[feature-wip](array-type) replicate impl for ColumnArray to support join with array column (#9070)

SQL with JOIN and columns ARRAY, will call function ColumnArray::replicate. At this pr,
we implement replicate for ARRAY type, to support SQL like this:
`SELECT count(lo_array),count(d_array),SUM(lo_extendedprice*lo_discount) AS REVENUE FROM  lineorder, date WHERE  lo_orderdate = d_datekey AND d_year = 1993 AND lo_discount BETWEEN 1 AND 3 AND lo_quantity < 25;`
This commit is contained in:
camby
2022-04-20 14:50:34 +08:00
committed by GitHub
parent df3a8545dc
commit a2edc6fd8b
5 changed files with 104 additions and 27 deletions

View File

@ -20,8 +20,6 @@
#include "vec/columns/column_array.h"
#include <string.h> // memcpy
#include "vec/columns/collator.h"
#include "vec/columns/column_const.h"
#include "vec/columns/column_nullable.h"
@ -493,14 +491,9 @@ void ColumnArray::insert_indices_from(const IColumn& src, const int* indices_beg
ColumnPtr ColumnArray::replicate(const Offsets& replicate_offsets) const {
if (replicate_offsets.empty()) return clone_empty();
// keep ColumnUInt8 for ColumnNullable::null_map
if (typeid_cast<const ColumnUInt8*>(data.get()))
return replicate_number<UInt8>(replicate_offsets);
if (typeid_cast<const ColumnUInt16*>(data.get()))
return replicate_number<UInt16>(replicate_offsets);
if (typeid_cast<const ColumnUInt32*>(data.get()))
return replicate_number<UInt32>(replicate_offsets);
if (typeid_cast<const ColumnUInt64*>(data.get()))
return replicate_number<UInt64>(replicate_offsets);
if (typeid_cast<const ColumnInt8*>(data.get()))
return replicate_number<Int8>(replicate_offsets);
if (typeid_cast<const ColumnInt16*>(data.get()))
@ -517,10 +510,40 @@ ColumnPtr ColumnArray::replicate(const Offsets& replicate_offsets) const {
if (typeid_cast<const ColumnConst*>(data.get())) return replicate_const(replicate_offsets);
if (typeid_cast<const ColumnNullable*>(data.get()))
return replicate_nullable(replicate_offsets);
//if (typeid_cast<const ColumnTuple *>(data.get())) return replicateTuple(replicate_offsets);
return replicate_generic(replicate_offsets);
}
void ColumnArray::replicate(const uint32_t* counts, size_t target_size, IColumn& column) const {
size_t col_size = size();
if (col_size == 0) {
return;
}
Offsets replicate_offsets(col_size);
size_t cur_offset = 0;
for (size_t i = 0; i < col_size; ++i) {
cur_offset += counts[i];
replicate_offsets[i] = cur_offset;
}
if (cur_offset != target_size) {
LOG(WARNING) << "ColumnArray replicate input target_size:" << target_size
<< " not equal SUM(counts):" << cur_offset;
return;
}
auto rep_res = replicate(replicate_offsets);
if (!rep_res) {
LOG(WARNING) << "ColumnArray replicate failed, replicate_offsets count="
<< replicate_offsets.size() << ", max=" << replicate_offsets.back();
return;
}
auto& rep_res_arr = typeid_cast<const ColumnArray&>(*rep_res);
ColumnArray& res_arr = typeid_cast<ColumnArray&>(column);
res_arr.data = rep_res_arr.get_data_ptr();
res_arr.offsets = rep_res_arr.get_offsets_ptr();
}
template <typename T>
ColumnPtr ColumnArray::replicate_number(const Offsets& replicate_offsets) const {
size_t col_size = size();