mirror of
https://github.com/ClickHouse/ClickHouse.git
synced 2024-11-23 16:12:01 +00:00
218 lines
6.1 KiB
C++
218 lines
6.1 KiB
C++
#pragma once
|
||
|
||
#include <string.h>
|
||
|
||
#include <DB/Columns/ColumnArray.h>
|
||
#include <DB/Columns/ColumnsNumber.h>
|
||
|
||
|
||
namespace DB
|
||
{
|
||
|
||
/** Cтолбeц значений типа "строка".
|
||
* Отличается от массива UInt8 только получением элемента (в виде String, а не Array)
|
||
*/
|
||
class ColumnString : public ColumnArray
|
||
{
|
||
private:
|
||
ColumnUInt8::Container_t & char_data;
|
||
|
||
public:
|
||
/** Создать пустой столбец строк, с типом значений */
|
||
ColumnString()
|
||
: ColumnArray(new ColumnUInt8()),
|
||
char_data(dynamic_cast<ColumnUInt8 &>(*data).getData())
|
||
{
|
||
}
|
||
|
||
std::string getName() const { return "ColumnString"; }
|
||
|
||
ColumnPtr cloneEmpty() const
|
||
{
|
||
return new ColumnString;
|
||
}
|
||
|
||
Field operator[](size_t n) const
|
||
{
|
||
size_t offset = n == 0 ? 0 : getOffsets()[n - 1];
|
||
size_t size = getOffsets()[n] - offset - 1;
|
||
const char * s = reinterpret_cast<const char *>(&dynamic_cast<const ColumnUInt8 &>(*data).getData()[offset]);
|
||
return String(s, size);
|
||
}
|
||
|
||
StringRef getDataAt(size_t n) const
|
||
{
|
||
return StringRef(&char_data[offsetAt(n)], sizeAt(n));
|
||
}
|
||
|
||
void insert(const Field & x)
|
||
{
|
||
const String & s = boost::get<const String &>(x);
|
||
size_t old_size = char_data.size();
|
||
size_t size_to_append = s.size() + 1;
|
||
|
||
char_data.resize(old_size + size_to_append);
|
||
memcpy(&char_data[old_size], s.c_str(), size_to_append);
|
||
getOffsets().push_back((getOffsets().size() == 0 ? 0 : getOffsets().back()) + size_to_append);
|
||
}
|
||
|
||
void insertFrom(const IColumn & src_, size_t n)
|
||
{
|
||
const ColumnString & src = static_cast<const ColumnString &>(src_);
|
||
size_t old_size = char_data.size();
|
||
size_t size_to_append = src.sizeAt(n);
|
||
size_t offset = src.offsetAt(n);
|
||
|
||
char_data.resize(old_size + size_to_append);
|
||
memcpy(&char_data[old_size], &src.char_data[offset], size_to_append);
|
||
getOffsets().push_back((getOffsets().size() == 0 ? 0 : getOffsets().back()) + size_to_append);
|
||
}
|
||
|
||
void filter(const Filter & filt)
|
||
{
|
||
size_t size = getOffsets().size();
|
||
if (size != filt.size())
|
||
throw Exception("Size of filter doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
|
||
|
||
if (size == 0)
|
||
return;
|
||
|
||
ColumnUInt8::Container_t tmp_chars;
|
||
Offsets_t tmp_offsets;
|
||
tmp_chars.reserve(char_data.size());
|
||
tmp_offsets.reserve(size);
|
||
|
||
Offset_t current_new_offset = 0;
|
||
|
||
for (size_t i = 0; i < size; ++i)
|
||
{
|
||
if (!filt[i])
|
||
continue;
|
||
|
||
size_t string_offset = i == 0 ? 0 : getOffsets()[i - 1];
|
||
size_t string_size = getOffsets()[i] - string_offset;
|
||
|
||
current_new_offset += string_size;
|
||
tmp_offsets.push_back(current_new_offset);
|
||
|
||
tmp_chars.resize(tmp_chars.size() + string_size);
|
||
memcpy(&tmp_chars[tmp_chars.size() - string_size], &char_data[string_offset], string_size);
|
||
}
|
||
|
||
tmp_chars.swap(char_data);
|
||
tmp_offsets.swap(getOffsets());
|
||
}
|
||
|
||
void permute(const Permutation & perm)
|
||
{
|
||
size_t size = getOffsets().size();
|
||
if (size != perm.size())
|
||
throw Exception("Size of permutation doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
|
||
|
||
if (size == 0)
|
||
return;
|
||
|
||
ColumnUInt8::Container_t tmp_chars(char_data.size());
|
||
Offsets_t tmp_offsets(size);
|
||
|
||
Offset_t current_new_offset = 0;
|
||
|
||
for (size_t i = 0; i < size; ++i)
|
||
{
|
||
size_t j = perm[i];
|
||
size_t string_offset = j == 0 ? 0 : getOffsets()[j - 1];
|
||
size_t string_size = getOffsets()[j] - string_offset;
|
||
|
||
memcpy(&tmp_chars[current_new_offset], &char_data[string_offset], string_size);
|
||
|
||
current_new_offset += string_size;
|
||
tmp_offsets[i] = current_new_offset;
|
||
}
|
||
|
||
tmp_chars.swap(char_data);
|
||
tmp_offsets.swap(getOffsets());
|
||
}
|
||
|
||
void insertDefault()
|
||
{
|
||
char_data.push_back(0);
|
||
getOffsets().push_back(getOffsets().size() == 0 ? 1 : (getOffsets().back() + 1));
|
||
}
|
||
|
||
int compareAt(size_t n, size_t m, const IColumn & rhs_) const
|
||
{
|
||
const ColumnString & rhs = static_cast<const ColumnString &>(rhs_);
|
||
|
||
/** Для производительности, строки сравниваются до первого нулевого байта.
|
||
* (если нулевой байт в середине строки, то то, что после него - игнорируется)
|
||
* Замечу, что завершающий нулевой байт всегда есть.
|
||
*/
|
||
return strcmp(
|
||
reinterpret_cast<const char *>(&char_data[offsetAt(n)]),
|
||
reinterpret_cast<const char *>(&rhs.char_data[rhs.offsetAt(m)]));
|
||
}
|
||
|
||
struct less
|
||
{
|
||
const ColumnString & parent;
|
||
less(const ColumnString & parent_) : parent(parent_) {}
|
||
bool operator()(size_t lhs, size_t rhs) const
|
||
{
|
||
return 0 > strcmp(
|
||
reinterpret_cast<const char *>(&parent.char_data[parent.offsetAt(lhs)]),
|
||
reinterpret_cast<const char *>(&parent.char_data[parent.offsetAt(rhs)]));
|
||
}
|
||
};
|
||
|
||
Permutation getPermutation() const
|
||
{
|
||
size_t s = getOffsets().size();
|
||
Permutation res(s);
|
||
for (size_t i = 0; i < s; ++i)
|
||
res[i] = i;
|
||
|
||
std::sort(res.begin(), res.end(), less(*this));
|
||
return res;
|
||
}
|
||
|
||
void replicate(const Offsets_t & replicate_offsets)
|
||
{
|
||
size_t col_size = size();
|
||
if (col_size != replicate_offsets.size())
|
||
throw Exception("Size of offsets doesn't match size of column.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
|
||
|
||
ColumnUInt8::Container_t tmp_chars;
|
||
Offsets_t tmp_offsets;
|
||
tmp_chars.reserve(char_data.size() / col_size * replicate_offsets.back());
|
||
tmp_offsets.reserve(replicate_offsets.back());
|
||
|
||
Offset_t prev_replicate_offset = 0;
|
||
Offset_t prev_string_offset = 0;
|
||
Offset_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 = getOffsets()[i] - prev_string_offset;
|
||
|
||
for (size_t j = 0; j < size_to_replicate; ++j)
|
||
{
|
||
current_new_offset += string_size;
|
||
tmp_offsets.push_back(current_new_offset);
|
||
|
||
tmp_chars.resize(tmp_chars.size() + string_size);
|
||
memcpy(&tmp_chars[tmp_chars.size() - string_size], &char_data[prev_string_offset], string_size);
|
||
}
|
||
|
||
prev_replicate_offset = replicate_offsets[i];
|
||
prev_string_offset = getOffsets()[i];
|
||
}
|
||
|
||
tmp_chars.swap(char_data);
|
||
tmp_offsets.swap(getOffsets());
|
||
}
|
||
};
|
||
|
||
|
||
}
|