database/doris
2
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity
Files
6746434770dd3108be2dff60b64bedf5f7560915
doris/be/src/olap/schema_change.cpp

2416 lines
107 KiB
C++
Raw Blame History

// Licensed to the Apache Software Foundation (ASF) under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you 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.
#include "olap/schema_change.h"
#include "common/status.h"
#include "gutil/integral_types.h"
#include "olap/merger.h"
#include "olap/olap_common.h"
#include "olap/row.h"
#include "olap/row_block.h"
#include "olap/row_cursor.h"
#include "olap/rowset/segment_v2/column_reader.h"
#include "olap/storage_engine.h"
#include "olap/tablet.h"
#include "olap/tablet_schema.h"
#include "olap/types.h"
#include "olap/wrapper_field.h"
#include "runtime/memory/mem_tracker.h"
#include "util/defer_op.h"
#include "vec/aggregate_functions/aggregate_function.h"
#include "vec/aggregate_functions/aggregate_function_reader.h"
#include "vec/aggregate_functions/aggregate_function_simple_factory.h"
#include "vec/columns/column.h"
#include "vec/core/block.h"
#include "vec/exprs/vexpr.h"
#include "vec/exprs/vexpr_context.h"
using std::nothrow;
namespace doris {
constexpr int ALTER_TABLE_BATCH_SIZE = 4096;
class RowBlockSorter {
public:
explicit RowBlockSorter(RowBlockAllocator* allocator);
virtual ~RowBlockSorter();
size_t num_rows() { return _swap_row_block != nullptr ? _swap_row_block->capacity() : 0; }
bool sort(RowBlock** row_block);
private:
static bool _row_cursor_comparator(const std::unique_ptr<RowCursor>& a,
const std::unique_ptr<RowCursor>& b) {
return compare_row(*a, *b) < 0;
}
RowBlockAllocator* _row_block_allocator;
RowBlock* _swap_row_block;
};
class RowBlockMerger {
public:
explicit RowBlockMerger(TabletSharedPtr tablet);
virtual ~RowBlockMerger();
bool merge(const std::vector<RowBlock*>& row_block_arr, RowsetWriter* rowset_writer,
uint64_t* merged_rows);
private:
struct MergeElement {
bool operator<(const MergeElement& other) const {
return compare_row(*row_cursor, *other.row_cursor) > 0;
}
const RowBlock* row_block;
RowCursor* row_cursor;
uint32_t row_block_index;
};
bool _make_heap(const std::vector<RowBlock*>& row_block_arr);
void _pop_heap();
TabletSharedPtr _tablet;
std::priority_queue<MergeElement> _heap;
};
class MultiBlockMerger {
public:
MultiBlockMerger(TabletSharedPtr tablet) : _tablet(tablet), _cmp(tablet) {}
Status merge(const std::vector<std::unique_ptr<vectorized::Block>>& blocks,
RowsetWriter* rowset_writer, uint64_t* merged_rows) {
int rows = 0;
for (auto& block : blocks) {
rows += block->rows();
}
if (!rows) {
return Status::OK();
}
std::vector<RowRef> row_refs;
row_refs.reserve(rows);
for (auto& block : blocks) {
for (uint16_t i = 0; i < block->rows(); i++) {
row_refs.emplace_back(block.get(), i);
}
}
// TODO: try to use pdqsort to replace std::sort
// The block version is incremental.
std::stable_sort(row_refs.begin(), row_refs.end(), _cmp);
auto finalized_block = _tablet->tablet_schema()->create_block();
int columns = finalized_block.columns();
*merged_rows += rows;
if (_tablet->keys_type() == KeysType::AGG_KEYS) {
auto tablet_schema = _tablet->tablet_schema();
int key_number = _tablet->num_key_columns();
std::vector<vectorized::AggregateFunctionPtr> agg_functions;
std::vector<vectorized::AggregateDataPtr> agg_places;
for (int i = key_number; i < columns; i++) {
vectorized::AggregateFunctionPtr function =
tablet_schema->column(i).get_aggregate_function(
{finalized_block.get_data_type(i)}, vectorized::AGG_LOAD_SUFFIX);
agg_functions.push_back(function);
// create aggregate data
vectorized::AggregateDataPtr place = new char[function->size_of_data()];
function->create(place);
agg_places.push_back(place);
}
for (int i = 0; i < rows; i++) {
auto row_ref = row_refs[i];
for (int j = key_number; j < columns; j++) {
auto column_ptr = row_ref.get_column(j).get();
agg_functions[j - key_number]->add(
agg_places[j - key_number],
const_cast<const vectorized::IColumn**>(&column_ptr), row_ref.position,
nullptr);
}
if (i == rows - 1 || _cmp.compare(row_refs[i], row_refs[i + 1])) {
for (int j = 0; j < key_number; j++) {
finalized_block.get_by_position(j).column->assume_mutable()->insert_from(
*row_ref.get_column(j), row_ref.position);
}
for (int j = key_number; j < columns; j++) {
agg_functions[j - key_number]->insert_result_into(
agg_places[j - key_number],
finalized_block.get_by_position(j).column->assume_mutable_ref());
agg_functions[j - key_number]->create(agg_places[j - key_number]);
}
if (i == rows - 1 || finalized_block.rows() == ALTER_TABLE_BATCH_SIZE) {
*merged_rows -= finalized_block.rows();
rowset_writer->add_block(&finalized_block);
finalized_block.clear_column_data();
}
}
}
for (int i = 0; i < columns - key_number; i++) {
agg_functions[i]->destroy(agg_places[i]);
delete[] agg_places[i];
}
} else {
std::vector<RowRef> pushed_row_refs;
if (_tablet->keys_type() == KeysType::DUP_KEYS) {
std::swap(pushed_row_refs, row_refs);
} else if (_tablet->keys_type() == KeysType::UNIQUE_KEYS) {
for (int i = 0; i < rows; i++) {
if (i == rows - 1 || _cmp.compare(row_refs[i], row_refs[i + 1])) {
pushed_row_refs.push_back(row_refs[i]);
}
}
}
// update real inserted row number
rows = pushed_row_refs.size();
*merged_rows -= rows;
for (int i = 0; i < rows; i += ALTER_TABLE_BATCH_SIZE) {
int limit = std::min(ALTER_TABLE_BATCH_SIZE, rows - i);
for (int idx = 0; idx < columns; idx++) {
auto column = finalized_block.get_by_position(idx).column->assume_mutable();
for (int j = 0; j < limit; j++) {
auto row_ref = pushed_row_refs[i + j];
column->insert_from(*row_ref.get_column(idx), row_ref.position);
}
}
rowset_writer->add_block(&finalized_block);
finalized_block.clear_column_data();
}
}
RETURN_IF_ERROR(rowset_writer->flush());
return Status::OK();
}
private:
struct RowRef {
RowRef(vectorized::Block* block_, uint16_t position_)
: block(block_), position(position_) {}
vectorized::ColumnPtr get_column(int index) const {
return block->get_by_position(index).column;
}
const vectorized::Block* block;
uint16_t position;
};
struct RowRefComparator {
RowRefComparator(TabletSharedPtr tablet) : _num_columns(tablet->num_key_columns()) {}
int compare(const RowRef& lhs, const RowRef& rhs) const {
return lhs.block->compare_at(lhs.position, rhs.position, _num_columns, *rhs.block, -1);
}
bool operator()(const RowRef& lhs, const RowRef& rhs) const {
return compare(lhs, rhs) < 0;
}
const size_t _num_columns;
};
TabletSharedPtr _tablet;
RowRefComparator _cmp;
};
RowBlockChanger::RowBlockChanger(TabletSchemaSPtr tablet_schema,
const DeleteHandler* delete_handler, DescriptorTbl desc_tbl)
: _desc_tbl(desc_tbl) {
_schema_mapping.resize(tablet_schema->num_columns());
_delete_handler = delete_handler;
}
RowBlockChanger::~RowBlockChanger() {
for (auto it = _schema_mapping.begin(); it != _schema_mapping.end(); ++it) {
SAFE_DELETE(it->default_value);
}
_schema_mapping.clear();
}
ColumnMapping* RowBlockChanger::get_mutable_column_mapping(size_t column_index) {
if (column_index >= _schema_mapping.size()) {
return nullptr;
}
return &(_schema_mapping[column_index]);
}
#define TYPE_REINTERPRET_CAST(FromType, ToType) \
{ \
size_t row_num = ref_block->row_block_info().row_num; \
for (size_t row = 0, mutable_row = 0; row < row_num; ++row) { \
char* ref_ptr = ref_block->field_ptr(row, ref_column); \
char* new_ptr = mutable_block->field_ptr(mutable_row++, i); \
*new_ptr = *ref_ptr; \
*(ToType*)(new_ptr + 1) = *(FromType*)(ref_ptr + 1); \
} \
break; \
}
#define LARGEINT_REINTERPRET_CAST(FromType, ToType) \
{ \
size_t row_num = ref_block->row_block_info().row_num; \
for (size_t row = 0, mutable_row = 0; row < row_num; ++row) { \
char* ref_ptr = ref_block->field_ptr(row, ref_column); \
char* new_ptr = mutable_block->field_ptr(mutable_row++, i); \
*new_ptr = *ref_ptr; \
ToType new_value = *(FromType*)(ref_ptr + 1); \
memcpy(new_ptr + 1, &new_value, sizeof(ToType)); \
} \
break; \
}
#define CONVERT_FROM_TYPE(from_type) \
{ \
switch (newtype) { \
case OLAP_FIELD_TYPE_TINYINT: \
TYPE_REINTERPRET_CAST(from_type, int8_t); \
case OLAP_FIELD_TYPE_UNSIGNED_TINYINT: \
TYPE_REINTERPRET_CAST(from_type, uint8_t); \
case OLAP_FIELD_TYPE_SMALLINT: \
TYPE_REINTERPRET_CAST(from_type, int16_t); \
case OLAP_FIELD_TYPE_UNSIGNED_SMALLINT: \
TYPE_REINTERPRET_CAST(from_type, uint16_t); \
case OLAP_FIELD_TYPE_INT: \
TYPE_REINTERPRET_CAST(from_type, int32_t); \
case OLAP_FIELD_TYPE_UNSIGNED_INT: \
TYPE_REINTERPRET_CAST(from_type, uint32_t); \
case OLAP_FIELD_TYPE_BIGINT: \
TYPE_REINTERPRET_CAST(from_type, int64_t); \
case OLAP_FIELD_TYPE_UNSIGNED_BIGINT: \
TYPE_REINTERPRET_CAST(from_type, uint64_t); \
case OLAP_FIELD_TYPE_LARGEINT: \
LARGEINT_REINTERPRET_CAST(from_type, int128_t); \
case OLAP_FIELD_TYPE_FLOAT: \
TYPE_REINTERPRET_CAST(from_type, float); \
case OLAP_FIELD_TYPE_DOUBLE: \
TYPE_REINTERPRET_CAST(from_type, double); \
default: \
LOG(WARNING) << "the column type which was altered to was unsupported." \
<< " origin_type=" \
<< ref_block->tablet_schema()->column(ref_column).type() \
<< ", alter_type=" << mutable_block->tablet_schema()->column(i).type(); \
return Status::OLAPInternalError(OLAP_ERR_SCHEMA_CHANGE_INFO_INVALID); \
} \
break; \
}
#define ASSIGN_DEFAULT_VALUE(length) \
case length: { \
for (size_t row = 0; row < ref_block.row_block_info().row_num; ++row) { \
memcpy(buf, _schema_mapping[i].default_value->ptr(), length); \
buf += length; \
} \
break; \
}
struct ConvertTypeMapHash {
size_t operator()(const std::pair<FieldType, FieldType>& pair) const {
return (pair.first + 31) ^ pair.second;
}
};
class ConvertTypeResolver {
DECLARE_SINGLETON(ConvertTypeResolver);
public:
bool get_convert_type_info(const FieldType from_type, const FieldType to_type) const {
return _convert_type_set.find(std::make_pair(from_type, to_type)) !=
_convert_type_set.end();
}
template <FieldType from_type, FieldType to_type>
void add_convert_type_mapping() {
_convert_type_set.emplace(std::make_pair(from_type, to_type));
}
private:
using convert_type_pair = std::pair<FieldType, FieldType>;
std::unordered_set<convert_type_pair, ConvertTypeMapHash> _convert_type_set;
DISALLOW_COPY_AND_ASSIGN(ConvertTypeResolver);
};
ConvertTypeResolver::ConvertTypeResolver() {
// from char type
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_TINYINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_SMALLINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_INT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_BIGINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_LARGEINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_FLOAT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_DOUBLE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_DATE>();
// supported type convert should annotate in doc:
// http://doris.apache.org/master/zh-CN/sql-reference/sql-statements/Data%20Definition/ALTER%20TABLE.html#description
// If type convert is supported here, you should check fe/src/main/java/org/apache/doris/catalog/ColumnType.java to supported it either
// from varchar type
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_TINYINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_SMALLINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_INT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_BIGINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_LARGEINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_FLOAT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_DOUBLE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_DATE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_VARCHAR, OLAP_FIELD_TYPE_STRING>();
// to varchar type
add_convert_type_mapping<OLAP_FIELD_TYPE_TINYINT, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_SMALLINT, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_INT, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_BIGINT, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_LARGEINT, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_FLOAT, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_DOUBLE, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_DECIMAL, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_VARCHAR>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_VARCHAR>();
// from string
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_TINYINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_SMALLINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_INT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_BIGINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_LARGEINT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_FLOAT>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_DOUBLE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_DATE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_VARCHAR>();
// to string
add_convert_type_mapping<OLAP_FIELD_TYPE_TINYINT, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_SMALLINT, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_INT, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_BIGINT, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_LARGEINT, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_FLOAT, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_DOUBLE, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_DECIMAL, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_CHAR, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_STRING, OLAP_FIELD_TYPE_STRING>();
add_convert_type_mapping<OLAP_FIELD_TYPE_DATE, OLAP_FIELD_TYPE_DATETIME>();
add_convert_type_mapping<OLAP_FIELD_TYPE_DATETIME, OLAP_FIELD_TYPE_DATE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_FLOAT, OLAP_FIELD_TYPE_DOUBLE>();
add_convert_type_mapping<OLAP_FIELD_TYPE_INT, OLAP_FIELD_TYPE_DATE>();
}
ConvertTypeResolver::~ConvertTypeResolver() = default;
bool to_bitmap(RowCursor* read_helper, RowCursor* write_helper, const TabletColumn& ref_column,
int field_idx, int ref_field_idx, MemPool* mem_pool) {
write_helper->set_not_null(field_idx);
BitmapValue bitmap;
if (!read_helper->is_null(ref_field_idx)) {
uint64_t origin_value;
char* src = read_helper->cell_ptr(ref_field_idx);
switch (ref_column.type()) {
case OLAP_FIELD_TYPE_TINYINT:
if (*(int8_t*)src < 0) {
LOG(WARNING) << "The input: " << *(int8_t*)src
<< " is not valid, to_bitmap only support bigint value from 0 to "
"18446744073709551615 currently";
return false;
}
origin_value = *(int8_t*)src;
break;
case OLAP_FIELD_TYPE_UNSIGNED_TINYINT:
origin_value = *(uint8_t*)src;
break;
case OLAP_FIELD_TYPE_SMALLINT:
if (*(int16_t*)src < 0) {
LOG(WARNING) << "The input: " << *(int16_t*)src
<< " is not valid, to_bitmap only support bigint value from 0 to "
"18446744073709551615 currently";
return false;
}
origin_value = *(int16_t*)src;
break;
case OLAP_FIELD_TYPE_UNSIGNED_SMALLINT:
origin_value = *(uint16_t*)src;
break;
case OLAP_FIELD_TYPE_INT:
if (*(int32_t*)src < 0) {
LOG(WARNING) << "The input: " << *(int32_t*)src
<< " is not valid, to_bitmap only support bigint value from 0 to "
"18446744073709551615 currently";
return false;
}
origin_value = *(int32_t*)src;
break;
case OLAP_FIELD_TYPE_UNSIGNED_INT:
origin_value = *(uint32_t*)src;
break;
case OLAP_FIELD_TYPE_BIGINT:
if (*(int64_t*)src < 0) {
LOG(WARNING) << "The input: " << *(int64_t*)src
<< " is not valid, to_bitmap only support bigint value from 0 to "
"18446744073709551615 currently";
return false;
}
origin_value = *(int64_t*)src;
break;
case OLAP_FIELD_TYPE_UNSIGNED_BIGINT:
origin_value = *(uint64_t*)src;
break;
default:
LOG(WARNING) << "the column type which was altered from was unsupported."
<< " from_type=" << ref_column.type();
return false;
}
bitmap.add(origin_value);
}
char* buf = reinterpret_cast<char*>(mem_pool->allocate(bitmap.getSizeInBytes()));
Slice dst(buf, bitmap.getSizeInBytes());
bitmap.write(dst.data);
write_helper->set_field_content(field_idx, reinterpret_cast<char*>(&dst), mem_pool);
return true;
}
bool hll_hash(RowCursor* read_helper, RowCursor* write_helper, const TabletColumn& ref_column,
int field_idx, int ref_field_idx, MemPool* mem_pool) {
write_helper->set_not_null(field_idx);
HyperLogLog hll;
if (!read_helper->is_null(ref_field_idx)) {
uint64_t hash_value;
switch (ref_column.type()) {
case OLAP_FIELD_TYPE_CHAR: {
int p = ref_column.length() - 1;
Slice* slice = reinterpret_cast<Slice*>(read_helper->cell_ptr(ref_field_idx));
char* buf = slice->data;
while (p >= 0 && buf[p] == '\0') {
p--;
}
slice->size = p + 1;
}
case OLAP_FIELD_TYPE_VARCHAR:
case OLAP_FIELD_TYPE_STRING: {
Slice slice = *reinterpret_cast<Slice*>(read_helper->cell_ptr(ref_field_idx));
hash_value = HashUtil::murmur_hash64A(slice.data, slice.size, HashUtil::MURMUR_SEED);
break;
}
case OLAP_FIELD_TYPE_BOOL:
case OLAP_FIELD_TYPE_TINYINT:
case OLAP_FIELD_TYPE_UNSIGNED_TINYINT:
case OLAP_FIELD_TYPE_SMALLINT:
case OLAP_FIELD_TYPE_UNSIGNED_SMALLINT:
case OLAP_FIELD_TYPE_INT:
case OLAP_FIELD_TYPE_UNSIGNED_INT:
case OLAP_FIELD_TYPE_BIGINT:
case OLAP_FIELD_TYPE_UNSIGNED_BIGINT:
case OLAP_FIELD_TYPE_LARGEINT:
case OLAP_FIELD_TYPE_FLOAT:
case OLAP_FIELD_TYPE_DOUBLE:
case OLAP_FIELD_TYPE_DISCRETE_DOUBLE:
case OLAP_FIELD_TYPE_DATE:
case OLAP_FIELD_TYPE_DATETIME: {
std::string ref_column_string =
read_helper->column_schema(ref_field_idx)
->type_info()
->to_string(read_helper->cell_ptr(ref_field_idx));
hash_value = HashUtil::murmur_hash64A(
ref_column_string.c_str(), ref_column_string.length(), HashUtil::MURMUR_SEED);
break;
}
default:
LOG(WARNING) << "fail to hll hash type : " << ref_column.type();
return false;
}
hll.update(hash_value);
}
std::string buf;
buf.resize(hll.max_serialized_size());
buf.resize(hll.serialize((uint8_t*)buf.c_str()));
Slice dst(buf);
write_helper->set_field_content(field_idx, reinterpret_cast<char*>(&dst), mem_pool);
return true;
}
bool count_field(RowCursor* read_helper, RowCursor* write_helper, const TabletColumn& ref_column,
int field_idx, int ref_field_idx, MemPool* mem_pool) {
write_helper->set_not_null(field_idx);
int64_t count = read_helper->is_null(ref_field_idx) ? 0 : 1;
write_helper->set_field_content(field_idx, (char*)&count, mem_pool);
return true;
}
Status RowBlockChanger::change_row_block(const RowBlock* ref_block, int32_t data_version,
RowBlock* mutable_block,
const uint64_t* filtered_rows) const {
if (mutable_block == nullptr) {
LOG(FATAL) << "mutable block is uninitialized.";
return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
} else if (mutable_block->tablet_schema()->num_columns() != _schema_mapping.size()) {
LOG(WARNING) << "mutable block does not match with schema mapping rules. "
<< "block_schema_size=" << mutable_block->tablet_schema()->num_columns()
<< ", mapping_schema_size=" << _schema_mapping.size();
return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
}
if (mutable_block->capacity() < ref_block->row_block_info().row_num) {
LOG(WARNING) << "mutable block is not large enough for storing the changed block. "
<< "mutable_block_size=" << mutable_block->capacity()
<< ", ref_block_row_num=" << ref_block->row_block_info().row_num;
return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
}
mutable_block->clear();
RowCursor write_helper;
if (write_helper.init(mutable_block->tablet_schema()) != Status::OK()) {
LOG(WARNING) << "fail to init rowcursor.";
return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
}
RowCursor read_helper;
if (read_helper.init(ref_block->tablet_schema()) != Status::OK()) {
LOG(WARNING) << "fail to init rowcursor.";
return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
}
// a.1 First determine whether the data needs to be filtered, and finally only those marked as 1 are left as needed
// For those without filter, it is equivalent to leave after setting all to 1
const uint32_t row_num = ref_block->row_block_info().row_num;
// a.2 Calculate the left row num
uint32_t new_row_num = row_num - *filtered_rows;
const bool filter_all = (new_row_num == 0);
MemPool* mem_pool = mutable_block->mem_pool();
// b. According to the previous filtering information, only processes that are also marked as 1
for (size_t i = 0, len = mutable_block->tablet_schema()->num_columns(); !filter_all && i < len;
++i) {
int32_t ref_column = _schema_mapping[i].ref_column;
if (_schema_mapping[i].ref_column >= 0) {
if (!_schema_mapping[i].materialized_function.empty()) {
bool (*_do_materialized_transform)(RowCursor*, RowCursor*, const TabletColumn&, int,
int, MemPool*) = nullptr;
if (_schema_mapping[i].materialized_function == "to_bitmap") {
_do_materialized_transform = to_bitmap;
} else if (_schema_mapping[i].materialized_function == "hll_hash") {
_do_materialized_transform = hll_hash;
} else if (_schema_mapping[i].materialized_function == "count_field") {
_do_materialized_transform = count_field;
} else {
LOG(WARNING) << "error materialized view function : "
<< _schema_mapping[i].materialized_function;
return Status::OLAPInternalError(OLAP_ERR_SCHEMA_CHANGE_INFO_INVALID);
}
VLOG_NOTICE << "_schema_mapping[" << i << "].materialized_function : "
<< _schema_mapping[i].materialized_function;
for (size_t row_index = 0, new_row_index = 0;
row_index < ref_block->row_block_info().row_num; ++row_index) {
mutable_block->get_row(new_row_index++, &write_helper);
ref_block->get_row(row_index, &read_helper);
if (!_do_materialized_transform(&read_helper, &write_helper,
ref_block->tablet_schema()->column(ref_column),
i, _schema_mapping[i].ref_column, mem_pool)) {
return Status::OLAPInternalError(OLAP_ERR_DATA_QUALITY_ERR);
}
}
continue;
}
// new column will be assigned as referenced column
// check if the type of new column is equal to the older's.
FieldType reftype = ref_block->tablet_schema()->column(ref_column).type();
FieldType newtype = mutable_block->tablet_schema()->column(i).type();
if (newtype == reftype) {
// Low efficiency, you can also directly calculate the variable length domain copy, but it will still destroy the package
for (size_t row_index = 0, new_row_index = 0;
row_index < ref_block->row_block_info().row_num; ++row_index) {
// Specify the new row index to be written (different from the read row_index)
mutable_block->get_row(new_row_index++, &write_helper);
ref_block->get_row(row_index, &read_helper);
if (read_helper.is_null(ref_column)) {
write_helper.set_null(i);
} else {
write_helper.set_not_null(i);
if (newtype == OLAP_FIELD_TYPE_CHAR) {
// if modify length of CHAR type, the size of slice should be equal
// to new length.
Slice* src = (Slice*)(read_helper.cell_ptr(ref_column));
size_t size = mutable_block->tablet_schema()->column(i).length();
char* buf = reinterpret_cast<char*>(mem_pool->allocate(size));
memset(buf, 0, size);
size_t copy_size = (size < src->size) ? size : src->size;
memcpy(buf, src->data, copy_size);
Slice dst(buf, size);
write_helper.set_field_content(i, reinterpret_cast<char*>(&dst),
mem_pool);
} else {
char* src = read_helper.cell_ptr(ref_column);
write_helper.set_field_content(i, src, mem_pool);
}
}
}
} else if (ConvertTypeResolver::instance()->get_convert_type_info(reftype, newtype)) {
for (size_t row_index = 0, new_row_index = 0;
row_index < ref_block->row_block_info().row_num; ++row_index) {
mutable_block->get_row(new_row_index++, &write_helper);
ref_block->get_row(row_index, &read_helper);
if (read_helper.is_null(ref_column)) {
write_helper.set_null(i);
} else {
write_helper.set_not_null(i);
const Field* ref_field = read_helper.column_schema(ref_column);
char* ref_value = read_helper.cell_ptr(ref_column);
Status st = write_helper.convert_from(i, ref_value, ref_field->type_info(),
mem_pool);
if (!st) {
LOG(WARNING)
<< "the column type which was altered from was unsupported."
<< "status:" << st.to_string() << ", from_type=" << reftype
<< ", to_type=" << newtype;
return st;
}
}
}
// Write column i from ref_column.
} else {
// copy and alter the field
// You can stay here for the time being, the new type does not involve type conversion for the time being
switch (reftype) {
case OLAP_FIELD_TYPE_TINYINT:
CONVERT_FROM_TYPE(int8_t);
case OLAP_FIELD_TYPE_UNSIGNED_TINYINT:
CONVERT_FROM_TYPE(uint8_t);
case OLAP_FIELD_TYPE_SMALLINT:
CONVERT_FROM_TYPE(int16_t);
case OLAP_FIELD_TYPE_UNSIGNED_SMALLINT:
CONVERT_FROM_TYPE(uint16_t);
case OLAP_FIELD_TYPE_INT:
CONVERT_FROM_TYPE(int32_t);
case OLAP_FIELD_TYPE_UNSIGNED_INT:
CONVERT_FROM_TYPE(uint32_t);
case OLAP_FIELD_TYPE_BIGINT:
CONVERT_FROM_TYPE(int64_t);
case OLAP_FIELD_TYPE_UNSIGNED_BIGINT:
CONVERT_FROM_TYPE(uint64_t);
case OLAP_FIELD_TYPE_LARGEINT:
CONVERT_FROM_TYPE(int128_t);
default:
LOG(WARNING) << "the column type which was altered from was unsupported."
<< " from_type="
<< ref_block->tablet_schema()->column(ref_column).type();
return Status::OLAPInternalError(OLAP_ERR_SCHEMA_CHANGE_INFO_INVALID);
}
if (newtype < reftype) {
VLOG_NOTICE << "type degraded while altering column. "
<< "column=" << mutable_block->tablet_schema()->column(i).name()
<< ", origin_type="
<< ref_block->tablet_schema()->column(ref_column).type()
<< ", alter_type="
<< mutable_block->tablet_schema()->column(i).type();
}
}
} else {
// New column, write default value
for (size_t row_index = 0, new_row_index = 0;
row_index < ref_block->row_block_info().row_num; ++row_index) {
mutable_block->get_row(new_row_index++, &write_helper);
if (_schema_mapping[i].default_value->is_null()) {
write_helper.set_null(i);
} else {
write_helper.set_not_null(i);
write_helper.set_field_content(i, _schema_mapping[i].default_value->ptr(),
mem_pool);
}
}
}
}
// NOTE The current row_num of mutable_block is still as much as ref
// (Actually, you can re-init into less when init, the new_row_num left by the filter)
// In split_table, there may be no data due to filtering
mutable_block->finalize(new_row_num);
return Status::OK();
}
#undef CONVERT_FROM_TYPE
#undef TYPE_REINTERPRET_CAST
#undef ASSIGN_DEFAULT_VALUE
Status RowBlockChanger::change_block(vectorized::Block* ref_block,
vectorized::Block* new_block) const {
if (new_block->columns() != _schema_mapping.size()) {
LOG(WARNING) << "block does not match with schema mapping rules. "
<< "block_schema_size=" << new_block->columns()
<< ", mapping_schema_size=" << _schema_mapping.size();
return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
}
ObjectPool pool;
RuntimeState* state = pool.add(new RuntimeState());
state->set_desc_tbl(&_desc_tbl);
RowDescriptor row_desc =
RowDescriptor(_desc_tbl.get_tuple_descriptor(_desc_tbl.get_row_tuples()[0]), false);
const int row_size = ref_block->rows();
const int column_size = new_block->columns();
// swap ref_block[key] and new_block[value]
std::map<int, int> swap_idx_map;
for (int idx = 0; idx < column_size; idx++) {
int ref_idx = _schema_mapping[idx].ref_column;
if (ref_idx < 0) {
// new column, write default value
auto value = _schema_mapping[idx].default_value;
auto column = new_block->get_by_position(idx).column->assume_mutable();
if (value->is_null()) {
DCHECK(column->is_nullable());
column->insert_many_defaults(row_size);
} else {
auto type_info = get_type_info(_schema_mapping[idx].new_column);
DefaultValueColumnIterator::insert_default_data(type_info.get(), value->size(),
value->ptr(), column, row_size);
}
} else if (_schema_mapping[idx].expr != nullptr) {
// calculate special materialized function, to_bitmap/hll_hash/count_field or cast expr
vectorized::VExprContext* ctx = nullptr;
RETURN_IF_ERROR(
vectorized::VExpr::create_expr_tree(&pool, *_schema_mapping[idx].expr, &ctx));
Defer defer {[&]() { ctx->close(state); }};
RETURN_IF_ERROR(ctx->prepare(state, row_desc));
RETURN_IF_ERROR(ctx->open(state));
int result_column_id = -1;
RETURN_IF_ERROR(ctx->execute(ref_block, &result_column_id));
DCHECK(ref_block->get_by_position(result_column_id).column->size() == row_size)
<< new_block->get_by_position(idx).name << " size invalid"
<< ", expect=" << row_size
<< ", real=" << ref_block->get_by_position(result_column_id).column->size();
if (ctx->root()->node_type() == TExprNodeType::CAST_EXPR) {
RETURN_IF_ERROR(
_check_cast_valid(ref_block->get_by_position(ref_idx).column,
ref_block->get_by_position(result_column_id).column));
}
swap_idx_map[result_column_id] = idx;
} else {
// same type, just swap column
swap_idx_map[ref_idx] = idx;
}
}
for (auto it : swap_idx_map) {
auto& ref_col = ref_block->get_by_position(it.first);
auto& new_col = new_block->get_by_position(it.second);
bool ref_col_nullable = ref_col.column->is_nullable();
bool new_col_nullable = new_col.column->is_nullable();
if (ref_col_nullable != new_col_nullable) {
// not nullable to nullable
if (new_col_nullable) {
auto* new_nullable_col = assert_cast<vectorized::ColumnNullable*>(
std::move(*new_col.column).mutate().get());
new_nullable_col->swap_nested_column(ref_col.column);
new_nullable_col->get_null_map_data().resize_fill(new_nullable_col->size());
} else {
// nullable to not nullable:
// suppose column `c_phone` is originally varchar(16) NOT NULL,
// then do schema change `alter table test modify column c_phone int not null`,
// the cast expr of schema change is `CastExpr(CAST String to Nullable(Int32))`,
// so need to handle nullable to not nullable here
auto* ref_nullable_col = assert_cast<vectorized::ColumnNullable*>(
std::move(*ref_col.column).mutate().get());
const auto* null_map = ref_nullable_col->get_null_map_column().get_data().data();
for (size_t i = 0; i < row_size; i++) {
if (null_map[i]) {
return Status::DataQualityError("is_null of data is changed!");
}
}
ref_nullable_col->swap_nested_column(new_col.column);
}
} else {
new_block->get_by_position(it.second).column.swap(
ref_block->get_by_position(it.first).column);
}
}
return Status::OK();
}
// This check is to prevent schema-change from causing data loss
Status RowBlockChanger::_check_cast_valid(vectorized::ColumnPtr ref_column,
vectorized::ColumnPtr new_column) const {
if (ref_column->is_nullable() != new_column->is_nullable()) {
if (ref_column->is_nullable()) {
return Status::DataQualityError("Can not change nullable column to not nullable");
} else {
auto* new_null_map =
vectorized::check_and_get_column<vectorized::ColumnNullable>(new_column)
->get_null_map_column()
.get_data()
.data();
bool is_changed = false;
for (size_t i = 0; i < ref_column->size(); i++) {
is_changed |= new_null_map[i];
}
if (is_changed) {
return Status::DataQualityError("is_null of data is changed!");
}
}
}
if (ref_column->is_nullable() && new_column->is_nullable()) {
auto* ref_null_map =
vectorized::check_and_get_column<vectorized::ColumnNullable>(ref_column)
->get_null_map_column()
.get_data()
.data();
auto* new_null_map =
vectorized::check_and_get_column<vectorized::ColumnNullable>(new_column)
->get_null_map_column()
.get_data()
.data();
bool is_changed = false;
for (size_t i = 0; i < ref_column->size(); i++) {
is_changed |= (ref_null_map[i] != new_null_map[i]);
}
if (is_changed) {
return Status::DataQualityError("is_null of data is changed!");
}
}
return Status::OK();
}
RowBlockSorter::RowBlockSorter(RowBlockAllocator* row_block_allocator)
: _row_block_allocator(row_block_allocator), _swap_row_block(nullptr) {}
RowBlockSorter::~RowBlockSorter() {
if (_swap_row_block) {
_row_block_allocator->release(_swap_row_block);
_swap_row_block = nullptr;
}
}
bool RowBlockSorter::sort(RowBlock** row_block) {
uint32_t row_num = (*row_block)->row_block_info().row_num;
bool null_supported = (*row_block)->row_block_info().null_supported;
if (_swap_row_block == nullptr || _swap_row_block->capacity() < row_num) {
if (_swap_row_block != nullptr) {
_row_block_allocator->release(_swap_row_block);
_swap_row_block = nullptr;
}
if (!_row_block_allocator->allocate(&_swap_row_block, row_num, null_supported)) {
LOG(WARNING) << "fail to allocate memory.";
return false;
}
}
RowCursor helper_row;
auto res = helper_row.init(_swap_row_block->tablet_schema());
if (!res) {
LOG(WARNING) << "row cursor init failed.res:" << res;
return false;
}
std::vector<std::unique_ptr<RowCursor>> row_cursor_list;
row_cursor_list.reserve((*row_block)->row_block_info().row_num);
// create an list of row cursor as long as the number of rows in data block.
for (size_t i = 0; i < (*row_block)->row_block_info().row_num; ++i) {
row_cursor_list.emplace_back(new (nothrow) RowCursor());
if (row_cursor_list[i] == nullptr) {
LOG(WARNING) << "failed to malloc RowCursor. size=" << sizeof(RowCursor);
return false;
}
if (row_cursor_list[i]->init((*row_block)->tablet_schema()) != Status::OK()) {
return false;
}
(*row_block)->get_row(i, row_cursor_list[i].get());
}
// Must use 'std::' because this class has a function whose name is sort too
std::stable_sort(row_cursor_list.begin(), row_cursor_list.end(), _row_cursor_comparator);
// copy the results sorted to temp row block.
_swap_row_block->clear();
for (size_t i = 0; i < row_cursor_list.size(); ++i) {
_swap_row_block->get_row(i, &helper_row);
copy_row(&helper_row, *row_cursor_list[i], _swap_row_block->mem_pool());
}
_swap_row_block->finalize(row_cursor_list.size());
// swap the row block for reducing memory allocating.
std::swap(*row_block, _swap_row_block);
return true;
}
RowBlockAllocator::RowBlockAllocator(TabletSchemaSPtr tablet_schema, size_t memory_limitation)
: _tablet_schema(tablet_schema),
_tracker(std::make_unique<MemTracker>("RowBlockAllocator")),
_row_len(tablet_schema->row_size()),
_memory_limitation(memory_limitation) {
VLOG_NOTICE << "RowBlockAllocator(). row_len=" << _row_len;
}
RowBlockAllocator::~RowBlockAllocator() {
if (_tracker->consumption() != 0) {
LOG(WARNING) << "memory lost in RowBlockAllocator. memory_size=" << _tracker->consumption();
}
}
Status RowBlockAllocator::allocate(RowBlock** row_block, size_t num_rows, bool null_supported) {
size_t row_block_size = _row_len * num_rows;
if (_memory_limitation > 0 && _tracker->consumption() + row_block_size > _memory_limitation) {
*row_block = nullptr;
return Status::OLAPInternalError(OLAP_ERR_FETCH_MEMORY_EXCEEDED);
}
// TODO(lijiao) : Why abandon the original m_row_block_buffer
*row_block = new (nothrow) RowBlock(_tablet_schema);
if (*row_block == nullptr) {
LOG(WARNING) << "failed to malloc RowBlock. size=" << sizeof(RowBlock);
return Status::OLAPInternalError(OLAP_ERR_MALLOC_ERROR);
}
RowBlockInfo row_block_info(0U, num_rows);
row_block_info.null_supported = null_supported;
(*row_block)->init(row_block_info);
_tracker->consume(row_block_size);
VLOG_NOTICE << "RowBlockAllocator::allocate() this=" << this << ", num_rows=" << num_rows
<< ", m_memory_allocated=" << _tracker->consumption()
<< ", row_block_addr=" << *row_block;
return Status::OK();
}
void RowBlockAllocator::release(RowBlock* row_block) {
if (row_block == nullptr) {
LOG(INFO) << "null row block released.";
return;
}
_tracker->release(row_block->capacity() * _row_len);
VLOG_NOTICE << "RowBlockAllocator::release() this=" << this
<< ", num_rows=" << row_block->capacity()
<< ", m_memory_allocated=" << _tracker->consumption()
<< ", row_block_addr=" << row_block;
delete row_block;
}
bool RowBlockAllocator::is_memory_enough_for_sorting(size_t num_rows, size_t allocated_rows) {
if (num_rows <= allocated_rows) {
return true;
}
size_t row_block_size = _row_len * (num_rows - allocated_rows);
return _tracker->consumption() + row_block_size < _memory_limitation;
}
RowBlockMerger::RowBlockMerger(TabletSharedPtr tablet) : _tablet(tablet) {}
RowBlockMerger::~RowBlockMerger() = default;
bool RowBlockMerger::merge(const std::vector<RowBlock*>& row_block_arr, RowsetWriter* rowset_writer,
uint64_t* merged_rows) {
uint64_t tmp_merged_rows = 0;
RowCursor row_cursor;
std::unique_ptr<MemPool> mem_pool(new MemPool());
std::unique_ptr<ObjectPool> agg_object_pool(new ObjectPool());
auto merge_error = [&]() -> bool {
while (!_heap.empty()) {
MergeElement element = _heap.top();
_heap.pop();
SAFE_DELETE(element.row_cursor);
}
return false;
};
if (row_cursor.init(_tablet->tablet_schema()) != Status::OK()) {
LOG(WARNING) << "fail to init row cursor.";
return merge_error();
}
if (!_make_heap(row_block_arr)) {
// There is error log in _make_heap, so no need to more log.
return merge_error();
}
row_cursor.allocate_memory_for_string_type(_tablet->tablet_schema());
while (_heap.size() > 0) {
init_row_with_others(&row_cursor, *(_heap.top().row_cursor), mem_pool.get(),
agg_object_pool.get());
_pop_heap();
if (KeysType::DUP_KEYS == _tablet->keys_type()) {
if (rowset_writer->add_row(row_cursor) != Status::OK()) {
LOG(WARNING) << "fail to add row to rowset writer.";
return merge_error();
}
continue;
}
while (!_heap.empty() && compare_row(row_cursor, *_heap.top().row_cursor) == 0) {
// TODO(zc): Currently we keep nullptr to indicate that this is a query path,
// we should fix this trick ASAP
agg_update_row(&row_cursor, *(_heap.top().row_cursor), nullptr);
++tmp_merged_rows;
_pop_heap();
}
agg_finalize_row(&row_cursor, mem_pool.get());
if (rowset_writer->add_row(row_cursor) != Status::OK()) {
LOG(WARNING) << "fail to add row to rowset writer.";
return merge_error();
}
// the memory allocate by mem pool has been copied,
// so we should release memory immediately
mem_pool->clear();
agg_object_pool.reset(new ObjectPool());
}
if (rowset_writer->flush() != Status::OK()) {
LOG(WARNING) << "failed to finalizing writer.";
return merge_error();
}
*merged_rows = tmp_merged_rows;
return true;
}
bool RowBlockMerger::_make_heap(const std::vector<RowBlock*>& row_block_arr) {
for (auto row_block : row_block_arr) {
MergeElement element;
element.row_block = row_block;
element.row_block_index = 0;
element.row_cursor = new (nothrow) RowCursor();
if (element.row_cursor == nullptr) {
LOG(FATAL) << "failed to malloc RowCursor. size=" << sizeof(RowCursor);
return false;
}
if (element.row_cursor->init(element.row_block->tablet_schema()) != Status::OK()) {
LOG(WARNING) << "failed to init row cursor.";
SAFE_DELETE(element.row_cursor);
return false;
}
element.row_block->get_row(element.row_block_index, element.row_cursor);
_heap.push(element);
}
return true;
}
void RowBlockMerger::_pop_heap() {
MergeElement element = _heap.top();
_heap.pop();
if (++element.row_block_index >= element.row_block->row_block_info().row_num) {
SAFE_DELETE(element.row_cursor);
return;
}
element.row_block->get_row(element.row_block_index, element.row_cursor);
_heap.push(element);
}
Status LinkedSchemaChange::process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer,
TabletSharedPtr new_tablet, TabletSharedPtr base_tablet,
TabletSchemaSPtr base_tablet_schema) {
// In some cases, there may be more than one type of rowset in a tablet,
// in which case the conversion cannot be done directly by linked schema change,
// but requires direct schema change to rewrite the data.
if (rowset_reader->type() != rowset_writer->type()) {
LOG(INFO) << "the type of rowset " << rowset_reader->rowset()->rowset_id()
<< " in base tablet is not same as type " << rowset_writer->type()
<< ", use direct schema change.";
return SchemaChangeHandler::get_sc_procedure(_row_block_changer, false, true)
->process(rowset_reader, rowset_writer, new_tablet, base_tablet,
base_tablet_schema);
} else {
Status status = rowset_writer->add_rowset_for_linked_schema_change(rowset_reader->rowset());
if (!status) {
LOG(WARNING) << "fail to convert rowset."
<< ", new_tablet=" << new_tablet->full_name()
<< ", version=" << rowset_writer->version().first << "-"
<< rowset_writer->version().second << ", error status " << status;
return status;
}
// copy delete bitmap to new tablet.
if (new_tablet->keys_type() == UNIQUE_KEYS &&
new_tablet->enable_unique_key_merge_on_write()) {
DeleteBitmap origin_delete_bitmap(base_tablet->tablet_id());
base_tablet->tablet_meta()->delete_bitmap().subset(
{rowset_reader->rowset()->rowset_id(), 0, 0},
{rowset_reader->rowset()->rowset_id(), UINT32_MAX, INT64_MAX},
&origin_delete_bitmap);
for (auto iter = origin_delete_bitmap.delete_bitmap.begin();
iter != origin_delete_bitmap.delete_bitmap.end(); ++iter) {
int ret = new_tablet->tablet_meta()->delete_bitmap().set(
{rowset_writer->rowset_id(), std::get<1>(iter->first),
std::get<2>(iter->first)},
iter->second);
DCHECK(ret == 1);
}
}
return Status::OK();
}
}
SchemaChangeDirectly::SchemaChangeDirectly(const RowBlockChanger& row_block_changer)
: _row_block_changer(row_block_changer), _row_block_allocator(nullptr), _cursor(nullptr) {}
SchemaChangeDirectly::~SchemaChangeDirectly() {
VLOG_NOTICE << "~SchemaChangeDirectly()";
SAFE_DELETE(_row_block_allocator);
SAFE_DELETE(_cursor);
}
bool SchemaChangeDirectly::_write_row_block(RowsetWriter* rowset_writer, RowBlock* row_block) {
for (uint32_t i = 0; i < row_block->row_block_info().row_num; i++) {
row_block->get_row(i, _cursor);
if (!rowset_writer->add_row(*_cursor)) {
LOG(WARNING) << "fail to write to new rowset for direct schema change";
return false;
}
}
return true;
}
Status reserve_block(std::unique_ptr<RowBlock, RowBlockDeleter>* block_handle_ptr, int row_num,
RowBlockAllocator* allocator) {
auto& block_handle = *block_handle_ptr;
if (block_handle == nullptr || block_handle->capacity() < row_num) {
// release old block and alloc new block
if (block_handle != nullptr) {
block_handle.reset();
}
RowBlock* new_row_block = nullptr;
auto res = allocator->allocate(&new_row_block, row_num, true);
RETURN_NOT_OK_LOG(res, "failed to allocate RowBlock.");
block_handle.reset(new_row_block);
} else {
block_handle->clear();
}
return Status::OK();
}
Status SchemaChangeDirectly::_inner_process(RowsetReaderSharedPtr rowset_reader,
RowsetWriter* rowset_writer, TabletSharedPtr new_tablet,
TabletSchemaSPtr base_tablet_schema) {
if (_row_block_allocator == nullptr) {
_row_block_allocator = new RowBlockAllocator(new_tablet->tablet_schema(), 0);
if (_row_block_allocator == nullptr) {
LOG(FATAL) << "failed to malloc RowBlockAllocator. size=" << sizeof(RowBlockAllocator);
return Status::OLAPInternalError(OLAP_ERR_INPUT_PARAMETER_ERROR);
}
}
if (nullptr == _cursor) {
_cursor = new (nothrow) RowCursor();
if (nullptr == _cursor) {
LOG(WARNING) << "fail to allocate row cursor.";
return Status::OLAPInternalError(OLAP_ERR_INPUT_PARAMETER_ERROR);
}
if (!_cursor->init(new_tablet->tablet_schema())) {
LOG(WARNING) << "fail to init row cursor.";
return Status::OLAPInternalError(OLAP_ERR_INPUT_PARAMETER_ERROR);
}
}
Status res = Status::OK();
VLOG_NOTICE << "init writer. new_tablet=" << new_tablet->full_name()
<< ", block_row_number=" << new_tablet->num_rows_per_row_block();
std::unique_ptr<RowBlock, RowBlockDeleter> new_row_block(nullptr, [&](RowBlock* block) {
if (block != nullptr) {
_row_block_allocator->release(block);
}
});
RowBlock* ref_row_block = nullptr;
rowset_reader->next_block(&ref_row_block);
while (ref_row_block != nullptr && ref_row_block->has_remaining()) {
// We will allocate blocks of the same size as before
// to ensure that the data can be stored
RETURN_NOT_OK(reserve_block(&new_row_block, ref_row_block->row_block_info().row_num,
_row_block_allocator));
// Change ref to new. This step is reasonable to say that it does need to wait for a large block, but theoretically it has nothing to do with the writer.
uint64_t filtered_rows = 0;
res = _row_block_changer.change_row_block(ref_row_block, rowset_reader->version().second,
new_row_block.get(), &filtered_rows);
RETURN_NOT_OK_LOG(res, "failed to change data in row block.");
// rows filtered by delete handler one by one
_add_filtered_rows(filtered_rows);
if (!_write_row_block(rowset_writer, new_row_block.get())) {
res = Status::OLAPInternalError(OLAP_ERR_SCHEMA_CHANGE_INFO_INVALID);
LOG(WARNING) << "failed to write row block.";
return res;
}
ref_row_block->clear();
rowset_reader->next_block(&ref_row_block);
}
if (!rowset_writer->flush()) {
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
return res;
}
Status VSchemaChangeDirectly::_inner_process(RowsetReaderSharedPtr rowset_reader,
RowsetWriter* rowset_writer,
TabletSharedPtr new_tablet,
TabletSchemaSPtr base_tablet_schema) {
do {
auto new_block =
std::make_unique<vectorized::Block>(new_tablet->tablet_schema()->create_block());
auto ref_block = std::make_unique<vectorized::Block>(base_tablet_schema->create_block());
rowset_reader->next_block(ref_block.get());
if (ref_block->rows() < 1) {
break;
}
RETURN_IF_ERROR(_changer.change_block(ref_block.get(), new_block.get()));
RETURN_IF_ERROR(rowset_writer->add_block(new_block.get()));
} while (true);
if (!rowset_writer->flush()) {
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
return Status::OK();
}
SchemaChangeWithSorting::SchemaChangeWithSorting(const RowBlockChanger& row_block_changer,
size_t memory_limitation)
: _row_block_changer(row_block_changer),
_memory_limitation(memory_limitation),
_temp_delta_versions(Version::mock()),
_row_block_allocator(nullptr) {}
SchemaChangeWithSorting::~SchemaChangeWithSorting() {
VLOG_NOTICE << "~SchemaChangeWithSorting()";
SAFE_DELETE(_row_block_allocator);
}
VSchemaChangeWithSorting::VSchemaChangeWithSorting(const RowBlockChanger& row_block_changer,
size_t memory_limitation)
: _changer(row_block_changer),
_memory_limitation(memory_limitation),
_temp_delta_versions(Version::mock()) {
_mem_tracker = std::make_unique<MemTracker>(fmt::format(
"VSchemaChangeWithSorting:changer={}", std::to_string(int64(&row_block_changer))));
}
Status SchemaChangeWithSorting::_inner_process(RowsetReaderSharedPtr rowset_reader,
RowsetWriter* rowset_writer,
TabletSharedPtr new_tablet,
TabletSchemaSPtr base_tablet_schema) {
if (_row_block_allocator == nullptr) {
_row_block_allocator =
new (nothrow) RowBlockAllocator(new_tablet->tablet_schema(), _memory_limitation);
if (_row_block_allocator == nullptr) {
LOG(FATAL) << "failed to malloc RowBlockAllocator. size=" << sizeof(RowBlockAllocator);
return Status::OLAPInternalError(OLAP_ERR_INPUT_PARAMETER_ERROR);
}
}
Status res = Status::OK();
RowsetSharedPtr rowset = rowset_reader->rowset();
RowBlockSorter row_block_sorter(_row_block_allocator);
// for internal sorting
RowBlock* new_row_block = nullptr;
std::vector<RowBlock*> row_block_arr;
// for external sorting
// src_rowsets to store the rowset generated by internal sorting
std::vector<RowsetSharedPtr> src_rowsets;
Defer defer {[&]() {
// remove the intermediate rowsets generated by internal sorting
for (auto& row_set : src_rowsets) {
StorageEngine::instance()->add_unused_rowset(row_set);
}
for (auto block : row_block_arr) {
_row_block_allocator->release(block);
}
row_block_arr.clear();
}};
_temp_delta_versions.first = _temp_delta_versions.second;
SegmentsOverlapPB segments_overlap = rowset->rowset_meta()->segments_overlap();
int64_t oldest_write_timestamp = rowset->oldest_write_timestamp();
int64_t newest_write_timestamp = rowset->newest_write_timestamp();
RowBlock* ref_row_block = nullptr;
rowset_reader->next_block(&ref_row_block);
while (ref_row_block != nullptr && ref_row_block->has_remaining()) {
auto st = _row_block_allocator->allocate(&new_row_block,
ref_row_block->row_block_info().row_num, true);
// if OLAP_ERR_FETCH_MEMORY_EXCEEDED == st.precise_code()
// that mean RowBlockAllocator::alocate() memory exceeded.
// But we can flush row_block_arr if row_block_arr is not empty.
// Don't return directly.
if (OLAP_ERR_MALLOC_ERROR == st.precise_code()) {
return Status::OLAPInternalError(OLAP_ERR_MALLOC_ERROR);
} else if (st) {
// do memory check for sorting, in case schema change task fail at row block sorting because of
// not doing internal sorting first
if (!_row_block_allocator->is_memory_enough_for_sorting(
ref_row_block->row_block_info().row_num, row_block_sorter.num_rows())) {
if (new_row_block != nullptr) {
_row_block_allocator->release(new_row_block);
new_row_block = nullptr;
}
}
}
if (new_row_block == nullptr) {
if (row_block_arr.empty()) {
LOG(WARNING) << "Memory limitation is too small for Schema Change."
<< "memory_limitation=" << _memory_limitation
<< "You can increase the memory "
<< "by changing the "
"Config.memory_limitation_per_thread_for_schema_change_bytes";
return Status::OLAPInternalError(OLAP_ERR_FETCH_MEMORY_EXCEEDED);
}
// enter here while memory limitation is reached.
RowsetSharedPtr rowset;
if (!_internal_sorting(
row_block_arr,
Version(_temp_delta_versions.second, _temp_delta_versions.second),
oldest_write_timestamp, newest_write_timestamp, new_tablet,
segments_overlap, &rowset)) {
LOG(WARNING) << "failed to sorting internally.";
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
src_rowsets.push_back(rowset);
for (auto block : row_block_arr) {
_row_block_allocator->release(block);
}
row_block_arr.clear();
// increase temp version
++_temp_delta_versions.second;
continue;
}
uint64_t filtered_rows = 0;
res = _row_block_changer.change_row_block(ref_row_block, rowset_reader->version().second,
new_row_block, &filtered_rows);
if (!res) {
row_block_arr.push_back(new_row_block);
LOG(WARNING) << "failed to change data in row block.";
return res;
}
_add_filtered_rows(filtered_rows);
if (new_row_block->row_block_info().row_num > 0) {
if (!row_block_sorter.sort(&new_row_block)) {
row_block_arr.push_back(new_row_block);
LOG(WARNING) << "failed to sort row block.";
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
row_block_arr.push_back(new_row_block);
} else {
LOG(INFO) << "new block num rows is: " << new_row_block->row_block_info().row_num;
_row_block_allocator->release(new_row_block);
new_row_block = nullptr;
}
ref_row_block->clear();
rowset_reader->next_block(&ref_row_block);
}
if (!row_block_arr.empty()) {
// enter here while memory limitation is reached.
RowsetSharedPtr rowset = nullptr;
if (!_internal_sorting(row_block_arr,
Version(_temp_delta_versions.second, _temp_delta_versions.second),
oldest_write_timestamp, newest_write_timestamp, new_tablet,
segments_overlap, &rowset)) {
LOG(WARNING) << "failed to sorting internally.";
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
src_rowsets.push_back(rowset);
for (auto block : row_block_arr) {
_row_block_allocator->release(block);
}
row_block_arr.clear();
// increase temp version
++_temp_delta_versions.second;
}
if (src_rowsets.empty()) {
res = rowset_writer->flush();
if (!res) {
LOG(WARNING) << "create empty version for schema change failed."
<< " version=" << rowset_writer->version().first << "-"
<< rowset_writer->version().second;
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
} else if (!_external_sorting(src_rowsets, rowset_writer, new_tablet)) {
LOG(WARNING) << "failed to sorting externally.";
return Status::OLAPInternalError(OLAP_ERR_ALTER_STATUS_ERR);
}
return res;
}
Status VSchemaChangeWithSorting::_inner_process(RowsetReaderSharedPtr rowset_reader,
RowsetWriter* rowset_writer,
TabletSharedPtr new_tablet,
TabletSchemaSPtr base_tablet_schema) {
// for internal sorting
std::vector<std::unique_ptr<vectorized::Block>> blocks;
// for external sorting
// src_rowsets to store the rowset generated by internal sorting
std::vector<RowsetSharedPtr> src_rowsets;
Defer defer {[&]() {
// remove the intermediate rowsets generated by internal sorting
for (auto& row_set : src_rowsets) {
StorageEngine::instance()->add_unused_rowset(row_set);
}
}};
RowsetSharedPtr rowset = rowset_reader->rowset();
SegmentsOverlapPB segments_overlap = rowset->rowset_meta()->segments_overlap();
int64_t oldest_write_timestamp = rowset->oldest_write_timestamp();
int64_t newest_write_timestamp = rowset->newest_write_timestamp();
_temp_delta_versions.first = _temp_delta_versions.second;
auto create_rowset = [&]() -> Status {
if (blocks.empty()) {
return Status::OK();
}
RowsetSharedPtr rowset;
RETURN_IF_ERROR(_internal_sorting(
blocks, Version(_temp_delta_versions.second, _temp_delta_versions.second),
oldest_write_timestamp, newest_write_timestamp, new_tablet, BETA_ROWSET,
segments_overlap, &rowset));
src_rowsets.push_back(rowset);
for (auto& block : blocks) {
_mem_tracker->release(block->allocated_bytes());
}
blocks.clear();
// increase temp version
_temp_delta_versions.second++;
return Status::OK();
};
auto new_block =
std::make_unique<vectorized::Block>(new_tablet->tablet_schema()->create_block());
do {
auto ref_block = std::make_unique<vectorized::Block>(base_tablet_schema->create_block());
rowset_reader->next_block(ref_block.get());
if (ref_block->rows() < 1) {
break;
}
RETURN_IF_ERROR(_changer.change_block(ref_block.get(), new_block.get()));
if (!_mem_tracker->check_limit(_memory_limitation, new_block->allocated_bytes())) {
RETURN_IF_ERROR(create_rowset());
if (!_mem_tracker->check_limit(_memory_limitation, new_block->allocated_bytes())) {
LOG(WARNING) << "Memory limitation is too small for Schema Change."
<< " _memory_limitation=" << _memory_limitation
<< ", new_block->allocated_bytes()=" << new_block->allocated_bytes()
<< ", consumption=" << _mem_tracker->consumption();
return Status::OLAPInternalError(OLAP_ERR_INPUT_PARAMETER_ERROR);
}
}
_mem_tracker->consume(new_block->allocated_bytes());
// move unique ptr
blocks.push_back(
std::make_unique<vectorized::Block>(new_tablet->tablet_schema()->create_block()));
swap(blocks.back(), new_block);
} while (true);
RETURN_IF_ERROR(create_rowset());
if (src_rowsets.empty()) {
RETURN_IF_ERROR(rowset_writer->flush());
} else {
RETURN_IF_ERROR(_external_sorting(src_rowsets, rowset_writer, new_tablet));
}
return Status::OK();
}
bool SchemaChangeWithSorting::_internal_sorting(
const std::vector<RowBlock*>& row_block_arr, const Version& version,
int64_t oldest_write_timestamp, int64_t newest_write_timestamp, TabletSharedPtr new_tablet,
SegmentsOverlapPB segments_overlap, RowsetSharedPtr* rowset) {
uint64_t merged_rows = 0;
RowBlockMerger merger(new_tablet);
VLOG_NOTICE << "init rowset builder. tablet=" << new_tablet->full_name()
<< ", block_row_size=" << new_tablet->num_rows_per_row_block();
std::unique_ptr<RowsetWriter> rowset_writer;
if (!new_tablet->create_rowset_writer(version, VISIBLE, segments_overlap,
new_tablet->tablet_schema(), oldest_write_timestamp,
newest_write_timestamp, &rowset_writer)) {
return false;
}
if (!merger.merge(row_block_arr, rowset_writer.get(), &merged_rows)) {
LOG(WARNING) << "failed to merge row blocks.";
new_tablet->data_dir()->remove_pending_ids(ROWSET_ID_PREFIX +
rowset_writer->rowset_id().to_string());
return false;
}
new_tablet->data_dir()->remove_pending_ids(ROWSET_ID_PREFIX +
rowset_writer->rowset_id().to_string());
_add_merged_rows(merged_rows);
*rowset = rowset_writer->build();
return true;
}
Status VSchemaChangeWithSorting::_internal_sorting(
const std::vector<std::unique_ptr<vectorized::Block>>& blocks, const Version& version,
int64_t oldest_write_timestamp, int64_t newest_write_timestamp, TabletSharedPtr new_tablet,
RowsetTypePB new_rowset_type, SegmentsOverlapPB segments_overlap, RowsetSharedPtr* rowset) {
uint64_t merged_rows = 0;
MultiBlockMerger merger(new_tablet);
std::unique_ptr<RowsetWriter> rowset_writer;
RETURN_IF_ERROR(new_tablet->create_rowset_writer(
version, VISIBLE, segments_overlap, new_tablet->tablet_schema(), oldest_write_timestamp,
newest_write_timestamp, &rowset_writer));
Defer defer {[&]() {
new_tablet->data_dir()->remove_pending_ids(ROWSET_ID_PREFIX +
rowset_writer->rowset_id().to_string());
}};
RETURN_IF_ERROR(merger.merge(blocks, rowset_writer.get(), &merged_rows));
_add_merged_rows(merged_rows);
*rowset = rowset_writer->build();
return Status::OK();
}
bool SchemaChangeWithSorting::_external_sorting(vector<RowsetSharedPtr>& src_rowsets,
RowsetWriter* rowset_writer,
TabletSharedPtr new_tablet) {
std::vector<RowsetReaderSharedPtr> rs_readers;
for (auto& rowset : src_rowsets) {
RowsetReaderSharedPtr rs_reader;
auto res = rowset->create_reader(&rs_reader);
if (!res) {
LOG(WARNING) << "failed to create rowset reader.";
return false;
}
rs_readers.push_back(rs_reader);
}
Merger::Statistics stats;
auto res = Merger::merge_rowsets(new_tablet, READER_ALTER_TABLE, new_tablet->tablet_schema(),
rs_readers, rowset_writer, &stats);
if (!res) {
LOG(WARNING) << "failed to merge rowsets. tablet=" << new_tablet->full_name()
<< ", version=" << rowset_writer->version().first << "-"
<< rowset_writer->version().second;
return false;
}
_add_merged_rows(stats.merged_rows);
_add_filtered_rows(stats.filtered_rows);
return true;
}
Status VSchemaChangeWithSorting::_external_sorting(vector<RowsetSharedPtr>& src_rowsets,
RowsetWriter* rowset_writer,
TabletSharedPtr new_tablet) {
std::vector<RowsetReaderSharedPtr> rs_readers;
for (auto& rowset : src_rowsets) {
RowsetReaderSharedPtr rs_reader;
RETURN_IF_ERROR(rowset->create_reader(&rs_reader));
rs_readers.push_back(rs_reader);
}
Merger::Statistics stats;
RETURN_IF_ERROR(Merger::vmerge_rowsets(new_tablet, READER_ALTER_TABLE,
new_tablet->tablet_schema(), rs_readers, rowset_writer,
&stats));
_add_merged_rows(stats.merged_rows);
_add_filtered_rows(stats.filtered_rows);
return Status::OK();
}
Status SchemaChangeHandler::process_alter_tablet_v2(const TAlterTabletReqV2& request) {
LOG(INFO) << "begin to do request alter tablet: base_tablet_id=" << request.base_tablet_id
<< ", new_tablet_id=" << request.new_tablet_id
<< ", alter_version=" << request.alter_version;
TabletSharedPtr base_tablet =
StorageEngine::instance()->tablet_manager()->get_tablet(request.base_tablet_id);
if (base_tablet == nullptr) {
LOG(WARNING) << "fail to find base tablet. base_tablet=" << request.base_tablet_id;
return Status::OLAPInternalError(OLAP_ERR_TABLE_NOT_FOUND);
}
// Lock schema_change_lock util schema change info is stored in tablet header
std::unique_lock<std::mutex> schema_change_lock(base_tablet->get_schema_change_lock(),
std::try_to_lock);
if (!schema_change_lock.owns_lock()) {
LOG(WARNING) << "failed to obtain schema change lock. "
<< "base_tablet=" << request.base_tablet_id;
return Status::OLAPInternalError(OLAP_ERR_TRY_LOCK_FAILED);
}
Status res = _do_process_alter_tablet_v2(request);
LOG(INFO) << "finished alter tablet process, res=" << res;
return res;
}
std::shared_mutex SchemaChangeHandler::_mutex;
std::unordered_set<int64_t> SchemaChangeHandler::_tablet_ids_in_converting;
std::set<std::string> SchemaChangeHandler::_supported_functions = {"hll_hash", "to_bitmap"};
// In the past schema change and rollup will create new tablet and will wait for txns starting before the task to finished
// It will cost a lot of time to wait and the task is very difficult to understand.
// In alter task v2, FE will call BE to create tablet and send an alter task to BE to convert historical data.
// The admin should upgrade all BE and then upgrade FE.
// Should delete the old code after upgrade finished.
Status SchemaChangeHandler::_do_process_alter_tablet_v2(const TAlterTabletReqV2& request) {
Status res = Status::OK();
TabletSharedPtr base_tablet =
StorageEngine::instance()->tablet_manager()->get_tablet(request.base_tablet_id);
if (base_tablet == nullptr) {
LOG(WARNING) << "fail to find base tablet. base_tablet=" << request.base_tablet_id;
return Status::OLAPInternalError(OLAP_ERR_TABLE_NOT_FOUND);
}
// new tablet has to exist
TabletSharedPtr new_tablet =
StorageEngine::instance()->tablet_manager()->get_tablet(request.new_tablet_id);
if (new_tablet == nullptr) {
LOG(WARNING) << "fail to find new tablet."
<< " new_tablet=" << request.new_tablet_id;
return Status::OLAPInternalError(OLAP_ERR_TABLE_NOT_FOUND);
}
// check if tablet's state is not_ready, if it is ready, it means the tablet already finished
// check whether the tablet's max continuous version == request.version
if (new_tablet->tablet_state() != TABLET_NOTREADY) {
res = _validate_alter_result(new_tablet, request);
LOG(INFO) << "tablet's state=" << new_tablet->tablet_state()
<< " the convert job already finished, check its version"
<< " res=" << res;
return res;
}
LOG(INFO) << "finish to validate alter tablet request. begin to convert data from base tablet "
"to new tablet"
<< " base_tablet=" << base_tablet->full_name()
<< " new_tablet=" << new_tablet->full_name();
std::shared_lock base_migration_rlock(base_tablet->get_migration_lock(), std::try_to_lock);
if (!base_migration_rlock.owns_lock()) {
return Status::OLAPInternalError(OLAP_ERR_RWLOCK_ERROR);
}
std::shared_lock new_migration_rlock(new_tablet->get_migration_lock(), std::try_to_lock);
if (!new_migration_rlock.owns_lock()) {
return Status::OLAPInternalError(OLAP_ERR_RWLOCK_ERROR);
}
std::vector<Version> versions_to_be_changed;
int64_t end_version = -1;
// reader_context is stack variables, it's lifetime should keep the same
// with rs_readers
RowsetReaderContext reader_context;
std::vector<RowsetReaderSharedPtr> rs_readers;
// delete handlers for new tablet
DeleteHandler delete_handler;
std::vector<ColumnId> return_columns;
// Create a new tablet schema, should merge with dropped columns in light weight schema change
TabletSchemaSPtr base_tablet_schema = std::make_shared<TabletSchema>();
base_tablet_schema->copy_from(*base_tablet->tablet_schema());
if (!request.columns.empty() && request.columns[0].col_unique_id >= 0) {
base_tablet_schema->clear_columns();
for (const auto& column : request.columns) {
base_tablet_schema->append_column(TabletColumn(column));
}
}
// Use tablet schema directly from base tablet, they are the newest schema, not contain
// dropped column during light weight schema change.
// But the tablet schema in base tablet maybe not the latest from FE, so that if fe pass through
// a tablet schema, then use request schema.
size_t num_cols = request.columns.empty() ? base_tablet->tablet_schema()->num_columns()
: request.columns.size();
return_columns.resize(num_cols);
for (int i = 0; i < num_cols; ++i) {
return_columns[i] = i;
}
// begin to find deltas to convert from base tablet to new tablet so that
// obtain base tablet and new tablet's push lock and header write lock to prevent loading data
{
std::lock_guard<std::mutex> base_tablet_lock(base_tablet->get_push_lock());
std::lock_guard<std::mutex> new_tablet_lock(new_tablet->get_push_lock());
std::lock_guard<std::shared_mutex> base_tablet_wlock(base_tablet->get_header_lock());
std::lock_guard<std::shared_mutex> new_tablet_wlock(new_tablet->get_header_lock());
do {
RowsetSharedPtr max_rowset;
// get history data to be converted and it will check if there is hold in base tablet
if (!_get_versions_to_be_changed(base_tablet, &versions_to_be_changed, &max_rowset)) {
LOG(WARNING) << "fail to get version to be changed. res=" << res;
break;
}
// should check the max_version >= request.alter_version, if not the convert is useless
if (max_rowset == nullptr || max_rowset->end_version() < request.alter_version) {
res = Status::InternalError(
"base tablet's max version={} is less than request version={}",
(max_rowset == nullptr ? 0 : max_rowset->end_version()),
request.alter_version);
break;
}
// before calculating version_to_be_changed,
// remove all data from new tablet, prevent to rewrite data(those double pushed when wait)
LOG(INFO) << "begin to remove all data from new tablet to prevent rewrite."
<< " new_tablet=" << new_tablet->full_name();
std::vector<RowsetSharedPtr> rowsets_to_delete;
std::vector<std::pair<Version, RowsetSharedPtr>> version_rowsets;
new_tablet->acquire_version_and_rowsets(&version_rowsets);
std::sort(version_rowsets.begin(), version_rowsets.end(),
[](const std::pair<Version, RowsetSharedPtr>& l,
const std::pair<Version, RowsetSharedPtr>& r) {
return l.first.first < r.first.first;
});
for (auto& pair : version_rowsets) {
if (pair.first.second <= max_rowset->end_version()) {
rowsets_to_delete.push_back(pair.second);
} else if (pair.first.first <= max_rowset->end_version()) {
// If max version is [X-10] and new tablet has version [7-9][10-12],
// we only can remove [7-9] from new tablet. If we add [X-10] to new tablet, it will has version
// cross: [X-10] [10-12].
// So, we should return OLAP_ERR_VERSION_ALREADY_MERGED for fast fail.
LOG(WARNING) << "New tablet has a version " << pair.first
<< " crossing base tablet's max_version="
<< max_rowset->end_version();
return Status::OLAPInternalError(OLAP_ERR_VERSION_ALREADY_MERGED);
}
}
std::vector<RowsetSharedPtr> empty_vec;
new_tablet->modify_rowsets(empty_vec, rowsets_to_delete);
// inherit cumulative_layer_point from base_tablet
// check if new_tablet.ce_point > base_tablet.ce_point?
new_tablet->set_cumulative_layer_point(-1);
// save tablet meta
new_tablet->save_meta();
for (auto& rowset : rowsets_to_delete) {
// do not call rowset.remove directly, using gc thread to delete it
StorageEngine::instance()->add_unused_rowset(rowset);
}
// init one delete handler
for (auto& version : versions_to_be_changed) {
end_version = std::max(end_version, version.second);
}
// acquire data sources correspond to history versions
base_tablet->capture_rs_readers(versions_to_be_changed, &rs_readers);
if (rs_readers.empty()) {
LOG(WARNING) << "fail to acquire all data sources. "
<< "version_num=" << versions_to_be_changed.size()
<< ", data_source_num=" << rs_readers.size();
res = Status::OLAPInternalError(OLAP_ERR_ALTER_DELTA_DOES_NOT_EXISTS);
break;
}
auto& all_del_preds = base_tablet->delete_predicates();
for (auto& delete_pred : all_del_preds) {
if (delete_pred->version().first > end_version) {
continue;
}
base_tablet_schema->merge_dropped_columns(
base_tablet->tablet_schema(delete_pred->version()));
}
res = delete_handler.init(base_tablet_schema, all_del_preds, end_version);
if (!res) {
LOG(WARNING) << "init delete handler failed. base_tablet="
<< base_tablet->full_name() << ", end_version=" << end_version;
break;
}
reader_context.reader_type = READER_ALTER_TABLE;
reader_context.tablet_schema = base_tablet_schema;
reader_context.need_ordered_result = true;
reader_context.delete_handler = &delete_handler;
reader_context.return_columns = &return_columns;
reader_context.sequence_id_idx = reader_context.tablet_schema->sequence_col_idx();
reader_context.is_unique = base_tablet->keys_type() == UNIQUE_KEYS;
reader_context.batch_size = ALTER_TABLE_BATCH_SIZE;
reader_context.is_vec = config::enable_vectorized_alter_table;
reader_context.delete_bitmap = &base_tablet->tablet_meta()->delete_bitmap();
reader_context.version = Version(0, end_version);
for (auto& rs_reader : rs_readers) {
res = rs_reader->init(&reader_context);
if (!res) {
LOG(WARNING) << "failed to init rowset reader: " << base_tablet->full_name();
break;
}
}
} while (false);
}
do {
if (!res) {
break;
}
SchemaChangeParams sc_params;
DescriptorTbl::create(&sc_params.pool, request.desc_tbl, &sc_params.desc_tbl);
sc_params.base_tablet = base_tablet;
sc_params.new_tablet = new_tablet;
sc_params.ref_rowset_readers = rs_readers;
sc_params.delete_handler = &delete_handler;
sc_params.base_tablet_schema = base_tablet_schema;
if (request.__isset.materialized_view_params) {
for (auto item : request.materialized_view_params) {
AlterMaterializedViewParam mv_param;
mv_param.column_name = item.column_name;
/*
* origin_column_name is always be set now,
* but origin_column_name may be not set in some materialized view function. eg:count(1)
*/
if (item.__isset.origin_column_name) {
mv_param.origin_column_name = item.origin_column_name;
}
/*
* TODO(lhy)
* Building the materialized view function for schema_change here based on defineExpr.
* This is a trick because the current storage layer does not support expression evaluation.
* We can refactor this part of the code until the uniform expression evaluates the logic.
* count distinct materialized view will set mv_expr with to_bitmap or hll_hash.
* count materialized view will set mv_expr with count.
*/
if (item.__isset.mv_expr) {
if (item.mv_expr.nodes[0].node_type == TExprNodeType::FUNCTION_CALL) {
mv_param.mv_expr = item.mv_expr.nodes[0].fn.name.function_name;
if (!_supported_functions.count(mv_param.mv_expr)) {
return Status::NotSupported("Unknow materialized view expr " +
mv_param.mv_expr);
}
} else if (item.mv_expr.nodes[0].node_type == TExprNodeType::CASE_EXPR) {
mv_param.mv_expr = "count_field";
}
mv_param.expr = std::make_shared<TExpr>(item.mv_expr);
}
sc_params.materialized_params_map.insert(
std::make_pair(item.column_name, mv_param));
}
}
{
std::lock_guard<std::shared_mutex> wrlock(_mutex);
_tablet_ids_in_converting.insert(new_tablet->tablet_id());
}
res = _convert_historical_rowsets(sc_params);
if (new_tablet->keys_type() != UNIQUE_KEYS ||
!new_tablet->enable_unique_key_merge_on_write() || !res) {
{
std::lock_guard<std::shared_mutex> wrlock(_mutex);
_tablet_ids_in_converting.erase(new_tablet->tablet_id());
}
}
if (!res) {
break;
}
// For unique with merge-on-write table, should process delete bitmap here.
// 1. During double write, the newly imported rowsets does not calculate
// delete bitmap and publish successfully.
// 2. After conversion, calculate delete bitmap for the rowsets imported
// during double write. During this period, new data can still be imported
// witout calculating delete bitmap and publish successfully.
// 3. Block the new publish, calculate the delete bitmap of the
// incremental rowsets.
// 4. Switch the tablet status to TABLET_RUNNING. The newly imported
// data will calculate delete bitmap.
if (new_tablet->keys_type() == UNIQUE_KEYS &&
new_tablet->enable_unique_key_merge_on_write()) {
// step 2
int64_t max_version = new_tablet->max_version().second;
std::vector<RowsetSharedPtr> rowsets;
if (end_version < max_version) {
LOG(INFO)
<< "alter table for unique with merge-on-write, calculate delete bitmap of "
<< "double write rowsets for version: " << end_version + 1 << "-"
<< max_version;
RETURN_IF_ERROR(new_tablet->capture_consistent_rowsets(
{end_version + 1, max_version}, &rowsets));
}
for (auto rowset_ptr : rowsets) {
if (rowset_ptr->version().second <= end_version) {
continue;
}
std::lock_guard<std::mutex> rwlock(new_tablet->get_rowset_update_lock());
std::shared_lock<std::shared_mutex> wrlock(new_tablet->get_header_lock());
RETURN_IF_ERROR(new_tablet->update_delete_bitmap_without_lock(rowset_ptr));
}
// step 3
std::lock_guard<std::mutex> rwlock(new_tablet->get_rowset_update_lock());
std::lock_guard<std::shared_mutex> new_wlock(new_tablet->get_header_lock());
int64_t new_max_version = new_tablet->max_version().second;
rowsets.clear();
if (max_version < new_max_version) {
LOG(INFO)
<< "alter table for unique with merge-on-write, calculate delete bitmap of "
<< "incremental rowsets for version: " << max_version + 1 << "-"
<< new_max_version;
RETURN_IF_ERROR(new_tablet->capture_consistent_rowsets(
{max_version + 1, new_max_version}, &rowsets));
}
for (auto rowset_ptr : rowsets) {
if (rowset_ptr->version().second <= max_version) {
continue;
}
RETURN_IF_ERROR(new_tablet->update_delete_bitmap_without_lock(rowset_ptr));
}
// step 4
{
std::lock_guard<std::shared_mutex> wrlock(_mutex);
_tablet_ids_in_converting.erase(new_tablet->tablet_id());
}
res = new_tablet->set_tablet_state(TabletState::TABLET_RUNNING);
if (!res) {
break;
}
new_tablet->save_meta();
} else {
// set state to ready
std::lock_guard<std::shared_mutex> new_wlock(new_tablet->get_header_lock());
res = new_tablet->set_tablet_state(TabletState::TABLET_RUNNING);
if (!res) {
break;
}
new_tablet->save_meta();
}
} while (false);
if (res) {
// _validate_alter_result should be outside the above while loop.
// to avoid requiring the header lock twice.
res = _validate_alter_result(new_tablet, request);
}
// if failed convert history data, then just remove the new tablet
if (!res) {
LOG(WARNING) << "failed to alter tablet. base_tablet=" << base_tablet->full_name()
<< ", drop new_tablet=" << new_tablet->full_name();
// do not drop the new tablet and its data. GC thread will
}
return res;
}
bool SchemaChangeHandler::tablet_in_converting(int64_t tablet_id) {
std::shared_lock rdlock(_mutex);
return _tablet_ids_in_converting.find(tablet_id) != _tablet_ids_in_converting.end();
}
Status SchemaChangeHandler::_get_versions_to_be_changed(
TabletSharedPtr base_tablet, std::vector<Version>* versions_to_be_changed,
RowsetSharedPtr* max_rowset) {
RowsetSharedPtr rowset = base_tablet->rowset_with_max_version();
if (rowset == nullptr) {
LOG(WARNING) << "Tablet has no version. base_tablet=" << base_tablet->full_name();
return Status::OLAPInternalError(OLAP_ERR_ALTER_DELTA_DOES_NOT_EXISTS);
}
*max_rowset = rowset;
RETURN_NOT_OK(base_tablet->capture_consistent_versions(Version(0, rowset->version().second),
versions_to_be_changed));
return Status::OK();
}
Status SchemaChangeHandler::_convert_historical_rowsets(const SchemaChangeParams& sc_params) {
LOG(INFO) << "begin to convert historical rowsets for new_tablet from base_tablet."
<< " base_tablet=" << sc_params.base_tablet->full_name()
<< ", new_tablet=" << sc_params.new_tablet->full_name();
// find end version
int32_t end_version = -1;
for (size_t i = 0; i < sc_params.ref_rowset_readers.size(); ++i) {
if (sc_params.ref_rowset_readers[i]->version().second > end_version) {
end_version = sc_params.ref_rowset_readers[i]->version().second;
}
}
// Add filter information in change, and filter column information will be set in _parse_request
// And filter some data every time the row block changes
RowBlockChanger rb_changer(sc_params.new_tablet->tablet_schema(), sc_params.delete_handler,
*sc_params.desc_tbl);
bool sc_sorting = false;
bool sc_directly = false;
// a.Parse the Alter request and convert it into an internal representation
Status res = _parse_request(sc_params, &rb_changer, &sc_sorting, &sc_directly);
auto process_alter_exit = [&]() -> Status {
{
// save tablet meta here because rowset meta is not saved during add rowset
std::lock_guard<std::shared_mutex> new_wlock(sc_params.new_tablet->get_header_lock());
sc_params.new_tablet->save_meta();
}
if (res) {
Version test_version(0, end_version);
res = sc_params.new_tablet->check_version_integrity(test_version);
}
LOG(INFO) << "finish converting rowsets for new_tablet from base_tablet. "
<< "base_tablet=" << sc_params.base_tablet->full_name()
<< ", new_tablet=" << sc_params.new_tablet->full_name();
return res;
};
if (!res) {
LOG(WARNING) << "failed to parse the request. res=" << res;
return process_alter_exit();
}
// b. Generate historical data converter
auto sc_procedure = get_sc_procedure(rb_changer, sc_sorting, sc_directly);
// c.Convert historical data
for (auto& rs_reader : sc_params.ref_rowset_readers) {
VLOG_TRACE << "begin to convert a history rowset. version=" << rs_reader->version().first
<< "-" << rs_reader->version().second;
// set status for monitor
// As long as there is a new_table as running, ref table is set as running
// NOTE If the first sub_table fails first, it will continue to go as normal here
TabletSharedPtr new_tablet = sc_params.new_tablet;
// When tablet create new rowset writer, it may change rowset type, in this case
// linked schema change will not be used.
std::unique_ptr<RowsetWriter> rowset_writer;
Status status = new_tablet->create_rowset_writer(
rs_reader->version(), VISIBLE,
rs_reader->rowset()->rowset_meta()->segments_overlap(), new_tablet->tablet_schema(),
rs_reader->oldest_write_timestamp(), rs_reader->newest_write_timestamp(),
&rowset_writer);
if (!status.ok()) {
res = Status::OLAPInternalError(OLAP_ERR_ROWSET_BUILDER_INIT);
return process_alter_exit();
}
if (res = sc_procedure->process(rs_reader, rowset_writer.get(), sc_params.new_tablet,
sc_params.base_tablet, sc_params.base_tablet_schema);
!res) {
LOG(WARNING) << "failed to process the version."
<< " version=" << rs_reader->version().first << "-"
<< rs_reader->version().second;
new_tablet->data_dir()->remove_pending_ids(ROWSET_ID_PREFIX +
rowset_writer->rowset_id().to_string());
return process_alter_exit();
}
new_tablet->data_dir()->remove_pending_ids(ROWSET_ID_PREFIX +
rowset_writer->rowset_id().to_string());
// Add the new version of the data to the header
// In order to prevent the occurrence of deadlock, we must first lock the old table, and then lock the new table
std::lock_guard<std::mutex> lock(sc_params.new_tablet->get_push_lock());
RowsetSharedPtr new_rowset = rowset_writer->build();
if (new_rowset == nullptr) {
LOG(WARNING) << "failed to build rowset, exit alter process";
return process_alter_exit();
}
res = sc_params.new_tablet->add_rowset(new_rowset);
if (res.precise_code() == OLAP_ERR_PUSH_VERSION_ALREADY_EXIST) {
LOG(WARNING) << "version already exist, version revert occurred. "
<< "tablet=" << sc_params.new_tablet->full_name() << ", version='"
<< rs_reader->version().first << "-" << rs_reader->version().second;
StorageEngine::instance()->add_unused_rowset(new_rowset);
res = Status::OK();
} else if (!res) {
LOG(WARNING) << "failed to register new version. "
<< " tablet=" << sc_params.new_tablet->full_name()
<< ", version=" << rs_reader->version().first << "-"
<< rs_reader->version().second;
StorageEngine::instance()->add_unused_rowset(new_rowset);
return process_alter_exit();
} else {
VLOG_NOTICE << "register new version. tablet=" << sc_params.new_tablet->full_name()
<< ", version=" << rs_reader->version().first << "-"
<< rs_reader->version().second;
}
VLOG_TRACE << "succeed to convert a history version."
<< " version=" << rs_reader->version().first << "-"
<< rs_reader->version().second;
}
// XXX:The SchemaChange state should not be canceled at this time, because the new Delta has to be converted to the old and new Schema version
return process_alter_exit();
}
// @static
// Analyze the mapping of the column and the mapping of the filter key
Status SchemaChangeHandler::_parse_request(const SchemaChangeParams& sc_params,
RowBlockChanger* rb_changer, bool* sc_sorting,
bool* sc_directly) {
TabletSharedPtr base_tablet = sc_params.base_tablet;
TabletSharedPtr new_tablet = sc_params.new_tablet;
TabletSchemaSPtr base_tablet_schema = sc_params.base_tablet_schema;
const std::unordered_map<std::string, AlterMaterializedViewParam>& materialized_function_map =
sc_params.materialized_params_map;
DescriptorTbl desc_tbl = *sc_params.desc_tbl;
// set column mapping
for (int i = 0, new_schema_size = new_tablet->tablet_schema()->num_columns();
i < new_schema_size; ++i) {
const TabletColumn& new_column = new_tablet->tablet_schema()->column(i);
const string& column_name = new_column.name();
ColumnMapping* column_mapping = rb_changer->get_mutable_column_mapping(i);
column_mapping->new_column = &new_column;
if (materialized_function_map.find(column_name) != materialized_function_map.end()) {
auto mvParam = materialized_function_map.find(column_name)->second;
column_mapping->materialized_function = mvParam.mv_expr;
column_mapping->expr = mvParam.expr;
int32_t column_index = base_tablet_schema->field_index(mvParam.origin_column_name);
if (column_index >= 0) {
column_mapping->ref_column = column_index;
continue;
} else {
LOG(WARNING) << "referenced column was missing. "
<< "[column=" << column_name << " referenced_column=" << column_index
<< "]";
return Status::OLAPInternalError(OLAP_ERR_CE_CMD_PARAMS_ERROR);
}
}
int32_t column_index = base_tablet_schema->field_index(column_name);
if (column_index >= 0) {
column_mapping->ref_column = column_index;
continue;
}
// Newly added column go here
column_mapping->ref_column = -1;
if (i < base_tablet_schema->num_short_key_columns()) {
*sc_directly = true;
}
RETURN_IF_ERROR(
_init_column_mapping(column_mapping, new_column, new_column.default_value()));
VLOG_TRACE << "A column with default value will be added after schema changing. "
<< "column=" << column_name << ", default_value=" << new_column.default_value();
}
// Check if re-aggregation is needed.
*sc_sorting = false;
// If the reference sequence of the Key column is out of order, it needs to be reordered
int num_default_value = 0;
for (int i = 0, new_schema_size = new_tablet->num_key_columns(); i < new_schema_size; ++i) {
ColumnMapping* column_mapping = rb_changer->get_mutable_column_mapping(i);
if (column_mapping->ref_column < 0) {
num_default_value++;
continue;
}
if (column_mapping->ref_column != i - num_default_value) {
*sc_sorting = true;
return Status::OK();
}
}
TabletSchemaSPtr new_tablet_schema = new_tablet->tablet_schema();
if (base_tablet_schema->keys_type() != new_tablet_schema->keys_type()) {
// only when base table is dup and mv is agg
// the rollup job must be reagg.
*sc_sorting = true;
return Status::OK();
}
// If the sort of key has not been changed but the new keys num is less then base's,
// the new table should be re agg.
// So we also need to set sc_sorting = true.
// A, B, C are keys(sort keys), D is value
// followings need resort:
// old keys: A B C D
// new keys: A B
if (new_tablet_schema->keys_type() != KeysType::DUP_KEYS &&
new_tablet->num_key_columns() < base_tablet_schema->num_key_columns()) {
// this is a table with aggregate key type, and num of key columns in new schema
// is less, which means the data in new tablet should be more aggregated.
// so we use sorting schema change to sort and merge the data.
*sc_sorting = true;
return Status::OK();
}
if (base_tablet_schema->num_short_key_columns() != new_tablet->num_short_key_columns()) {
// the number of short_keys changed, can't do linked schema change
*sc_directly = true;
return Status::OK();
}
for (size_t i = 0; i < new_tablet->num_columns(); ++i) {
ColumnMapping* column_mapping = rb_changer->get_mutable_column_mapping(i);
if (column_mapping->ref_column < 0) {
continue;
} else {
auto column_new = new_tablet_schema->column(i);
auto column_old = base_tablet_schema->column(column_mapping->ref_column);
if (column_new.type() != column_old.type() ||
column_new.precision() != column_old.precision() ||
column_new.frac() != column_old.frac() ||
column_new.length() != column_old.length() ||
column_new.is_bf_column() != column_old.is_bf_column() ||
column_new.has_bitmap_index() != column_old.has_bitmap_index()) {
*sc_directly = true;
return Status::OK();
}
}
}
if (!sc_params.delete_handler->empty()) {
// there exists delete condition in header, can't do linked schema change
*sc_directly = true;
}
if (base_tablet->tablet_meta()->preferred_rowset_type() !=
new_tablet->tablet_meta()->preferred_rowset_type()) {
// If the base_tablet and new_tablet rowset types are different, just use directly type
*sc_directly = true;
}
return Status::OK();
}
Status SchemaChangeHandler::_init_column_mapping(ColumnMapping* column_mapping,
const TabletColumn& column_schema,
const std::string& value) {
column_mapping->default_value = WrapperField::create(column_schema);
if (column_mapping->default_value == nullptr) {
return Status::OLAPInternalError(OLAP_ERR_MALLOC_ERROR);
}
if (column_schema.is_nullable() && value.length() == 0) {
column_mapping->default_value->set_null();
} else {
column_mapping->default_value->from_string(value, column_schema.precision(),
column_schema.frac());
}
return Status::OK();
}
Status SchemaChangeHandler::_validate_alter_result(TabletSharedPtr new_tablet,
const TAlterTabletReqV2& request) {
Version max_continuous_version = {-1, 0};
new_tablet->max_continuous_version_from_beginning(&max_continuous_version);
LOG(INFO) << "find max continuous version of tablet=" << new_tablet->full_name()
<< ", start_version=" << max_continuous_version.first
<< ", end_version=" << max_continuous_version.second;
if (max_continuous_version.second < request.alter_version) {
return Status::InternalError("result version={} is less than request version={}",
max_continuous_version.second, request.alter_version);
}
std::vector<std::pair<Version, RowsetSharedPtr>> version_rowsets;
{
std::shared_lock rdlock(new_tablet->get_header_lock());
new_tablet->acquire_version_and_rowsets(&version_rowsets);
}
for (auto& pair : version_rowsets) {
RowsetSharedPtr rowset = pair.second;
if (!rowset->check_file_exist()) {
return Status::OLAPInternalError(OLAP_ERR_FILE_NOT_EXIST);
}
}
return Status::OK();
}
} // namespace doris
Reference in New Issue View Git Blame Copy Permalink