ColumnPtr ColumnArray::indexImpl(const PaddedPODArray<T> & indexes, size_t limit) const
{
if (limit == 0)
return ColumnArray::create(data);
/// Convert indexes to UInt64 in case of overflow.
auto nested_indexes_column = ColumnUInt64::create();
PaddedPODArray<UInt64> & nested_indexes = nested_indexes_column->getData();
nested_indexes.reserve(getOffsets().back());
auto res = ColumnArray::create(data->cloneEmpty());
Offsets & res_offsets = res->getOffsets();
res_offsets.resize(limit);
size_t current_offset = 0;
for (size_t i = 0; i < limit; ++i)
{
for (size_t j = 0; j < sizeAt(indexes[i]); ++j)
nested_indexes.push_back(offsetAt(indexes[i]) + j);
current_offset += sizeAt(indexes[i]);
res_offsets[i] = current_offset;
}
if (current_offset != 0)
res->data = data->index(*nested_indexes_column, current_offset);
return res;
}
int ColumnString::compareAtWithCollation(size_t n, size_t m, const IColumn & rhs_, const Collator & collator) const
{
const ColumnString & rhs = static_cast<const ColumnString &>(rhs_);
return collator.compare(
reinterpret_cast<const char *>(&chars[offsetAt(n)]), sizeAt(n),
reinterpret_cast<const char *>(&rhs.chars[rhs.offsetAt(m)]), rhs.sizeAt(m));
}
void ColumnArray::popBack(size_t n)
{
auto & offsets_data = getOffsets();
size_t nested_n = offsets_data.back() - offsetAt(offsets_data.size() - n);
if (nested_n)
getData().popBack(nested_n);
offsets_data.resize_assume_reserved(offsets_data.size() - n);
}
void ColumnArray::updateHashWithValue(size_t n, SipHash & hash) const
{
size_t array_size = sizeAt(n);
size_t offset = offsetAt(n);
hash.update(array_size);
for (size_t i = 0; i < array_size; ++i)
getData().updateHashWithValue(offset + i, hash);
}
ColumnPtr ColumnArray::filterGeneric(const Filter & filt, ssize_t result_size_hint) const
{
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 std::make_shared<ColumnArray>(data);
Filter nested_filt(getOffsets().back());
for (size_t i = 0; i < size; ++i)
{
if (filt[i])
memset(&nested_filt[offsetAt(i)], 1, sizeAt(i));
else
memset(&nested_filt[offsetAt(i)], 0, sizeAt(i));
}
std::shared_ptr<ColumnArray> res = std::make_shared<ColumnArray>(data);
ssize_t nested_result_size_hint = 0;
if (result_size_hint < 0)
nested_result_size_hint = result_size_hint;
else if (result_size_hint && result_size_hint < 1000000000 && data->size() < 1000000000) /// Избегаем переполнения.
nested_result_size_hint = result_size_hint * data->size() / size;
res->data = data->filter(nested_filt, nested_result_size_hint);
Offsets_t & res_offsets = res->getOffsets();
if (result_size_hint)
res_offsets.reserve(result_size_hint > 0 ? result_size_hint : size);
size_t current_offset = 0;
for (size_t i = 0; i < size; ++i)
{
if (filt[i])
{
current_offset += sizeAt(i);
res_offsets.push_back(current_offset);
}
}
return res;
}
ColumnPtr ColumnArray::filterGeneric(const Filter & filt, ssize_t result_size_hint) const
{
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 ColumnArray::create(data);
Filter nested_filt(getOffsets().back());
for (size_t i = 0; i < size; ++i)
{
if (filt[i])
memset(&nested_filt[offsetAt(i)], 1, sizeAt(i));
else
memset(&nested_filt[offsetAt(i)], 0, sizeAt(i));
}
auto res = ColumnArray::create(data->cloneEmpty());
ssize_t nested_result_size_hint = 0;
if (result_size_hint < 0)
nested_result_size_hint = result_size_hint;
else if (result_size_hint && result_size_hint < 1000000000 && data->size() < 1000000000) /// Avoid overflow.
nested_result_size_hint = result_size_hint * data->size() / size;
res->data = data->filter(nested_filt, nested_result_size_hint);
Offsets & res_offsets = res->getOffsets();
if (result_size_hint)
res_offsets.reserve(result_size_hint > 0 ? result_size_hint : size);
size_t current_offset = 0;
for (size_t i = 0; i < size; ++i)
{
if (filt[i])
{
current_offset += sizeAt(i);
res_offsets.push_back(current_offset);
}
}
return res;
}
void ColumnArray::get(size_t n, Field & res) const
{
size_t offset = offsetAt(n);
size_t size = sizeAt(n);
res = Array(size);
Array & res_arr = DB::get<Array &>(res);
for (size_t i = 0; i < size; ++i)
getData().get(offset + i, res_arr[i]);
}
Field ColumnArray::operator[](size_t n) const
{
size_t offset = offsetAt(n);
size_t size = sizeAt(n);
Array res(size);
for (size_t i = 0; i < size; ++i)
res[i] = getData()[offset + i];
return res;
}
StringRef ColumnArray::serializeValueIntoArena(size_t n, Arena & arena, char const *& begin) const
{
size_t array_size = sizeAt(n);
size_t offset = offsetAt(n);
char * pos = arena.allocContinue(sizeof(array_size), begin);
memcpy(pos, &array_size, sizeof(array_size));
size_t values_size = 0;
for (size_t i = 0; i < array_size; ++i)
values_size += getData().serializeValueIntoArena(offset + i, arena, begin).size;
return StringRef(begin, sizeof(array_size) + values_size);
}
int ColumnArray::compareAt(size_t n, size_t m, const IColumn & rhs_, int nan_direction_hint) const
{
const ColumnArray & rhs = static_cast<const ColumnArray &>(rhs_);
/// Suboptimal
size_t lhs_size = sizeAt(n);
size_t rhs_size = rhs.sizeAt(m);
size_t min_size = std::min(lhs_size, rhs_size);
for (size_t i = 0; i < min_size; ++i)
if (int res = getData().compareAt(offsetAt(n) + i, rhs.offsetAt(m) + i, *rhs.data.get(), nan_direction_hint))
return res;
return lhs_size < rhs_size
? -1
: (lhs_size == rhs_size
? 0
: 1);
}
StringRef ColumnArray::getDataAt(size_t n) const
{
/** Returns the range of memory that covers all elements of the array.
* Works for arrays of fixed length values.
* For arrays of strings and arrays of arrays, the resulting chunk of memory may not be one-to-one correspondence with the elements,
* since it contains only the data laid in succession, but not the offsets.
*/
size_t offset_of_first_elem = offsetAt(n);
StringRef first = getData().getDataAtWithTerminatingZero(offset_of_first_elem);
size_t array_size = sizeAt(n);
if (array_size == 0)
return StringRef(first.data, 0);
size_t offset_of_last_elem = getOffsets()[n] - 1;
StringRef last = getData().getDataAtWithTerminatingZero(offset_of_last_elem);
return StringRef(first.data, last.data + last.size - first.data);
}
ColumnPtr ColumnArray::permute(const Permutation & perm, size_t limit) const
{
size_t size = getOffsets().size();
if (limit == 0)
limit = size;
else
limit = std::min(size, limit);
if (perm.size() < limit)
throw Exception("Size of permutation is less than required.", ErrorCodes::SIZES_OF_COLUMNS_DOESNT_MATCH);
if (limit == 0)
return std::make_shared<ColumnArray>(data);
Permutation nested_perm(getOffsets().back());
std::shared_ptr<ColumnArray> res = std::make_shared<ColumnArray>(data->cloneEmpty());
Offsets_t & res_offsets = res->getOffsets();
res_offsets.resize(limit);
size_t current_offset = 0;
for (size_t i = 0; i < limit; ++i)
{
for (size_t j = 0; j < sizeAt(perm[i]); ++j)
nested_perm[current_offset + j] = offsetAt(perm[i]) + j;
current_offset += sizeAt(perm[i]);
res_offsets[i] = current_offset;
}
if (current_offset != 0)
res->data = data->permute(nested_perm, current_offset);
return res;
}
