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
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
|
#pragma once
#include <Functions/IFunction.h>
#include <Functions/FunctionHelpers.h>
#include <Columns/ColumnsNumber.h>
#include <Columns/ColumnNullable.h>
#include <Common/assert_cast.h>
#include <DataTypes/DataTypeDate.h>
#include <DataTypes/DataTypeDate32.h>
#include <DataTypes/DataTypeDateTime.h>
#include <DataTypes/DataTypeDateTime64.h>
#include <DataTypes/DataTypesNumber.h>
#include <DataTypes/NumberTraits.h>
#include <DataTypes/DataTypeNullable.h>
namespace DB
{
namespace ErrorCodes
{
extern const int ILLEGAL_TYPE_OF_ARGUMENT;
}
template <bool is_first_line_zero>
struct FunctionRunningDifferenceName;
template <>
struct FunctionRunningDifferenceName<true>
{
static constexpr auto name = "runningDifference";
};
template <>
struct FunctionRunningDifferenceName<false>
{
static constexpr auto name = "runningDifferenceStartingWithFirstValue";
};
/** Calculate difference of consecutive values in columns.
* So, result of function depends on partition of data to columns and on order of data in columns.
*/
template <bool is_first_line_zero>
class FunctionRunningDifferenceImpl : public IFunction
{
private:
/// It is possible to track value from previous columns, to calculate continuously across all columns. Not implemented.
template <typename Src, typename Dst>
static NO_SANITIZE_UNDEFINED void process(const PaddedPODArray<Src> & src, PaddedPODArray<Dst> & dst, const NullMap * null_map)
{
size_t size = src.size();
dst.resize(size);
if (size == 0)
return;
/// It is possible to SIMD optimize this loop. By no need for that in practice.
Src prev{};
bool has_prev_value = false;
for (size_t i = 0; i < size; ++i)
{
if (null_map && (*null_map)[i])
{
dst[i] = Dst{};
continue;
}
if (!has_prev_value)
{
dst[i] = is_first_line_zero ? static_cast<Dst>(0) : static_cast<Dst>(src[i]);
prev = src[i];
has_prev_value = true;
}
else
{
auto cur = src[i];
/// Overflow is Ok.
dst[i] = static_cast<Dst>(cur) - prev;
prev = cur;
}
}
}
/// Result type is same as result of subtraction of argument types.
template <typename SrcFieldType>
using DstFieldType = typename NumberTraits::ResultOfSubtraction<SrcFieldType, SrcFieldType>::Type;
/// Call polymorphic lambda with tag argument of concrete field type of src_type.
template <typename F>
void dispatchForSourceType(const IDataType & src_type, F && f) const
{
WhichDataType which(src_type);
if (which.isUInt8())
f(UInt8());
else if (which.isUInt16())
f(UInt16());
else if (which.isUInt32())
f(UInt32());
else if (which.isUInt64())
f(UInt64());
else if (which.isUInt128())
f(UInt128());
else if (which.isUInt256())
f(UInt256());
else if (which.isInt8())
f(Int8());
else if (which.isInt16())
f(Int16());
else if (which.isInt32())
f(Int32());
else if (which.isInt64())
f(Int64());
else if (which.isInt128())
f(Int128());
else if (which.isInt256())
f(Int256());
else if (which.isFloat32())
f(Float32());
else if (which.isFloat64())
f(Float64());
else if (which.isDate())
f(DataTypeDate::FieldType());
else if (which.isDate32())
f(DataTypeDate32::FieldType());
else if (which.isDateTime())
f(DataTypeDateTime::FieldType());
else
throw Exception(ErrorCodes::ILLEGAL_TYPE_OF_ARGUMENT, "Argument for function {} must have numeric type.", getName());
}
public:
static constexpr auto name = FunctionRunningDifferenceName<is_first_line_zero>::name;
static FunctionPtr create(ContextPtr)
{
return std::make_shared<FunctionRunningDifferenceImpl<is_first_line_zero>>();
}
String getName() const override
{
return name;
}
bool isStateful() const override
{
return true;
}
size_t getNumberOfArguments() const override
{
return 1;
}
bool isDeterministic() const override
{
return false;
}
bool isDeterministicInScopeOfQuery() const override
{
return false;
}
bool isSuitableForShortCircuitArgumentsExecution(const DataTypesWithConstInfo & /*arguments*/) const override { return false; }
bool useDefaultImplementationForNulls() const override { return false; }
DataTypePtr getReturnTypeImpl(const DataTypes & arguments) const override
{
DataTypePtr res;
dispatchForSourceType(*removeNullable(arguments[0]), [&](auto field_type_tag)
{
res = std::make_shared<DataTypeNumber<DstFieldType<decltype(field_type_tag)>>>();
});
if (arguments[0]->isNullable())
res = makeNullable(res);
return res;
}
ColumnPtr executeImpl(const ColumnsWithTypeAndName & arguments, const DataTypePtr & result_type, size_t input_rows_count) const override
{
const auto & src = arguments.at(0);
/// When column is constant, its difference is zero.
if (isColumnConst(*src.column))
return result_type->createColumnConstWithDefaultValue(input_rows_count);
auto res_column = removeNullable(result_type)->createColumn();
const auto * src_column = src.column.get();
ColumnPtr null_map_column = nullptr;
const NullMap * null_map = nullptr;
if (const auto * nullable_column = checkAndGetColumn<ColumnNullable>(src_column))
{
src_column = &nullable_column->getNestedColumn();
null_map_column = nullable_column->getNullMapColumnPtr();
null_map = &nullable_column->getNullMapData();
}
dispatchForSourceType(*removeNullable(src.type), [&](auto field_type_tag)
{
using SrcFieldType = decltype(field_type_tag);
process(assert_cast<const ColumnVector<SrcFieldType> &>(*src_column).getData(),
assert_cast<ColumnVector<DstFieldType<SrcFieldType>> &>(*res_column).getData(), null_map);
});
if (null_map_column)
return ColumnNullable::create(std::move(res_column), null_map_column);
else
return res_column;
}
};
}
|
