blob: 734eb2be9aac971967328ab4025216200315efa6 [file] [log] [blame]
// 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.
// This file is copied from
// https://github.com/ClickHouse/ClickHouse/blob/master/src/Columns/ColumnStr<T>.cpp
// and modified by Doris
#include "vec/columns/column_string.h"
#include <algorithm>
#include <boost/iterator/iterator_facade.hpp>
#include "util/memcpy_inlined.h"
#include "util/simd/bits.h"
#include "vec/columns/columns_common.h"
#include "vec/common/arena.h"
#include "vec/common/assert_cast.h"
#include "vec/common/memcmp_small.h"
#include "vec/common/unaligned.h"
#include "vec/core/sort_block.h"
namespace doris::vectorized {
template <typename T>
void ColumnStr<T>::sanity_check() const {
#ifndef NDEBUG
sanity_check_simple();
auto count = (int)offsets.size();
for (int i = 0; i < count; ++i) {
if (offsets[i] < offsets[i - 1]) {
throw Exception(Status::InternalError("row count: {}, offsets[{}]: {}, offsets[{}]: {}",
count, i, offsets[i], i - 1, offsets[i - 1]));
}
}
#endif
}
template <typename T>
void ColumnStr<T>::sanity_check_simple() const {
#ifndef NDEBUG
auto count = (int)offsets.size();
if (chars.size() != offsets[count - 1]) {
throw Exception(Status::InternalError("row count: {}, chars.size(): {}, offset[{}]: {}",
count, chars.size(), count - 1, offsets[count - 1]));
}
if (offsets[-1] != 0) {
throw Exception(Status::InternalError("wrong offsets[-1]: {}", offsets[-1]));
}
#endif
}
template <typename T>
MutableColumnPtr ColumnStr<T>::clone_resized(size_t to_size) const {
auto res = ColumnStr<T>::create();
if (to_size == 0) {
return res;
}
size_t from_size = size();
if (to_size <= from_size) {
/// Just cut column.
res->offsets.assign(offsets.begin(), offsets.begin() + to_size);
res->chars.assign(chars.begin(), chars.begin() + offsets[to_size - 1]);
} else {
/// Copy column and append empty strings for extra elements.
if (from_size > 0) {
res->offsets.assign(offsets.begin(), offsets.end());
res->chars.assign(chars.begin(), chars.end());
}
res->offsets.resize_fill(to_size, chars.size());
}
res->sanity_check_simple();
return res;
}
template <typename T>
MutableColumnPtr ColumnStr<T>::get_shrinked_column() {
auto shrinked_column = ColumnStr<T>::create();
shrinked_column->get_offsets().reserve(offsets.size());
shrinked_column->get_chars().reserve(chars.size());
for (int i = 0; i < size(); i++) {
StringRef str = get_data_at(i);
reinterpret_cast<ColumnStr<T>*>(shrinked_column.get())
->insert_data(str.data, strnlen(str.data, str.size));
}
return shrinked_column;
}
// This method is only called by MutableBlock::merge_ignore_overflow
// by hash join operator to collect build data to avoid
// the total string length of a ColumnStr<uint32_t> column exceeds the 4G limit.
//
// After finishing collecting build data, a ColumnStr<uint32_t> column
// will be converted to ColumnStr<uint64_t> if the total string length
// exceeds the 4G limit by calling Block::replace_if_overflow.
template <typename T>
void ColumnStr<T>::insert_range_from_ignore_overflow(const doris::vectorized::IColumn& src,
size_t start, size_t length) {
if (length == 0) {
return;
}
const auto& src_concrete = assert_cast<const ColumnStr<T>&>(src);
if (start + length > src_concrete.offsets.size()) {
throw doris::Exception(doris::ErrorCode::INTERNAL_ERROR,
"Parameter out of bound in "
"IColumnStr<T>::insert_range_from_ignore_overflow method.");
}
size_t nested_offset = src_concrete.offset_at(start);
size_t nested_length = src_concrete.offsets[start + length - 1] - nested_offset;
size_t old_chars_size = chars.size();
chars.resize(old_chars_size + nested_length);
memcpy(&chars[old_chars_size], &src_concrete.chars[nested_offset], nested_length);
if (start == 0 && offsets.empty()) {
offsets.assign(src_concrete.offsets.begin(), src_concrete.offsets.begin() + length);
} else {
size_t old_size = offsets.size();
size_t prev_max_offset = offsets.back(); /// -1th index is Ok, see PaddedPODArray
offsets.resize(old_size + length);
for (size_t i = 0; i < length; ++i) {
// unsinged integer overflow is well defined in C++,
// so we don't need to check the overflow here.
offsets[old_size + i] =
src_concrete.offsets[start + i] - nested_offset + prev_max_offset;
}
}
#ifndef NDEBUG
auto count = int64_t(offsets.size());
// offsets may overflow, so we make chars.size() as T to do same overflow check
if (offsets.back() != T(chars.size())) {
throw Exception(Status::InternalError("row count: {}, chars.size(): {}, offset[{}]: {}",
count, chars.size(), count - 1, offsets[count - 1]));
}
#endif
}
template <typename T>
void ColumnStr<T>::insert_range_from(const IColumn& src, size_t start, size_t length) {
if (length == 0) {
return;
}
auto do_insert = [&](const auto& src_concrete) {
const auto& src_offsets = src_concrete.get_offsets();
const auto& src_chars = src_concrete.get_chars();
if (start + length > src_offsets.size()) {
throw doris::Exception(
doris::ErrorCode::INTERNAL_ERROR,
"Parameter out of bound in IColumnStr<T>::insert_range_from method.");
}
size_t nested_offset = src_offsets[static_cast<ssize_t>(start) - 1];
size_t nested_length = src_offsets[start + length - 1] - nested_offset;
size_t old_chars_size = chars.size();
check_chars_length(old_chars_size + nested_length, offsets.size() + length);
chars.resize(old_chars_size + nested_length);
memcpy(&chars[old_chars_size], &src_chars[nested_offset], nested_length);
using OffsetsType = std::decay_t<decltype(src_offsets)>;
if (std::is_same_v<T, typename OffsetsType::value_type> && start == 0 && offsets.empty()) {
offsets.assign(src_offsets.begin(), src_offsets.begin() + length);
} else {
size_t old_size = offsets.size();
size_t prev_max_offset = offsets.back(); /// -1th index is Ok, see PaddedPODArray
offsets.resize(old_size + length);
for (size_t i = 0; i < length; ++i) {
offsets[old_size + i] = src_offsets[start + i] - nested_offset + prev_max_offset;
}
}
};
// insert_range_from maybe called by ColumnArray::insert_indices_from(which is used by hash join operator),
// so we need to support both ColumnStr<uint32_t> and ColumnStr<uint64_t>
if (src.is_column_string64()) {
do_insert(assert_cast<const ColumnStr<uint64_t>&>(src));
} else {
do_insert(assert_cast<const ColumnStr<uint32_t>&>(src));
}
sanity_check_simple();
}
template <typename T>
void ColumnStr<T>::insert_many_from(const IColumn& src, size_t position, size_t length) {
const auto& string_column = assert_cast<const ColumnStr<T>&>(src);
auto [data_val, data_length] = string_column.get_data_at(position);
size_t old_chars_size = chars.size();
check_chars_length(old_chars_size + data_length * length, offsets.size() + length);
chars.resize(old_chars_size + data_length * length);
auto old_size = offsets.size();
offsets.resize(old_size + length);
auto start_pos = old_size;
auto end_pos = old_size + length;
auto prev_pos = old_chars_size;
for (; start_pos < end_pos; ++start_pos) {
memcpy(&chars[prev_pos], data_val, data_length);
offsets[start_pos] = prev_pos + data_length;
prev_pos = prev_pos + data_length;
}
}
template <typename T>
void ColumnStr<T>::insert_indices_from(const IColumn& src, const uint32_t* indices_begin,
const uint32_t* indices_end) {
auto do_insert = [&](const auto& src_str) {
const auto* __restrict src_offset_data = src_str.get_offsets().data();
auto old_char_size = chars.size();
size_t total_chars_size = old_char_size;
auto dst_offsets_pos = offsets.size();
offsets.resize(offsets.size() + indices_end - indices_begin);
auto* dst_offsets_data = offsets.data();
for (const auto* x = indices_begin; x != indices_end; ++x) {
int64_t src_offset = *x;
total_chars_size += src_offset_data[src_offset] - src_offset_data[src_offset - 1];
dst_offsets_data[dst_offsets_pos++] = total_chars_size;
}
check_chars_length(total_chars_size, offsets.size());
chars.resize(total_chars_size);
const auto* __restrict src_data_ptr = src_str.get_chars().data();
auto* dst_data_ptr = chars.data();
size_t dst_chars_pos = old_char_size;
for (const auto* x = indices_begin; x != indices_end; ++x) {
int64_t src_offset = *x;
const size_t size_to_append =
src_offset_data[src_offset] - src_offset_data[src_offset - 1];
const size_t offset = src_offset_data[src_offset - 1];
memcpy_inlined(dst_data_ptr + dst_chars_pos, src_data_ptr + offset, size_to_append);
dst_chars_pos += size_to_append;
}
};
if (src.is_column_string64()) {
do_insert(assert_cast<const ColumnStr<uint64_t>&>(src));
} else {
do_insert(assert_cast<const ColumnStr<uint32_t>&>(src));
}
sanity_check_simple();
}
template <typename T>
void ColumnStr<T>::update_crcs_with_value(uint32_t* __restrict hashes, doris::PrimitiveType type,
uint32_t rows, uint32_t offset,
const uint8_t* __restrict null_data) const {
auto s = rows;
DCHECK(s == size());
if (null_data == nullptr) {
for (size_t i = 0; i < s; i++) {
auto data_ref = get_data_at(i);
hashes[i] = HashUtil::zlib_crc_hash(data_ref.data, data_ref.size, hashes[i]);
}
} else {
for (size_t i = 0; i < s; i++) {
if (null_data[i] == 0) {
auto data_ref = get_data_at(i);
hashes[i] = HashUtil::zlib_crc_hash(data_ref.data, data_ref.size, hashes[i]);
}
}
}
}
template <typename T>
ColumnPtr ColumnStr<T>::filter(const IColumn::Filter& filt, ssize_t result_size_hint) const {
if constexpr (std::is_same_v<UInt32, T>) {
if (offsets.size() == 0) {
return ColumnStr<T>::create();
}
auto res = ColumnStr<T>::create();
Chars& res_chars = res->chars;
IColumn::Offsets& res_offsets = res->offsets;
filter_arrays_impl<UInt8, IColumn::Offset>(chars, offsets, res_chars, res_offsets, filt,
result_size_hint);
return res;
} else {
throw doris::Exception(doris::ErrorCode::INTERNAL_ERROR,
"should not call filter in ColumnStr<UInt64>");
}
}
template <typename T>
size_t ColumnStr<T>::filter(const IColumn::Filter& filter) {
if constexpr (std::is_same_v<UInt32, T>) {
CHECK_EQ(filter.size(), offsets.size());
if (offsets.size() == 0) {
resize(0);
return 0;
}
auto res = filter_arrays_impl<UInt8, IColumn::Offset>(chars, offsets, filter);
sanity_check_simple();
return res;
} else {
throw doris::Exception(doris::ErrorCode::INTERNAL_ERROR,
"should not call filter in ColumnStr<UInt64>");
}
}
template <typename T>
Status ColumnStr<T>::filter_by_selector(const uint16_t* sel, size_t sel_size, IColumn* col_ptr) {
if constexpr (std::is_same_v<UInt32, T>) {
auto* col = static_cast<ColumnStr<T>*>(col_ptr);
Chars& res_chars = col->chars;
IColumn::Offsets& res_offsets = col->offsets;
IColumn::Filter filter;
filter.resize_fill(offsets.size(), 0);
for (size_t i = 0; i < sel_size; i++) {
filter[sel[i]] = 1;
}
filter_arrays_impl<UInt8, IColumn::Offset>(chars, offsets, res_chars, res_offsets, filter,
sel_size);
return Status::OK();
} else {
return Status::InternalError("should not call filter_by_selector in ColumnStr<UInt64>");
}
}
template <typename T>
ColumnPtr ColumnStr<T>::permute(const IColumn::Permutation& perm, size_t limit) const {
size_t size = offsets.size();
if (limit == 0) {
limit = size;
} else {
limit = std::min(size, limit);
}
if (perm.size() < limit) {
throw doris::Exception(doris::ErrorCode::INTERNAL_ERROR,
"Size of permutation is less than required.");
__builtin_unreachable();
}
if (limit == 0) {
return ColumnStr<T>::create();
}
auto res = ColumnStr<T>::create();
Chars& res_chars = res->chars;
auto& res_offsets = res->offsets;
if (limit == size) {
res_chars.resize(chars.size());
} else {
size_t new_chars_size = 0;
for (size_t i = 0; i < limit; ++i) {
new_chars_size += size_at(perm[i]);
}
res_chars.resize(new_chars_size);
}
res_offsets.resize(limit);
T current_new_offset = 0;
for (size_t i = 0; i < limit; ++i) {
size_t j = perm[i];
size_t string_offset = offsets[j - 1];
size_t string_size = offsets[j] - string_offset;
memcpy_small_allow_read_write_overflow15(&res_chars[current_new_offset],
&chars[string_offset], string_size);
current_new_offset += string_size;
res_offsets[i] = current_new_offset;
}
sanity_check_simple();
return res;
}
template <typename T>
StringRef ColumnStr<T>::serialize_value_into_arena(size_t n, Arena& arena,
char const*& begin) const {
auto string_size(size_at(n));
auto offset(offset_at(n));
StringRef res;
res.size = sizeof(string_size) + string_size;
char* pos = arena.alloc_continue(res.size, begin);
memcpy(pos, &string_size, sizeof(string_size));
memcpy(pos + sizeof(string_size), &chars[offset], string_size);
res.data = pos;
return res;
}
template <typename T>
const char* ColumnStr<T>::deserialize_and_insert_from_arena(const char* pos) {
const auto string_size = unaligned_load<uint32_t>(pos);
pos += sizeof(string_size);
const size_t old_size = chars.size();
const size_t new_size = old_size + string_size;
check_chars_length(new_size, offsets.size() + 1);
chars.resize(new_size);
memcpy(chars.data() + old_size, pos, string_size);
offsets.push_back(new_size);
sanity_check_simple();
return pos + string_size;
}
template <typename T>
size_t ColumnStr<T>::get_max_row_byte_size() const {
size_t max_size = 0;
size_t num_rows = offsets.size();
for (size_t i = 0; i < num_rows; ++i) {
max_size = std::max(max_size, size_t(size_at(i)));
}
return max_size + sizeof(uint32_t);
}
template <typename T>
void ColumnStr<T>::serialize_vec(std::vector<StringRef>& keys, size_t num_rows,
size_t max_row_byte_size) const {
for (size_t i = 0; i < num_rows; ++i) {
auto offset(offset_at(i));
auto string_size(size_at(i));
auto* ptr = const_cast<char*>(keys[i].data + keys[i].size);
memcpy_fixed<uint32_t>(ptr, (char*)&string_size);
memcpy(ptr + sizeof(string_size), &chars[offset], string_size);
keys[i].size += sizeof(string_size) + string_size;
}
}
template <typename T>
void ColumnStr<T>::serialize_vec_with_null_map(std::vector<StringRef>& keys, size_t num_rows,
const UInt8* null_map) const {
DCHECK(null_map != nullptr);
const bool has_null = simd::contain_byte(null_map, num_rows, 1);
if (has_null) {
for (size_t i = 0; i < num_rows; ++i) {
char* __restrict dest = const_cast<char*>(keys[i].data + keys[i].size);
// serialize null first
memcpy(dest, null_map + i, sizeof(uint8_t));
if (null_map[i] == 0) {
UInt32 offset(offset_at(i));
UInt32 string_size(size_at(i));
memcpy_fixed<uint32_t>(dest + 1, (char*)&string_size);
memcpy(dest + 1 + sizeof(string_size), &chars[offset], string_size);
keys[i].size += sizeof(string_size) + string_size + sizeof(UInt8);
} else {
keys[i].size += sizeof(UInt8);
}
}
} else {
// All rows are not null, serialize null & value
for (size_t i = 0; i < num_rows; ++i) {
char* __restrict dest = const_cast<char*>(keys[i].data + keys[i].size);
// serialize null first
memcpy(dest, null_map + i, sizeof(uint8_t));
UInt32 offset(offset_at(i));
UInt32 string_size(size_at(i));
memcpy_fixed<uint32_t>(dest + 1, (char*)&string_size);
memcpy(dest + 1 + sizeof(string_size), &chars[offset], string_size);
keys[i].size += sizeof(string_size) + string_size + sizeof(UInt8);
}
}
}
template <typename T>
void ColumnStr<T>::deserialize_vec(std::vector<StringRef>& keys, const size_t num_rows) {
for (size_t i = 0; i != num_rows; ++i) {
const auto* original_ptr = keys[i].data;
keys[i].data = deserialize_and_insert_from_arena(original_ptr);
keys[i].size -= keys[i].data - original_ptr;
}
}
template <typename T>
void ColumnStr<T>::deserialize_vec_with_null_map(std::vector<StringRef>& keys,
const size_t num_rows, const uint8_t* null_map) {
for (size_t i = 0; i != num_rows; ++i) {
if (null_map[i] == 0) {
const auto* original_ptr = keys[i].data;
keys[i].data = deserialize_and_insert_from_arena(original_ptr);
keys[i].size -= keys[i].data - original_ptr;
} else {
insert_default();
}
}
sanity_check_simple();
}
template <typename T>
template <bool positive>
struct ColumnStr<T>::less {
const ColumnStr<T>& parent;
explicit less(const ColumnStr<T>& parent_) : parent(parent_) {}
bool operator()(size_t lhs, size_t rhs) const {
int res = memcmp_small_allow_overflow15(
parent.chars.data() + parent.offset_at(lhs), parent.size_at(lhs),
parent.chars.data() + parent.offset_at(rhs), parent.size_at(rhs));
return positive ? (res < 0) : (res > 0);
}
};
template <typename T>
void ColumnStr<T>::get_permutation(bool reverse, size_t limit, int /*nan_direction_hint*/,
IColumn::Permutation& res) const {
size_t s = offsets.size();
res.resize(s);
for (size_t i = 0; i < s; ++i) {
res[i] = i;
}
if (reverse) {
pdqsort(res.begin(), res.end(), less<false>(*this));
} else {
pdqsort(res.begin(), res.end(), less<true>(*this));
}
}
template <typename T>
ColumnPtr ColumnStr<T>::replicate(const IColumn::Offsets& replicate_offsets) const {
size_t col_size = size();
column_match_offsets_size(col_size, replicate_offsets.size());
auto res = ColumnStr<T>::create();
if (0 == col_size) {
return res;
}
Chars& res_chars = res->chars;
auto& res_offsets = res->offsets;
res_chars.reserve(chars.size() / col_size * replicate_offsets.back());
res_offsets.reserve(replicate_offsets.back());
T prev_replicate_offset = 0;
T prev_string_offset = 0;
T current_new_offset = 0;
for (size_t i = 0; i < col_size; ++i) {
size_t size_to_replicate = replicate_offsets[i] - prev_replicate_offset;
size_t string_size = offsets[i] - prev_string_offset;
for (size_t j = 0; j < size_to_replicate; ++j) {
current_new_offset += string_size;
res_offsets.push_back(current_new_offset);
res_chars.resize(res_chars.size() + string_size);
memcpy_small_allow_read_write_overflow15(&res_chars[res_chars.size() - string_size],
&chars[prev_string_offset], string_size);
check_chars_length(res_chars.size(), res_offsets.size());
}
prev_replicate_offset = replicate_offsets[i];
prev_string_offset = offsets[i];
}
check_chars_length(res_chars.size(), res_offsets.size());
sanity_check_simple();
return res;
}
template <typename T>
void ColumnStr<T>::reserve(size_t n) {
offsets.reserve(n);
chars.reserve(n);
sanity_check_simple();
}
template <typename T>
void ColumnStr<T>::resize(size_t n) {
auto origin_size = size();
if (origin_size > n) {
offsets.resize(n);
chars.resize(offsets[n - 1]);
} else if (origin_size < n) {
insert_many_defaults(n - origin_size);
}
sanity_check_simple();
}
template <typename T>
void ColumnStr<T>::sort_column(const ColumnSorter* sorter, EqualFlags& flags,
IColumn::Permutation& perms, EqualRange& range,
bool last_column) const {
sorter->sort_column(static_cast<const ColumnStr<T>&>(*this), flags, perms, range, last_column);
}
template <typename T>
void ColumnStr<T>::compare_internal(size_t rhs_row_id, const IColumn& rhs, int nan_direction_hint,
int direction, std::vector<uint8>& cmp_res,
uint8* __restrict filter) const {
auto sz = offsets.size();
DCHECK(cmp_res.size() == sz);
const auto& cmp_base =
assert_cast<const ColumnStr<T>&, TypeCheckOnRelease::DISABLE>(rhs).get_data_at(
rhs_row_id);
size_t begin = simd::find_zero(cmp_res, 0);
while (begin < sz) {
size_t end = simd::find_one(cmp_res, begin + 1);
for (size_t row_id = begin; row_id < end; row_id++) {
auto value_a = get_data_at(row_id);
// need to covert to unsigned char, orelse the compare semantic is not consistent
// with other member functions, e.g. get_permutation and compare_at,
// and will result wrong result.
int res = memcmp_small_allow_overflow15((Char*)value_a.data, value_a.size,
(Char*)cmp_base.data, cmp_base.size);
cmp_res[row_id] = res != 0;
filter[row_id] = res * direction < 0;
}
begin = simd::find_zero(cmp_res, end + 1);
}
}
template <typename T>
ColumnPtr ColumnStr<T>::convert_column_if_overflow() {
if (std::is_same_v<T, UInt32> && chars.size() > config::string_overflow_size) {
auto new_col = ColumnStr<uint64_t>::create();
const auto length = offsets.size();
std::swap(new_col->get_chars(), chars);
new_col->get_offsets().resize(length);
auto& large_offsets = new_col->get_offsets();
size_t loc = 0;
// TODO: recheck to SIMD the code
// if offset overflow. will be lower than offsets[loc - 1]
while (offsets[loc] >= offsets[loc - 1] && loc < length) {
large_offsets[loc] = offsets[loc];
loc++;
}
while (loc < length) {
large_offsets[loc] = (offsets[loc] - offsets[loc - 1]) + large_offsets[loc - 1];
loc++;
}
offsets.clear();
new_col->sanity_check_simple();
return new_col;
}
return this->get_ptr();
}
template class ColumnStr<uint32_t>;
template class ColumnStr<uint64_t>;
} // namespace doris::vectorized