key)) {
return this->small_data[i].second;
}
}
auto iter = this->inner_cont.find(key);
if (iter != this->inner_cont.end()) {
return iter->second;
} else {
for (int i = 0; i < N; ++i) {
if (!this->small_data_allocated[i]) {
this->small_data_allocated[i] = true;
this->helper.assign(this->small_data[i], {key, T()});
return this->small_data[i].second;
}
}
return this->inner_cont[key];
}
}
};
template <typename Key, int N = 1>
class small_unordered_set : public small_container<Key, Key, vvl::unordered_set<Key>, value_type_helper_set<Key>, N> {};
// For the given data key, look up the layer_data instance from given layer_data_map
template <typename DATA_T>
DATA_T *GetLayerDataPtr(void *data_key, small_unordered_map<void *, DATA_T *, 2> &layer_data_map) {
/* TODO: We probably should lock here, or have caller lock */
DATA_T *&got = layer_data_map[data_key];
if (got == nullptr) {
got = new DATA_T;
}
return got;
}
template <typename DATA_T>
void FreeLayerDataPtr(void *data_key, small_unordered_map<void *, DATA_T *, 2> &layer_data_map) {
delete layer_data_map[data_key];
layer_data_map.erase(data_key);
}
// For the given data key, look up the layer_data instance from given layer_data_map
template <typename DATA_T>
DATA_T *GetLayerDataPtr(void *data_key, std::unordered_map<void *, DATA_T *> &layer_data_map) {
DATA_T *debug_data;
/* TODO: We probably should lock here, or have caller lock */
auto got = layer_data_map.find(data_key);
if (got == layer_data_map.end()) {
debug_data = new DATA_T;
layer_data_map[(void *)data_key] = debug_data;
} else {
debug_data = got->second;
}
return debug_data;
}
template <typename DATA_T>
void FreeLayerDataPtr(void *data_key, std::unordered_map<void *, DATA_T *> &layer_data_map) {
auto got = layer_data_map.find(data_key);
assert(got != layer_data_map.end());
delete got->second;
layer_data_map.erase(got);
}
namespace vvl {
inline constexpr std::in_place_t in_place{};
// Partial implementation of std::span for C++11
template <typename T>
class span {
public:
using pointer = T *;
using const_pointer = T const *;
using iterator = pointer;
using const_iterator = const_pointer;
span() = default;
span(pointer start, size_t n) : data_(start), count_(n) {}
template <typename Iterator>
span(Iterator start, Iterator end) : data_(&(*start)), count_(end - start) {}
template <typename Container>
span(Container &c) : data_(c.data()), count_(c.size()) {}
iterator begin() { return data_; }
const_iterator begin() const { return data_; }
iterator end() { return data_ + count_; }
const_iterator end() const { return data_ + count_; }
T &operator[](int i) { return data_[i]; }
const T &operator[](int i) const { return data_[i]; }
T &front() { return *data_; }
const T &front() const { return *data_; }
T &back() { return *(data_ + (count_ - 1)); }
const T &back() const { return *(data_ + (count_ - 1)); }
size_t size() const { return count_; }
bool empty() const { return count_ == 0; }
pointer data() { return data_; }
const_pointer data() const { return data_; }
private:
pointer data_ = {};
size_t count_ = 0;
};
//
// Allow type inference that using the constructor doesn't allow in C++11
template <typename T>
span<T> make_span(T *begin, size_t count) {
return span<T>(begin, count);
}
template <typename T>
span<T> make_span(T *begin, T *end) {
return make_span<T>(begin, end);
}
template <typename BaseType>
using base_type =
typename std::remove_reference<typename std::remove_const<typename std::remove_pointer<BaseType>::type>::type>::type;
// Helper for thread local Validate -> Record phase data
// Define T unique to each entrypoint which will persist data
// Use only in with singleton (leaf) validation objects
// State machine transition state changes of payload relative to TlsGuard object lifecycle:
// State INIT: bool(payload_)
// State RESET: NOT bool(payload_)
// * PreCallValidate* phase
// * Initialized with skip (in PreCallValidate*)
// * RESET -> INIT
// * Destruct with skip == true
// * INIT -> RESET
// * PreCallRecord* phase (optional IF PostCallRecord present)
// * Initialized w/o skip (set "persist_" IFF PostCallRecord present)
// * Must be in state INIT
// * Destruct with NOT persist_
// * INIT -> RESET
// * PreCallRecord* phase (optional IF PreCallRecord present)
// * Initialized w/o skip ("persist_" *must be false)
// * Must be in state INIT
// * Destruct
// * INIT -> RESET
struct TlsGuardPersist {};
template <typename T>
class TlsGuard {
public:
// For use on inital references -- Validate phase
template <typename... Args>
TlsGuard(bool *skip, Args &&...args) : skip_(skip), persist_(false) {
// Record phase calls are required to clean up payload
assert(!payload_);
payload_.emplace(std::forward<Args>(args)...);
}
// For use on non-terminal persistent references (PreRecord phase IFF PostRecord is also present.
TlsGuard(const TlsGuardPersist &) : skip_(nullptr), persist_(true) { assert(payload_); }
// For use on terminal persistent references
// Validate phase calls are required to setup payload
// PreCallRecord calls are required to preserve (persist_) payload, if PostCallRecord calls will use
TlsGuard() : skip_(nullptr), persist_(false) { assert(payload_); }
~TlsGuard() {
assert(payload_);
if (!persist_ && (!skip_ || *skip_)) payload_.reset();
}
T &operator*() & {
assert(payload_);
return *payload_;
}
const T &operator*() const & {
assert(payload_);
return *payload_;
}
T &&operator*() && {
assert(payload_);
return std::move(*payload_);
}
T *operator->() { return &(*payload_); }
private:
inline thread_local static std::optional<T> payload_{};
bool *skip_;
bool persist_;
};
// Only use this if you aren't planning to use what you would have gotten from a find.
template <typename Container, typename Key = typename Container::key_type>
bool Contains(const Container &container, const Key &key) {
return container.find(key) != container.cend();
}
// EraseIf is not implemented as std::erase(std::remove_if(...), ...) for two reasons:
// 1) Robin Hood containers don't support two-argument erase functions
// 2) STL remove_if requires the predicate to be const w.r.t the value-type, and std::erase_if doesn't AFAICT
template <typename Container, typename Predicate>
typename Container::size_type EraseIf(Container &c, Predicate &&p) {
const auto before_size = c.size();
auto pos = c.begin();
while (pos != c.end()) {
if (p(*pos)) {
pos = c.erase(pos);
} else {
++pos;
}
}
return before_size - c.size();
}
template <typename T>
constexpr T MaxTypeValue([[maybe_unused]] T) {
return std::numeric_limits<T>::max();
}
template <typename T>
constexpr T MinTypeValue([[maybe_unused]] T) {
return std::numeric_limits<T>::min();
}
// Typesafe UINT32_MAX
constexpr auto kU32Max = std::numeric_limits<uint32_t>::max();
// Typesafe UINT64_MAX
constexpr auto kU64Max = std::numeric_limits<uint64_t>::max();
// Typesafe INT32_MAX
constexpr auto kI32Max = std::numeric_limits<int32_t>::max();
// Typesafe INT64_MAX
constexpr auto kI64Max = std::numeric_limits<int64_t>::max();
template <typename T>
T GetQuotientCeil(T numerator, T denominator) {
denominator = std::max(denominator, T{1});
return static_cast<T>(std::ceil(static_cast<double>(numerator) / static_cast<double>(denominator)));
}
} // namespace vvl