1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
|
#include <Common/Exception.h>
#include <Columns/ColumnsNumber.h>
#include <DataTypes/DataTypesNumber.h>
#include <Storages/System/StorageSystemNumbers.h>
#include <Storages/SelectQueryInfo.h>
#include <Processors/ISource.h>
#include <QueryPipeline/Pipe.h>
#include <Processors/LimitTransform.h>
namespace DB
{
namespace
{
class NumbersSource : public ISource
{
public:
NumbersSource(UInt64 block_size_, UInt64 offset_, UInt64 step_)
: ISource(createHeader()), block_size(block_size_), next(offset_), step(step_) {}
String getName() const override { return "Numbers"; }
protected:
Chunk generate() override
{
auto column = ColumnUInt64::create(block_size);
ColumnUInt64::Container & vec = column->getData();
size_t curr = next; /// The local variable for some reason works faster (>20%) than member of class.
UInt64 * pos = vec.data(); /// This also accelerates the code.
UInt64 * end = &vec[block_size];
while (pos < end)
*pos++ = curr++;
next += step;
progress(column->size(), column->byteSize());
return { Columns {std::move(column)}, block_size };
}
private:
UInt64 block_size;
UInt64 next;
UInt64 step;
static Block createHeader()
{
return { ColumnWithTypeAndName(ColumnUInt64::create(), std::make_shared<DataTypeUInt64>(), "number") };
}
};
struct NumbersMultiThreadedState
{
std::atomic<UInt64> counter;
explicit NumbersMultiThreadedState(UInt64 offset) : counter(offset) {}
};
using NumbersMultiThreadedStatePtr = std::shared_ptr<NumbersMultiThreadedState>;
class NumbersMultiThreadedSource : public ISource
{
public:
NumbersMultiThreadedSource(NumbersMultiThreadedStatePtr state_, UInt64 block_size_, UInt64 max_counter_)
: ISource(createHeader())
, state(std::move(state_))
, block_size(block_size_)
, max_counter(max_counter_) {}
String getName() const override { return "NumbersMt"; }
protected:
Chunk generate() override
{
if (block_size == 0)
return {};
UInt64 curr = state->counter.fetch_add(block_size, std::memory_order_relaxed);
if (curr >= max_counter)
return {};
if (curr + block_size > max_counter)
block_size = max_counter - curr;
auto column = ColumnUInt64::create(block_size);
ColumnUInt64::Container & vec = column->getData();
UInt64 * pos = vec.data();
UInt64 * end = &vec[block_size];
while (pos < end)
*pos++ = curr++;
progress(column->size(), column->byteSize());
return { Columns {std::move(column)}, block_size };
}
private:
NumbersMultiThreadedStatePtr state;
UInt64 block_size;
UInt64 max_counter;
static Block createHeader()
{
return { ColumnWithTypeAndName(ColumnUInt64::create(), std::make_shared<DataTypeUInt64>(), "number") };
}
};
}
StorageSystemNumbers::StorageSystemNumbers(const StorageID & table_id, bool multithreaded_, std::optional<UInt64> limit_, UInt64 offset_, bool even_distribution_)
: IStorage(table_id), multithreaded(multithreaded_), even_distribution(even_distribution_), limit(limit_), offset(offset_)
{
StorageInMemoryMetadata storage_metadata;
storage_metadata.setColumns(ColumnsDescription({{"number", std::make_shared<DataTypeUInt64>()}}));
setInMemoryMetadata(storage_metadata);
}
Pipe StorageSystemNumbers::read(
const Names & column_names,
const StorageSnapshotPtr & storage_snapshot,
SelectQueryInfo & query_info,
ContextPtr /*context*/,
QueryProcessingStage::Enum /*processed_stage*/,
size_t max_block_size,
size_t num_streams)
{
storage_snapshot->check(column_names);
if (limit && *limit < max_block_size)
{
max_block_size = static_cast<size_t>(*limit);
multithreaded = false;
}
if (!multithreaded)
num_streams = 1;
Pipe pipe;
if (num_streams > 1 && !even_distribution && limit)
{
auto state = std::make_shared<NumbersMultiThreadedState>(offset);
UInt64 max_counter = offset + *limit;
for (size_t i = 0; i < num_streams; ++i)
{
auto source = std::make_shared<NumbersMultiThreadedSource>(state, max_block_size, max_counter);
if (i == 0)
{
auto rows_appr = *limit;
if (query_info.limit > 0 && query_info.limit < rows_appr)
rows_appr = query_info.limit;
source->addTotalRowsApprox(rows_appr);
}
pipe.addSource(std::move(source));
}
return pipe;
}
for (size_t i = 0; i < num_streams; ++i)
{
auto source = std::make_shared<NumbersSource>(max_block_size, offset + i * max_block_size, num_streams * max_block_size);
if (limit && i == 0)
{
auto rows_appr = *limit;
if (query_info.limit > 0 && query_info.limit < rows_appr)
rows_appr = query_info.limit;
source->addTotalRowsApprox(rows_appr);
}
pipe.addSource(std::move(source));
}
if (limit)
{
size_t i = 0;
/// This formula is how to split 'limit' elements to 'num_streams' chunks almost uniformly.
pipe.addSimpleTransform([&](const Block & header)
{
++i;
return std::make_shared<LimitTransform>(
header, *limit * i / num_streams - *limit * (i - 1) / num_streams, 0);
});
}
return pipe;
}
}
|
