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
221
222
223
224
225
226
use std::mem::size_of;
use odbc_sys::{Date, Numeric, Time, Timestamp};
use crate::{Bit, DataType};
/// Describes a column of a [`crate::buffers::ColumnarBuffer`].
///
/// While related to the [`crate::DataType`] of the column this is bound to, the Buffer type is
/// different as it does not describe the type of the data source but the format the data is going
/// to be represented in memory. While the data source is often considered to choose the buffer
/// type, the kind of processing which is supposed to be applied to the data may be even more
/// important when choosing a buffer for the cursor type. E.g. if you intend to print a date to
/// standard output, it may be more reasonable to bind it as `Text` rather than `Date`.
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum BufferDesc {
/// Variable sized binary buffer, holding up to `max_bytes` bytes per value.
Binary {
/// Maximum number of bytes per value.
max_bytes: usize,
},
/// Text buffer holding strings with binary length of up to `max_str_len`.
///
/// Consider an upper bound choosing this based on the information in a [`DataType::Varchar`]
/// column. E.g. PostgreSQL may return a field size of several GiB for individual values if a
/// column is specified as `TEXT`, or Microsoft SQL Server may return `0` for a column of type
/// `VARCHAR(max)`. In such situations, if values are truly that large, bulk fetching data is
/// not recommended, but streaming individual fields one by one. Usually though, the actual
/// cells of the table in the database contain much shorter values. The best thing todo is to
/// adapt the database schema to better reflect the actual size of the values. Lacking control
/// over the database schema, you can always choose a smaller buffer size than initializing the
/// buffer in disagreement with the database schema.
Text {
/// Maximum string length. Terminating zero is excluded, i.e. memory for it will be
/// implicitly allocated if required.
max_str_len: usize,
},
/// UTF-16 encoded text buffer holding strings with length of up to `max_str_len`. Length is in
/// terms of 2-Byte characters.
WText {
/// Maximum string length. Terminating zero is excluded, i.e. memory for it will be
/// implicitly allocated if required.
max_str_len: usize,
},
/// 64 bit floating point
F64 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// 32 bit floating point
F32 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Describes a buffer holding [`crate::sys::Date`] values.
Date {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Describes a buffer holding [`crate::sys::Time`] values.
Time {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Describes a buffer holding [`crate::sys::Timestamp`] values.
Timestamp {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Signed 8 Bit integer
I8 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Signed 16 Bit integer
I16 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Signed 32 Bit integer
I32 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Signed 64 Bit integer
I64 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Unsigned 8 Bit integer
U8 {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Can either be zero or one
Bit {
/// This indicates whether or not the buffer will be able to represent NULL values. This
/// will cause an indicator buffer to be bound.
nullable: bool,
},
/// Use [`crate::sys::Numeric`] to represent Numeric values. Note that not all driver support
/// Numeric types. Even if they do, they may not respect the `scale` and `precision` values
/// unless they are explicit set in the Applicatior Parameter Descriptor (APD) for inserting or
/// the Application Row Descriptor (ARD). This currently would require unsafe code. Using text
/// buffers to insert or fetch Numeric values works more reliable.
///
/// In my tests so far using Numeric buffers with PostgreSQL works for both inserting and
/// fetching values. With Microsoft SQL Server, it defaults to scale `0` and can not even be
/// changed by manipulating the ARD / APD via unsafe code for insertion.
///
/// With MariaDB inserting works out of the box, yet fetching does default to scale `0` and
/// would require manipulating the ARD.
Numeric,
}
impl BufferDesc {
pub fn from_data_type(data_type: DataType, nullable: bool) -> Option<Self> {
let buffer_desc = match data_type {
DataType::Numeric { precision, scale }
| DataType::Decimal { precision, scale } if scale == 0 && precision < 3 => BufferDesc::I8 { nullable },
DataType::Numeric { precision, scale }
| DataType::Decimal { precision, scale } if scale == 0 && precision < 10 => BufferDesc::I32 { nullable },
DataType::Numeric { precision, scale }
| DataType::Decimal { precision, scale } if scale == 0 && precision < 19 => BufferDesc::I64 { nullable },
DataType::Integer => BufferDesc::I32 { nullable },
DataType::SmallInt => BufferDesc::I16 { nullable },
DataType::Float { precision: 0..=24 } | DataType::Real => BufferDesc::F32 { nullable },
DataType::Float { precision: 25..=53 } |DataType::Double => BufferDesc::F64 { nullable },
DataType::Date => BufferDesc::Date { nullable },
DataType::Time { precision: 0 } => BufferDesc::Time { nullable },
DataType::Timestamp { precision: _ } => BufferDesc::Timestamp { nullable },
DataType::BigInt => BufferDesc::I64 { nullable },
DataType::TinyInt => BufferDesc::I8 { nullable },
DataType::Bit => BufferDesc::Bit { nullable },
DataType::Varbinary { length }
| DataType::Binary { length }
| DataType::LongVarbinary { length } => length.map(|l| BufferDesc::Binary { max_bytes: l.get() })?,
DataType::Varchar { length }
| DataType::WVarchar { length }
// Currently no special buffers for fixed lengths text implemented.
| DataType::WChar {length }
| DataType::Char { length }
| DataType::WLongVarchar { length }
| DataType::LongVarchar { length } => {
length.map(|length| BufferDesc::Text { max_str_len : length.get() } )?
},
// Specialized buffers for Numeric and decimal are not yet supported.
| DataType::Numeric { precision: _, scale: _ }
| DataType::Decimal { precision: _, scale: _ }
| DataType::Time { precision: _ } => BufferDesc::Text { max_str_len: data_type.display_size().unwrap().get() },
DataType::Unknown
| DataType::Float { precision: _ }
| DataType::Other { data_type: _, column_size: _, decimal_digits: _ } => return None,
};
Some(buffer_desc)
}
/// Element size of buffer if bound as a columnar row. Can be used to estimate memory for
/// columnar bindings.
pub fn bytes_per_row(&self) -> usize {
let size_indicator = |nullable: bool| if nullable { size_of::<isize>() } else { 0 };
match *self {
BufferDesc::Binary { max_bytes: length } => length + size_indicator(true),
BufferDesc::Text { max_str_len } => max_str_len + 1 + size_indicator(true),
BufferDesc::WText { max_str_len } => (max_str_len + 1) * 2 + size_indicator(true),
BufferDesc::F64 { nullable } => size_of::<f64>() + size_indicator(nullable),
BufferDesc::F32 { nullable } => size_of::<f32>() + size_indicator(nullable),
BufferDesc::Date { nullable } => size_of::<Date>() + size_indicator(nullable),
BufferDesc::Time { nullable } => size_of::<Time>() + size_indicator(nullable),
BufferDesc::Timestamp { nullable } => size_of::<Timestamp>() + size_indicator(nullable),
BufferDesc::I8 { nullable } => size_of::<i8>() + size_indicator(nullable),
BufferDesc::I16 { nullable } => size_of::<i16>() + size_indicator(nullable),
BufferDesc::I32 { nullable } => size_of::<i32>() + size_indicator(nullable),
BufferDesc::I64 { nullable } => size_of::<i64>() + size_indicator(nullable),
BufferDesc::U8 { nullable } => size_of::<u8>() + size_indicator(nullable),
BufferDesc::Bit { nullable } => size_of::<Bit>() + size_indicator(nullable),
BufferDesc::Numeric => size_of::<Numeric>(),
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
#[cfg(target_pointer_width = "64")] // Indicator size is platform dependent.
fn bytes_per_row() {
assert_eq!(5 + 8, BufferDesc::Binary { max_bytes: 5 }.bytes_per_row());
assert_eq!(
5 + 1 + 8,
BufferDesc::Text { max_str_len: 5 }.bytes_per_row()
);
assert_eq!(
10 + 2 + 8,
BufferDesc::WText { max_str_len: 5 }.bytes_per_row()
);
assert_eq!(6, BufferDesc::Date { nullable: false }.bytes_per_row());
assert_eq!(6, BufferDesc::Time { nullable: false }.bytes_per_row());
assert_eq!(
16,
BufferDesc::Timestamp { nullable: false }.bytes_per_row()
);
assert_eq!(1, BufferDesc::Bit { nullable: false }.bytes_per_row());
assert_eq!(1 + 8, BufferDesc::Bit { nullable: true }.bytes_per_row());
assert_eq!(4, BufferDesc::F32 { nullable: false }.bytes_per_row());
assert_eq!(8, BufferDesc::F64 { nullable: false }.bytes_per_row());
assert_eq!(1, BufferDesc::I8 { nullable: false }.bytes_per_row());
assert_eq!(2, BufferDesc::I16 { nullable: false }.bytes_per_row());
assert_eq!(4, BufferDesc::I32 { nullable: false }.bytes_per_row());
assert_eq!(8, BufferDesc::I64 { nullable: false }.bytes_per_row());
assert_eq!(1, BufferDesc::U8 { nullable: false }.bytes_per_row());
assert_eq!(19, BufferDesc::Numeric.bytes_per_row());
}
}