upd: first version of database

This commit is contained in:
2023-12-04 07:52:18 +00:00
parent a7e167d30d
commit 01864d5a5e
3 changed files with 223 additions and 14 deletions

View File

@ -0,0 +1,53 @@
#ifndef BPT_DATABASE_HPP
#define BPT_DATABASE_HPP
#include <string>
#include <vector>
#include "drivearray.hpp"
#include "key2index.hpp"
template <class StorageType>
class DriveMultiMap_string {
private:
String2Index Indexer;
DriveArray<StorageType, 2, 100> Storage;
bool is_open = false;
public:
DriveMultiMap_string() = default;
void OpenFile(const std::string __file_name) noexcept {
Indexer.OpenFile(__file_name + ".idx");
Storage.OpenFile(__file_name + ".dat");
is_open = true;
}
std::vector<StorageType> Find(const std::string &key) noexcept {
if (!is_open) return {};
std::vector<StorageType> ret;
std::vector<int> idxs = std::move(Indexer.Find(key));
for (auto idx : idxs) {
StorageType tmp;
Storage.read(tmp, idx);
ret.push_back(tmp);
}
return std::move(ret);
}
void Delete(const std::string &key, const StorageType &value) noexcept {
if (!is_open) return;
std::vector<int> idxs = std::move(Indexer.Find(key));
for (auto idx : idxs) {
StorageType tmp;
Storage.read(tmp, idx);
if (tmp == value) {
Storage.Delete(idx);
Indexer.Delete(key, idx);
return;
}
}
}
void Insert(const std::string &key, StorageType &value) noexcept {
if (!is_open) return;
int idx = Storage.write(value);
Indexer.Insert(key, idx);
}
};
#endif // BPT_DATABASE_HPP

View File

@ -13,7 +13,7 @@
#include <stack>
#include <string>
template <class T, const int info_len = 2>
template <class T, const int info_len = 2, const int kRefreshThreshold = 100>
class DriveArray {
private:
static const int kPageSize = 4096;
@ -26,10 +26,8 @@ class DriveArray {
((info_len + 2) * sizeof(int) + kPageSize - 1) / kPageSize * kPageSize;
std::stack<int> free_mem;
int total_mem = 0;
const int kRefreshThreshold = 100;
unsigned int forced_refresh = 0;
std::mutex mtx;
void reallocate(bool include_resync = false) {
void reallocate(bool include_resync = false) noexcept {
size_t length_needed =
raw_data_begin +
(sizeofT * total_mem + kPageSize - 1) / kPageSize * kPageSize;
@ -43,7 +41,7 @@ class DriveArray {
virtual_mem = mmap(nullptr, file_length, PROT_READ | PROT_WRITE, MAP_SHARED,
file_descriptor, 0);
}
void ForceRefresh() {
void ForceRefresh() noexcept {
munmap(virtual_mem, file_length);
virtual_mem = mmap(nullptr, file_length, PROT_READ | PROT_WRITE, MAP_SHARED,
file_descriptor, 0);
@ -51,6 +49,7 @@ class DriveArray {
public:
DriveArray() = default;
inline bool IsOpen() const noexcept { return file_descriptor >= 0; }
~DriveArray() {
reallocate(true);
int stk_data_begin =
@ -65,11 +64,17 @@ class DriveArray {
}
munmap(virtual_mem, file_length);
close(file_descriptor);
file_descriptor = -1;
}
bool operator=(const DriveArray &) = delete;
DriveArray(const std::string &file_name) : file_name(file_name) {
void OpenFile(const std::string &file_name) {
if (file_name == "") return;
if (file_descriptor >= 0) {
munmap(virtual_mem, file_length);
close(file_descriptor);
file_descriptor = -1;
}
file_descriptor =
open(file_name.c_str(), O_CREAT | O_RDWR, S_IRUSR | S_IWUSR);
struct stat file_state;
@ -93,8 +98,9 @@ class DriveArray {
for (int i = 0; i < free_mem_cnt; i++) {
free_mem.push(*(p++));
}
// madvise(virtual_mem + stk_data_begin, free_mem_cnt * sizeof(int),
// MADV_FREE);
}
DriveArray(const std::string &file_name) : file_name(file_name) {
OpenFile(file_name);
}
void initialise(std::string FN = "") {
@ -102,6 +108,7 @@ class DriveArray {
if (file_descriptor >= 0) {
munmap(virtual_mem, file_length);
close(file_descriptor);
file_descriptor = -1;
}
file_descriptor =
open(file_name.c_str(), O_CREAT | O_RDWR, S_IRUSR | S_IWUSR);
@ -115,7 +122,7 @@ class DriveArray {
for (int i = 0; i < info_len; i++) *((int *)(virtual_mem) + i) = 0;
}
void get_info(int &tmp, int n) {
void get_info(int &tmp, int n) noexcept {
if (n > info_len) return;
tmp = *((int *)(virtual_mem) + n - 1);
if (++forced_refresh >= kRefreshThreshold) {
@ -124,7 +131,7 @@ class DriveArray {
}
}
void write_info(int tmp, int n) {
void write_info(int tmp, int n) noexcept {
if (n > info_len) return;
*((int *)(virtual_mem) + n - 1) = tmp;
if (++forced_refresh >= kRefreshThreshold) {
@ -133,7 +140,7 @@ class DriveArray {
}
}
int write(T &t) {
int write(T &t) noexcept {
int index = -1;
if (!free_mem.empty()) {
index = free_mem.top();
@ -144,7 +151,7 @@ class DriveArray {
return index;
}
void update(T &t, const int index) {
void update(T &t, const int index) noexcept {
reallocate();
void *data_begin = virtual_mem + raw_data_begin + sizeofT * (index - 1);
std::memmove(data_begin, &t, sizeofT);
@ -155,7 +162,7 @@ class DriveArray {
}
}
void read(T &t, const int index) {
void read(T &t, const int index) noexcept {
reallocate();
void *data_begin = virtual_mem + raw_data_begin + sizeofT * (index - 1);
std::memmove(&t, data_begin, sizeofT);
@ -166,7 +173,7 @@ class DriveArray {
}
}
void Delete(int index) { free_mem.push(index); }
void Delete(int index) noexcept { free_mem.push(index); }
};
#endif // BPT_DriveArray_HPP

View File

@ -0,0 +1,149 @@
#ifndef BPT_KEY2INDEX_HPP
#define BPT_KEY2INDEX_HPP
#include <cassert>
#include <functional>
#include <string>
#include <vector>
#include "drivearray.hpp"
class String2Index {
private:
static const int kBucketSize = 262142;
static const int kPageSize = 4096;
struct Node {
char str[66];
int val;
Node() = default;
Node(const std::string &_str, int _val) : val(_val) {
assert(_str.length() <= 64);
strcpy(str, _str.c_str());
}
};
static const int kNodesPerBlock =
(kPageSize - 2 * sizeof(int)) / sizeof(Node);
struct Block {
int tot, nxt_idx;
Node data[kNodesPerBlock];
char padding[kPageSize - 2 * sizeof(int) - sizeof(Node) * (kNodesPerBlock)];
Block() : tot(0), nxt_idx(0) {}
Block(int _tot, int _nxt_idx) : tot(_tot), nxt_idx(_nxt_idx) {}
};
static_assert(kNodesPerBlock >= 1, "kNodesPerBlock error");
static_assert(sizeof(Block) == kPageSize, "Block Size error");
DriveArray<Block, kBucketSize, 10> mem;
int *hash_table = nullptr;
std::string file_name;
inline size_t Hash(std::string str) noexcept {
const static std::string salt1 = "mL;]-=eT";
const static std::string salt2 = "9B<mF_me";
str = salt1 + str + salt2;
size_t ret = 0;
for (int i = 0; i < str.length(); ++i) ret = ret * 131 + str[i];
return ret;
}
public:
String2Index() = default;
void OpenFile(const std::string __file_name) noexcept {
file_name = __file_name;
if (file_name == "") return;
if (mem.IsOpen())
for (int i = 0; i < kBucketSize; ++i)
mem.write_info(hash_table[i], i + 1);
mem.OpenFile(file_name);
delete[] hash_table;
hash_table = new int[kBucketSize];
for (int i = 0; i < kBucketSize; ++i) mem.get_info(hash_table[i], i + 1);
}
String2Index(const std::string __file_name) : file_name(__file_name) {
OpenFile(file_name);
}
void initialise(std::string FN = "") noexcept {
if (FN != "") file_name = FN;
if (mem.IsOpen())
for (int i = 0; i < kBucketSize; ++i)
mem.write_info(hash_table[i], i + 1);
delete[] hash_table;
mem.initialise(file_name);
hash_table = new int[kBucketSize]();
}
~String2Index() {
for (int i = 0; i < kBucketSize; ++i) mem.write_info(hash_table[i], i + 1);
delete[] hash_table;
}
void Insert(const std::string &str, int val) noexcept {
size_t hash_val = Hash(str);
int idx = hash_table[hash_val % kBucketSize];
Block *blk_ptr = new Block;
if (idx == 0) {
Block __Init_Block;
idx = mem.write(__Init_Block);
hash_table[hash_val % kBucketSize] = idx;
assert(idx >= 1);
}
mem.read(*blk_ptr, idx);
if (blk_ptr->tot == kNodesPerBlock) {
Block __New_Head_Block(0, idx);
idx = mem.write(__New_Head_Block);
hash_table[hash_val % kBucketSize] = idx;
mem.read(*blk_ptr, idx);
}
blk_ptr->data[blk_ptr->tot++] = Node(str, val);
mem.update(*blk_ptr, idx);
delete blk_ptr;
}
void Delete(const std::string &str, int val) noexcept {
size_t hash_val = Hash(str);
int idx = hash_table[hash_val % kBucketSize];
Block *blk_ptr = new Block;
while (idx != 0) {
mem.read(*blk_ptr, idx);
for (int i = 0; i < blk_ptr->tot; ++i) {
if (blk_ptr->data[i].str == str && blk_ptr->data[i].val == val) {
int headidx = hash_table[hash_val % kBucketSize];
if (headidx == idx) {
blk_ptr->data[i] = blk_ptr->data[--blk_ptr->tot];
mem.update(*blk_ptr, idx);
} else {
Block *head_blk_ptr = new Block;
mem.read(*head_blk_ptr, headidx);
blk_ptr->data[i] = head_blk_ptr->data[--head_blk_ptr->tot];
if (head_blk_ptr->tot == 0) {
hash_table[hash_val % kBucketSize] = head_blk_ptr->nxt_idx;
mem.Delete(headidx);
} else
mem.update(*head_blk_ptr, headidx);
mem.update(*blk_ptr, idx);
delete head_blk_ptr;
}
delete blk_ptr;
return;
}
}
idx = blk_ptr->nxt_idx;
}
delete blk_ptr;
}
std::vector<int> Find(const std::string &str) noexcept {
std::vector<int> ret;
size_t hash_val = Hash(str);
int idx = hash_table[hash_val % kBucketSize];
Block *blk_ptr = new Block;
while (idx != 0) {
mem.read(*blk_ptr, idx);
for (int i = 0; i < blk_ptr->tot; ++i) {
if (blk_ptr->data[i].str == str) {
ret.push_back(blk_ptr->data[i].val);
}
}
idx = blk_ptr->nxt_idx;
}
delete blk_ptr;
return std::move(ret);
}
};
#endif // BPT_KEY2INDEX_HPP