ladybird/Userland/Libraries/LibSQL/Database.cpp
Jelle Raaijmakers 6601ff9d65 LibSQL: Redesign heap storage to support arbitrary amounts of data
Previously, `Heap` would store serialized data in blocks of 1024 bytes
regardless of the actual length. Data longer than 1024 bytes was
silently truncated causing database corruption.

This changes the heap storage to prefix every block with two new fields:
the total data size in bytes, and the next block to retrieve if the data
is longer than what can be stored inside a single block. By chaining
blocks together, we can store arbitrary amounts of data without needing
to change anything of the logic in the rest of LibSQL.

As part of these changes, the "free list" is also removed from the heap
awaiting an actual implementation: it was never used.

Note that this bumps the database version from 3 to 4, and as such
invalidates (deletes) any database opened with LibSQL that is not
version 4.
2023-04-23 18:08:17 -04:00

256 lines
7.7 KiB
C++

/*
* Copyright (c) 2021, Jan de Visser <jan@de-visser.net>
* Copyright (c) 2021, Mahmoud Mandour <ma.mandourr@gmail.com>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <AK/DeprecatedString.h>
#include <AK/NonnullRefPtr.h>
#include <LibSQL/BTree.h>
#include <LibSQL/Database.h>
#include <LibSQL/Heap.h>
#include <LibSQL/Meta.h>
#include <LibSQL/Row.h>
#include <LibSQL/Tuple.h>
namespace SQL {
Database::Database(DeprecatedString name)
: m_heap(Heap::construct(move(name)))
, m_serializer(m_heap)
{
}
ResultOr<void> Database::open()
{
TRY(m_heap->open());
m_schemas = BTree::construct(m_serializer, SchemaDef::index_def()->to_tuple_descriptor(), m_heap->schemas_root());
m_schemas->on_new_root = [&]() {
m_heap->set_schemas_root(m_schemas->root());
};
m_tables = BTree::construct(m_serializer, TableDef::index_def()->to_tuple_descriptor(), m_heap->tables_root());
m_tables->on_new_root = [&]() {
m_heap->set_tables_root(m_tables->root());
};
m_table_columns = BTree::construct(m_serializer, ColumnDef::index_def()->to_tuple_descriptor(), m_heap->table_columns_root());
m_table_columns->on_new_root = [&]() {
m_heap->set_table_columns_root(m_table_columns->root());
};
m_open = true;
auto ensure_schema_exists = [&](auto schema_name) -> ResultOr<NonnullRefPtr<SchemaDef>> {
if (auto result = get_schema(schema_name); result.is_error()) {
if (result.error().error() != SQLErrorCode::SchemaDoesNotExist)
return result.release_error();
auto schema_def = SchemaDef::construct(schema_name);
TRY(add_schema(*schema_def));
return schema_def;
} else {
return result.release_value();
}
};
(void)TRY(ensure_schema_exists("default"sv));
auto master_schema = TRY(ensure_schema_exists("master"sv));
if (auto result = get_table("master"sv, "internal_describe_table"sv); result.is_error()) {
if (result.error().error() != SQLErrorCode::TableDoesNotExist)
return result.release_error();
auto internal_describe_table = TableDef::construct(master_schema, "internal_describe_table");
internal_describe_table->append_column("Name", SQLType::Text);
internal_describe_table->append_column("Type", SQLType::Text);
TRY(add_table(*internal_describe_table));
}
return {};
}
Database::~Database() = default;
ErrorOr<void> Database::commit()
{
VERIFY(is_open());
TRY(m_heap->flush());
return {};
}
ResultOr<void> Database::add_schema(SchemaDef const& schema)
{
VERIFY(is_open());
if (!m_schemas->insert(schema.key()))
return Result { SQLCommand::Unknown, SQLErrorCode::SchemaExists, schema.name() };
return {};
}
Key Database::get_schema_key(DeprecatedString const& schema_name)
{
auto key = SchemaDef::make_key();
key["schema_name"] = schema_name;
return key;
}
ResultOr<NonnullRefPtr<SchemaDef>> Database::get_schema(DeprecatedString const& schema)
{
VERIFY(is_open());
auto schema_name = schema;
if (schema.is_empty())
schema_name = "default"sv;
Key key = get_schema_key(schema_name);
if (auto it = m_schema_cache.find(key.hash()); it != m_schema_cache.end())
return it->value;
auto schema_iterator = m_schemas->find(key);
if (schema_iterator.is_end() || (*schema_iterator != key))
return Result { SQLCommand::Unknown, SQLErrorCode::SchemaDoesNotExist, schema_name };
auto schema_def = SchemaDef::construct(*schema_iterator);
m_schema_cache.set(key.hash(), schema_def);
return schema_def;
}
ResultOr<void> Database::add_table(TableDef& table)
{
VERIFY(is_open());
if (!m_tables->insert(table.key()))
return Result { SQLCommand::Unknown, SQLErrorCode::TableExists, table.name() };
for (auto& column : table.columns()) {
if (!m_table_columns->insert(column->key()))
VERIFY_NOT_REACHED();
}
return {};
}
Key Database::get_table_key(DeprecatedString const& schema_name, DeprecatedString const& table_name)
{
auto key = TableDef::make_key(get_schema_key(schema_name));
key["table_name"] = table_name;
return key;
}
ResultOr<NonnullRefPtr<TableDef>> Database::get_table(DeprecatedString const& schema, DeprecatedString const& name)
{
VERIFY(is_open());
auto schema_name = schema;
if (schema.is_empty())
schema_name = "default"sv;
Key key = get_table_key(schema_name, name);
if (auto it = m_table_cache.find(key.hash()); it != m_table_cache.end())
return it->value;
auto table_iterator = m_tables->find(key);
if (table_iterator.is_end() || (*table_iterator != key))
return Result { SQLCommand::Unknown, SQLErrorCode::TableDoesNotExist, DeprecatedString::formatted("{}.{}", schema_name, name) };
auto schema_def = TRY(get_schema(schema));
auto table_def = TableDef::construct(schema_def, name);
table_def->set_pointer((*table_iterator).pointer());
m_table_cache.set(key.hash(), table_def);
auto table_hash = table_def->hash();
auto column_key = ColumnDef::make_key(table_def);
for (auto it = m_table_columns->find(column_key); !it.is_end() && ((*it)["table_hash"].to_int<u32>() == table_hash); ++it)
table_def->append_column(*it);
return table_def;
}
ErrorOr<Vector<Row>> Database::select_all(TableDef& table)
{
VERIFY(m_table_cache.get(table.key().hash()).has_value());
Vector<Row> ret;
for (auto pointer = table.pointer(); pointer; pointer = ret.last().next_pointer())
ret.append(m_serializer.deserialize_block<Row>(pointer, table, pointer));
return ret;
}
ErrorOr<Vector<Row>> Database::match(TableDef& table, Key const& key)
{
VERIFY(m_table_cache.get(table.key().hash()).has_value());
Vector<Row> ret;
// TODO Match key against indexes defined on table. If found,
// use the index instead of scanning the table.
for (auto pointer = table.pointer(); pointer;) {
auto row = m_serializer.deserialize_block<Row>(pointer, table, pointer);
if (row.match(key))
ret.append(row);
pointer = ret.last().next_pointer();
}
return ret;
}
ErrorOr<void> Database::insert(Row& row)
{
VERIFY(m_table_cache.get(row.table().key().hash()).has_value());
// TODO: implement table constraints such as unique, foreign key, etc.
row.set_pointer(m_heap->request_new_block_index());
row.set_next_pointer(row.table().pointer());
TRY(update(row));
// TODO update indexes defined on table.
auto table_key = row.table().key();
table_key.set_pointer(row.pointer());
VERIFY(m_tables->update_key_pointer(table_key));
row.table().set_pointer(row.pointer());
return {};
}
ErrorOr<void> Database::remove(Row& row)
{
auto& table = row.table();
VERIFY(m_table_cache.get(table.key().hash()).has_value());
if (table.pointer() == row.pointer()) {
auto table_key = table.key();
table_key.set_pointer(row.next_pointer());
m_tables->update_key_pointer(table_key);
table.set_pointer(row.next_pointer());
return {};
}
for (auto pointer = table.pointer(); pointer;) {
auto current = m_serializer.deserialize_block<Row>(pointer, table, pointer);
if (current.next_pointer() == row.pointer()) {
current.set_next_pointer(row.next_pointer());
TRY(update(current));
break;
}
pointer = current.next_pointer();
}
return {};
}
ErrorOr<void> Database::update(Row& tuple)
{
VERIFY(m_table_cache.get(tuple.table().key().hash()).has_value());
// TODO: implement table constraints such as unique, foreign key, etc.
m_serializer.reset();
m_serializer.serialize_and_write<Tuple>(tuple);
// TODO update indexes defined on table.
return {};
}
}