1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-07-27 17:27:35 +00:00

LibSQL: Remove unused HashIndex and related classes

The features provided by these classes should be used eventually, but so
far we've been maintaining these classes for over 2 years without any
actual use. We can restore them when it comes time to actually use them.
This commit is contained in:
Timothy Flynn 2023-08-07 11:27:12 -04:00 committed by Tim Flynn
parent 1151ba333a
commit 1b40bf9783
6 changed files with 0 additions and 956 deletions

View file

@ -15,7 +15,6 @@ set(SOURCES
BTree.cpp
BTreeIterator.cpp
Database.cpp
HashIndex.cpp
Heap.cpp
Index.cpp
Key.cpp

View file

@ -11,10 +11,6 @@ class BTree;
class BTreeIterator;
class ColumnDef;
class Database;
class HashBucket;
class HashDirectoryNode;
class HashIndex;
class HashIndexIterator;
class Heap;
class Index;
class IndexNode;

View file

@ -1,434 +0,0 @@
/*
* Copyright (c) 2021, Jan de Visser <jan@de-visser.net>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <LibSQL/HashIndex.h>
#include <LibSQL/Heap.h>
#include <LibSQL/Key.h>
#include <LibSQL/Serializer.h>
namespace SQL {
HashDirectoryNode::HashDirectoryNode(HashIndex& index, u32 node_number, size_t offset)
: IndexNode(index.node_pointer(node_number))
, m_hash_index(index)
, m_node_number(node_number)
, m_offset(offset)
{
}
HashDirectoryNode::HashDirectoryNode(HashIndex& index, u32 pointer)
: IndexNode(pointer)
, m_hash_index(index)
{
}
void HashDirectoryNode::deserialize(Serializer& serializer)
{
dbgln_if(SQL_DEBUG, "Deserializing Hash Directory Node");
m_hash_index.m_global_depth = serializer.deserialize<u32>();
auto size = serializer.deserialize<u32>();
dbgln_if(SQL_DEBUG, "Global Depth {}, #Bucket pointers {}", m_hash_index.global_depth(), size);
auto next_node = serializer.deserialize<u32>();
if (next_node) {
dbgln_if(SQL_DEBUG, "Next node {}", next_node);
m_hash_index.m_nodes.append(next_node);
} else {
dbgln_if(SQL_DEBUG, "This is the last directory node");
m_is_last = true;
}
for (auto ix = 0u; ix < size; ix++) {
auto bucket_pointer = serializer.deserialize<u32>();
auto local_depth = serializer.deserialize<u32>();
dbgln_if(SQL_DEBUG, "--Index {} bucket pointer {} local depth {}", ix, bucket_pointer, local_depth);
m_hash_index.append_bucket(ix, local_depth, bucket_pointer);
}
}
void HashDirectoryNode::serialize(Serializer& serializer) const
{
dbgln_if(SQL_DEBUG, "Serializing directory node #{}. Offset {}", m_node_number, m_offset);
serializer.serialize<u32>((u32)m_hash_index.global_depth());
serializer.serialize<u32>(number_of_pointers());
dbgln_if(SQL_DEBUG, "Global depth {}, #bucket pointers {}", m_hash_index.global_depth(), number_of_pointers());
u32 next_node;
if (m_node_number < (m_hash_index.m_nodes.size() - 1)) {
next_node = m_hash_index.m_nodes[m_node_number + 1];
dbgln_if(SQL_DEBUG, "Next directory node pointer {}", next_node);
} else {
next_node = 0u;
dbgln_if(SQL_DEBUG, "This is the last directory node");
}
serializer.serialize<u32>(next_node);
for (auto ix = 0u; ix < number_of_pointers(); ix++) {
auto& bucket = m_hash_index.m_buckets[m_offset + ix];
dbgln_if(SQL_DEBUG, "Bucket index #{} block_index {} local depth {} size {}", ix, bucket->block_index(), bucket->local_depth(), bucket->size());
serializer.serialize<u32>(bucket->block_index());
serializer.serialize<u32>(bucket->local_depth());
}
}
HashBucket::HashBucket(HashIndex& hash_index, Block::Index index, u32 local_depth, Block::Index pointer)
: IndexNode(pointer)
, m_hash_index(hash_index)
, m_local_depth(local_depth)
, m_index(index)
{
}
void HashBucket::serialize(Serializer& serializer) const
{
dbgln_if(SQL_DEBUG, "Serializing bucket: block_index {}, index #{}, local depth {} size {}",
block_index(), index(), local_depth(), size());
serializer.serialize<u32>(local_depth());
serializer.serialize<u32>(size());
for (auto& key : m_entries)
serializer.serialize<Key>(key);
}
void HashBucket::deserialize(Serializer& serializer)
{
if (m_inflated || !block_index())
return;
dbgln_if(SQL_DEBUG, "Inflating Hash Bucket {}", block_index());
m_local_depth = serializer.deserialize<u32>();
dbgln_if(SQL_DEBUG, "Bucket Local Depth {}", m_local_depth);
auto size = serializer.deserialize<u32>();
dbgln_if(SQL_DEBUG, "Bucket has {} keys", size);
for (auto ix = 0u; ix < size; ix++) {
auto key = serializer.deserialize<Key>(m_hash_index.descriptor());
dbgln_if(SQL_DEBUG, "Key {}: {}", ix, key.to_deprecated_string());
m_entries.append(key);
}
m_inflated = true;
}
size_t HashBucket::length() const
{
size_t len = 2 * sizeof(u32);
for (auto& key : m_entries)
len += key.length();
return len;
}
Optional<u32> HashBucket::get(Key& key)
{
auto optional_index = find_key_in_bucket(key);
if (optional_index.has_value()) {
auto& k = m_entries[optional_index.value()];
key.set_block_index(k.block_index());
return k.block_index();
}
return {};
}
bool HashBucket::insert(Key const& key)
{
if (!m_inflated)
m_hash_index.serializer().deserialize_block_to(block_index(), *this);
if (find_key_in_bucket(key).has_value())
return false;
if (length() + key.length() > Block::DATA_SIZE) {
dbgln_if(SQL_DEBUG, "Adding key {} would make length exceed block size", key.to_deprecated_string());
return false;
}
m_entries.append(key);
m_hash_index.serializer().serialize_and_write(*this);
return true;
}
Optional<size_t> HashBucket::find_key_in_bucket(Key const& key)
{
for (auto ix = 0u; ix < size(); ix++) {
auto& k = entries()[ix];
if (k == key)
return ix;
}
return {};
}
HashBucket const* HashBucket::next_bucket()
{
for (auto ix = m_index + 1; ix < m_hash_index.size(); ix++) {
auto bucket = m_hash_index.get_bucket_by_index(ix);
m_hash_index.serializer().deserialize_block_to<HashBucket>(bucket->block_index(), *bucket);
if (bucket->size())
return bucket;
}
return nullptr;
}
HashBucket const* HashBucket::previous_bucket()
{
for (auto ix = m_index - 1; ix > 0; ix--) {
auto bucket = m_hash_index.get_bucket_by_index(ix);
if (bucket->block_index() > 0)
return bucket;
}
return nullptr;
}
Vector<Key> const& HashBucket::entries()
{
if (!m_inflated)
m_hash_index.serializer().deserialize_block_to(block_index(), *this);
return m_entries;
}
Key const& HashBucket::operator[](size_t ix)
{
if (!m_inflated)
m_hash_index.serializer().deserialize_block_to(block_index(), *this);
return m_entries[ix];
}
Key const& HashBucket::operator[](size_t ix) const
{
return m_entries[ix];
}
void HashBucket::list_bucket()
{
warnln("Bucket #{} size {} local depth {} block_index {}{}",
index(), size(), local_depth(), block_index(), (block_index() > 0 ? "" : " (VIRTUAL)"));
for (auto& key : entries())
warnln(" {} hash {}", key.to_deprecated_string(), key.hash());
}
HashIndex::HashIndex(Serializer& serializer, NonnullRefPtr<TupleDescriptor> const& descriptor, Block::Index first_node)
: Index(serializer, descriptor, true, first_node)
{
if (first_node == 0)
set_block_index(request_new_block_index());
if (serializer.has_block(first_node)) {
Block::Index block_index = first_node;
do {
VERIFY(serializer.has_block(block_index));
auto node = serializer.deserialize_block<HashDirectoryNode>(block_index, *this, block_index);
if (node.is_last())
break;
block_index = m_nodes.last(); // FIXME Ugly
} while (block_index);
} else {
auto bucket = append_bucket(0u, 1u, request_new_block_index());
bucket->m_inflated = true;
serializer.serialize_and_write(*bucket);
bucket = append_bucket(1u, 1u, request_new_block_index());
bucket->m_inflated = true;
serializer.serialize_and_write(*bucket);
m_nodes.append(first_node);
write_directory();
}
}
HashBucket* HashIndex::get_bucket(u32 index)
{
VERIFY(index < m_buckets.size());
auto divisor = size() / 2;
while (m_buckets[index]->block_index() == 0) {
VERIFY(divisor > 1);
index = index % divisor;
divisor /= 2;
}
auto& bucket = m_buckets[index];
return bucket;
}
HashBucket* HashIndex::get_bucket_for_insert(Key const& key)
{
auto key_hash = key.hash();
do {
dbgln_if(SQL_DEBUG, "HashIndex::get_bucket_for_insert({}) bucket {} of {}", key.to_deprecated_string(), key_hash % size(), size());
auto bucket = get_bucket(key_hash % size());
if (bucket->length() + key.length() < Block::DATA_SIZE)
return bucket;
dbgln_if(SQL_DEBUG, "Bucket is full (bucket size {}/length {} key length {}). Expanding directory", bucket->size(), bucket->length(), key.length());
// We previously doubled the directory but the target bucket is
// still at an older depth. Create new buckets at the current global
// depth and allocate the contents of the existing buckets to the
// newly created ones:
while (bucket->local_depth() < global_depth()) {
auto base_index = bucket->index();
auto step = 1 << (global_depth() - bucket->local_depth());
auto total_moved = 0;
for (auto ix = base_index + step; ix < size(); ix += step) {
auto& sub_bucket = m_buckets[ix];
sub_bucket->set_local_depth(bucket->local_depth() + 1);
auto moved = 0;
for (auto entry_index = (int)bucket->m_entries.size() - 1; entry_index >= 0; entry_index--) {
if (bucket->m_entries[entry_index].hash() % size() == ix) {
if (!sub_bucket->block_index())
sub_bucket->set_block_index(request_new_block_index());
sub_bucket->insert(bucket->m_entries.take(entry_index));
moved++;
}
}
if (moved > 0) {
dbgln_if(SQL_DEBUG, "Moved {} entries from bucket #{} to #{}", moved, base_index, ix);
serializer().serialize_and_write(*sub_bucket);
}
total_moved += moved;
}
if (total_moved)
dbgln_if(SQL_DEBUG, "Redistributed {} entries from bucket #{}", total_moved, base_index);
else
dbgln_if(SQL_DEBUG, "Nothing redistributed from bucket #{}", base_index);
bucket->set_local_depth(bucket->local_depth() + 1);
serializer().serialize_and_write(*bucket);
write_directory();
auto bucket_after_redistribution = get_bucket(key_hash % size());
if (bucket_after_redistribution->length() + key.length() < Block::DATA_SIZE)
return bucket_after_redistribution;
}
expand();
} while (true);
VERIFY_NOT_REACHED();
}
void HashIndex::expand()
{
auto sz = size();
dbgln_if(SQL_DEBUG, "Expanding directory from {} to {} buckets", sz, 2 * sz);
for (auto i = 0u; i < sz; i++) {
auto bucket = get_bucket(i);
bucket = append_bucket(sz + i, bucket->local_depth(), 0u);
bucket->m_inflated = true;
}
m_global_depth++;
write_directory();
}
void HashIndex::write_directory()
{
auto num_nodes_required = (size() / HashDirectoryNode::max_pointers_in_node()) + 1;
while (m_nodes.size() < num_nodes_required)
m_nodes.append(request_new_block_index());
size_t offset = 0u;
size_t num_node = 0u;
while (offset < size()) {
HashDirectoryNode node(*this, num_node, offset);
serializer().serialize_and_write(node);
offset += node.number_of_pointers();
}
}
HashBucket* HashIndex::append_bucket(u32 index, u32 local_depth, u32 pointer)
{
m_buckets.append(make<HashBucket>(*this, index, local_depth, pointer));
return m_buckets.last();
}
HashBucket* HashIndex::get_bucket_by_index(u32 index)
{
if (index >= size())
return nullptr;
return m_buckets[index];
}
Optional<u32> HashIndex::get(Key& key)
{
auto hash = key.hash();
auto bucket_index = hash % size();
dbgln_if(SQL_DEBUG, "HashIndex::get({}) bucket_index {}", key.to_deprecated_string(), bucket_index);
auto bucket = get_bucket(bucket_index);
if constexpr (SQL_DEBUG)
bucket->list_bucket();
return bucket->get(key);
}
bool HashIndex::insert(Key const& key)
{
dbgln_if(SQL_DEBUG, "HashIndex::insert({})", key.to_deprecated_string());
auto bucket = get_bucket_for_insert(key);
bucket->insert(key);
if constexpr (SQL_DEBUG)
bucket->list_bucket();
return true;
}
HashIndexIterator HashIndex::begin()
{
return HashIndexIterator(get_bucket(0));
}
HashIndexIterator HashIndex::end()
{
return HashIndexIterator::end();
}
HashIndexIterator HashIndex::find(Key const& key)
{
auto hash = key.hash();
auto bucket_index = hash % size();
auto bucket = get_bucket(bucket_index);
auto optional_index = bucket->find_key_in_bucket(key);
if (!optional_index.has_value())
return end();
return HashIndexIterator(bucket, optional_index.value());
}
void HashIndex::list_hash()
{
warnln("Number of buckets: {} (Global depth {})", size(), global_depth());
warn("Directory pointer(s): ");
for (auto ptr : m_nodes)
warn("{}, ", ptr);
warnln();
for (auto& bucket : m_buckets)
bucket->list_bucket();
}
HashIndexIterator::HashIndexIterator(HashBucket const* bucket, size_t index)
: m_current(bucket)
, m_index(index)
{
VERIFY(!m_current || !index || (index < m_current->size()));
while (m_current && (m_current->size() == 0)) {
m_current = m_current->next_bucket();
m_index = 0;
}
}
HashIndexIterator HashIndexIterator::next()
{
if (is_end())
return *this;
if (m_index < (m_current->size() - 1))
return HashIndexIterator(m_current.ptr(), m_index + 1);
return HashIndexIterator(m_current->next_bucket());
}
HashIndexIterator HashIndexIterator::previous()
{
TODO();
}
bool HashIndexIterator::operator==(HashIndexIterator const& other) const
{
if (is_end())
return other.is_end();
if (other.is_end())
return false;
VERIFY(&other.m_current->hash_index() == &m_current->hash_index());
return (m_current.ptr() == other.m_current.ptr()) && (m_index == other.m_index);
}
bool HashIndexIterator::operator==(Key const& other) const
{
if (is_end())
return false;
if (other.is_null())
return false;
return (**this).compare(other);
}
}

View file

@ -1,177 +0,0 @@
/*
* Copyright (c) 2021, Jan de Visser <jan@de-visser.net>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#pragma once
#include <AK/WeakPtr.h>
#include <LibCore/EventReceiver.h>
#include <LibSQL/Forward.h>
#include <LibSQL/Heap.h>
#include <LibSQL/Index.h>
#include <LibSQL/Key.h>
namespace SQL {
/**
* The HashIndex class is a straightforward implementation of a persisted
* extendible hash table (see
* https://en.wikipedia.org/wiki/Extendible_hashing).
*/
class HashBucket : public IndexNode
, public Weakable<HashBucket> {
public:
HashBucket(HashIndex&, Block::Index index, u32 local_depth, Block::Index pointer);
~HashBucket() override = default;
Optional<u32> get(Key&);
bool insert(Key const&);
Vector<Key> const& entries();
Key const& operator[](size_t);
Key const& operator[](size_t ix) const;
[[nodiscard]] u32 local_depth() const { return m_local_depth; }
[[nodiscard]] u32 size() { return entries().size(); }
[[nodiscard]] size_t length() const;
[[nodiscard]] u32 size() const { return m_entries.size(); }
[[nodiscard]] Block::Index index() const { return m_index; }
void serialize(Serializer&) const;
void deserialize(Serializer&);
[[nodiscard]] HashIndex const& hash_index() const { return m_hash_index; }
[[nodiscard]] HashBucket const* next_bucket();
[[nodiscard]] HashBucket const* previous_bucket();
void list_bucket();
private:
Optional<size_t> find_key_in_bucket(Key const&);
void set_index(Block::Index index) { m_index = index; }
void set_local_depth(u32 depth) { m_local_depth = depth; }
HashIndex& m_hash_index;
u32 m_local_depth { 1 };
Block::Index m_index { 0 };
Vector<Key> m_entries;
bool m_inflated { false };
friend HashIndex;
};
class HashIndex : public Index {
C_OBJECT(HashIndex);
public:
~HashIndex() override = default;
Optional<u32> get(Key&);
bool insert(Key const&);
bool insert(Key const&& entry) { return insert(entry); }
HashIndexIterator find(Key const&);
HashIndexIterator begin();
static HashIndexIterator end();
[[nodiscard]] u32 global_depth() const { return m_global_depth; }
[[nodiscard]] u32 size() const { return 1 << m_global_depth; }
[[nodiscard]] HashBucket* get_bucket(u32);
[[nodiscard]] u32 node_pointer(u32 node_number) const { return m_nodes[node_number]; }
[[nodiscard]] u32 first_node_pointer() const { return m_nodes[0]; }
[[nodiscard]] size_t nodes() const { return m_nodes.size(); }
void list_hash();
private:
HashIndex(Serializer&, NonnullRefPtr<TupleDescriptor> const&, Block::Index);
void expand();
void write_directory();
HashBucket* append_bucket(u32 index, u32 local_depth, u32 pointer);
HashBucket* get_bucket_for_insert(Key const&);
[[nodiscard]] HashBucket* get_bucket_by_index(u32 index);
u32 m_global_depth { 1 };
Vector<u32> m_nodes;
Vector<OwnPtr<HashBucket>> m_buckets;
friend HashBucket;
friend HashDirectoryNode;
};
class HashDirectoryNode : public IndexNode {
public:
HashDirectoryNode(HashIndex&, u32, size_t);
HashDirectoryNode(HashIndex&, u32);
HashDirectoryNode(HashDirectoryNode const& other) = default;
void deserialize(Serializer&);
void serialize(Serializer&) const;
[[nodiscard]] u32 number_of_pointers() const { return min(max_pointers_in_node(), m_hash_index.size() - m_offset); }
[[nodiscard]] bool is_last() const { return m_is_last; }
static constexpr size_t max_pointers_in_node() { return (Block::DATA_SIZE - 3 * sizeof(u32)) / (2 * sizeof(u32)); }
private:
HashIndex& m_hash_index;
size_t m_node_number { 0 };
size_t m_offset { 0 };
bool m_is_last { false };
};
class HashIndexIterator {
public:
[[nodiscard]] bool is_end() const { return !m_current; }
bool operator==(HashIndexIterator const& other) const;
bool operator==(Key const& other) const;
HashIndexIterator operator++()
{
*this = next();
return *this;
}
HashIndexIterator operator++(int)
{
*this = next();
return *this;
}
HashIndexIterator operator--()
{
*this = previous();
return *this;
}
HashIndexIterator const operator--(int)
{
*this = previous();
return *this;
}
Key const& operator*() const
{
VERIFY(!is_end());
return (*m_current)[m_index];
}
Key const& operator->() const
{
VERIFY(!is_end());
return (*m_current)[m_index];
}
HashIndexIterator& operator=(HashIndexIterator const&) = default;
HashIndexIterator(HashIndexIterator const&) = default;
private:
HashIndexIterator() = default;
explicit HashIndexIterator(HashBucket const*, size_t key_index = 0);
static HashIndexIterator end() { return HashIndexIterator(); }
[[nodiscard]] HashIndexIterator next();
[[nodiscard]] HashIndexIterator previous();
[[nodiscard]] Key key() const { return **this; }
WeakPtr<HashBucket> m_current;
size_t m_index { 0 };
friend HashIndex;
};
}