/
query_subtask.rs
328 lines (296 loc) · 11.4 KB
/
query_subtask.rs
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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
// This file is part of https://github.com/SpringQL/SpringQL which is licensed under MIT OR Apache-2.0. See file LICENSE-MIT or LICENSE-APACHE for full license details.
pub(super) mod collect_subtask;
pub(super) mod group_aggregate_window_subtask;
pub(super) mod join_subtask;
pub(super) mod projection_subtask;
use std::sync::{Arc, Mutex, MutexGuard};
use rand::{
prelude::{SliceRandom, SmallRng},
SeedableRng,
};
use crate::{
error::Result,
expr_resolver::ExprResolver,
pipeline::{name::ColumnName, stream_model::StreamModel},
stream_engine::{
autonomous_executor::task::window::{
aggregate::AggrWindow, join_window::JoinWindow, panes::pane::join_pane::JoinDir,
},
command::query_plan::{query_plan_operation::LowerOps, QueryPlan},
},
stream_engine::{
autonomous_executor::{
performance_metrics::metrics_update_command::metrics_update_by_task_execution::InQueueMetricsUpdateByCollect,
task::{task_context::TaskContext, tuple::Tuple},
},
SqlValue,
},
stream_engine::{
autonomous_executor::{
performance_metrics::metrics_update_command::metrics_update_by_task_execution::{
InQueueMetricsUpdateByTask, WindowInFlowByWindowTask,
},
row::{column::stream_column::StreamColumns, column_values::ColumnValues, Row},
},
command::query_plan::query_plan_operation::JoinOp,
},
};
use self::{
collect_subtask::CollectSubtask, group_aggregate_window_subtask::GroupAggregateWindowSubtask,
join_subtask::JoinSubtask, projection_subtask::ProjectionSubtask,
};
/// Process input row 1-by-1.
#[derive(Debug)]
pub(in crate::stream_engine::autonomous_executor) struct QuerySubtask {
expr_resolver: ExprResolver,
projection_subtask: ProjectionSubtask,
group_aggr_window_subtask: Option<GroupAggregateWindowSubtask>,
// TODO recursive JOIN
join: Option<(
JoinSubtask,
CollectSubtask, // right stream
)>,
left_collect_subtask: CollectSubtask, // left stream
rng: Mutex<SmallRng>,
}
#[derive(Clone, Debug, new)]
pub(in crate::stream_engine::autonomous_executor) struct SqlValues(Vec<SqlValue>);
impl SqlValues {
/// ```text
/// column_order = (c2, c3, c1)
/// stream_shape = (c1, c2, c3)
///
/// |
/// v
///
/// (fields[1], fields[2], fields[0])
/// ```
///
/// # Panics
///
/// - Tuple fields and column_order have different length.
/// - Type mismatch between `self.fields` (ordered) and `stream_shape`
/// - Duplicate column names in `column_order`
pub(in crate::stream_engine::autonomous_executor) fn into_row(
self,
stream_model: Arc<StreamModel>,
column_order: Vec<ColumnName>,
) -> Row {
assert_eq!(self.0.len(), column_order.len());
let column_values = self.mk_column_values(column_order);
let stream_columns = StreamColumns::new(stream_model, column_values)
.expect("type or shape mismatch? must be checked on pump creation");
Row::new(stream_columns)
}
fn mk_column_values(self, column_order: Vec<ColumnName>) -> ColumnValues {
let mut column_values = ColumnValues::default();
for (column_name, value) in column_order.into_iter().zip(self.0.into_iter()) {
column_values
.insert(column_name, value)
.expect("duplicate column name");
}
column_values
}
}
#[derive(Debug, new)]
pub(in crate::stream_engine::autonomous_executor) struct QuerySubtaskOut {
pub(in crate::stream_engine::autonomous_executor) values_seq: Vec<SqlValues>,
pub(in crate::stream_engine::autonomous_executor) in_queue_metrics_update:
InQueueMetricsUpdateByTask,
}
impl QuerySubtask {
pub(in crate::stream_engine::autonomous_executor) fn new(plan: QueryPlan) -> Self {
let rng =
Mutex::new(SmallRng::from_rng(rand::thread_rng()).expect("this generally won't fail"));
let (left_collect_subtask, join) = Self::subtasks_from_lower_ops(plan.lower_ops);
let group_aggr_window_subtask = plan
.upper_ops
.group_aggr_window
.map(|op| GroupAggregateWindowSubtask::new(op.window_param, op.op_param));
let projection_subtask = ProjectionSubtask::new(plan.upper_ops.projection.expr_labels);
Self {
expr_resolver: plan.expr_resolver,
projection_subtask,
group_aggr_window_subtask,
left_collect_subtask,
join,
rng,
}
}
/// (left collect subtask, Option<(join subtask, right collect subtask)>)
fn subtasks_from_lower_ops(
lower_ops: LowerOps,
) -> (CollectSubtask, Option<(JoinSubtask, CollectSubtask)>) {
match lower_ops.join {
JoinOp::Collect(collect_op) => {
let collect_subtask = CollectSubtask::from_collect_op(collect_op);
(collect_subtask, None)
}
JoinOp::JoinWindow(join_window_op) => {
let left_collect_subtask = CollectSubtask::from_collect_op(join_window_op.left);
let right_collect_subtask = CollectSubtask::from_collect_op(join_window_op.right);
let join_subtask =
JoinSubtask::new(join_window_op.window_param, join_window_op.join_param);
(
left_collect_subtask,
Some((join_subtask, right_collect_subtask)),
)
}
}
}
/// # Returns
///
/// None when input queue does not exist or is empty.
///
/// # Failures
///
/// TODO
pub(in crate::stream_engine::autonomous_executor) fn run(
&self,
context: &TaskContext,
) -> Result<Option<QuerySubtaskOut>> {
match self.run_lower_ops(context) {
None => Ok(None),
Some((lower_tuples, in_queue_metrics_update_by_task)) => {
let (values_seq, in_queue_metrics_update) =
self.run_upper_ops(lower_tuples, in_queue_metrics_update_by_task)?;
Ok(Some(QuerySubtaskOut::new(
values_seq,
in_queue_metrics_update,
)))
}
}
}
fn run_upper_ops(
&self,
tuples: Vec<Tuple>,
in_queue_metrics_update_by_lower: InQueueMetricsUpdateByTask,
) -> Result<(Vec<SqlValues>, InQueueMetricsUpdateByTask)> {
let (values_seq, window_in_flow_upper_total) = tuples.into_iter().fold(
Ok((Vec::new(), WindowInFlowByWindowTask::zero())),
|res, tuple| {
let (mut values_seq_acc, window_in_flow_acc) = res?;
let (mut values_seq, window_in_flow) = self.run_upper_ops_inner(tuple)?;
values_seq_acc.append(&mut values_seq);
Ok((values_seq_acc, window_in_flow_acc + window_in_flow))
},
)?;
let in_queue_metrics_update_by_task = InQueueMetricsUpdateByTask::new(
in_queue_metrics_update_by_lower.by_collect,
Some(window_in_flow_upper_total + in_queue_metrics_update_by_lower.window_in_flow),
);
Ok((values_seq, in_queue_metrics_update_by_task))
}
fn run_upper_ops_inner(
&self,
tuple: Tuple,
) -> Result<(Vec<SqlValues>, WindowInFlowByWindowTask)> {
if let Some(group_aggr_window_subtask) = &self.group_aggr_window_subtask {
let (aggregated_and_grouping_values_seq, window_in_flow) =
group_aggr_window_subtask.run(&self.expr_resolver, tuple);
let values_seq = aggregated_and_grouping_values_seq
.into_iter()
.map(|aggregated_and_grouping_values| {
self.projection_subtask
.run_with_aggr(aggregated_and_grouping_values)
})
.collect::<Result<Vec<_>>>()?;
Ok((values_seq, window_in_flow))
} else {
let values = self
.projection_subtask
.run_without_aggr(&self.expr_resolver, &tuple)?;
Ok((vec![values], WindowInFlowByWindowTask::zero()))
}
}
/// # Returns
///
/// None when input queue does not exist or is empty or JOIN op does not emit output yet.
fn run_lower_ops(
&self,
context: &TaskContext,
) -> Option<(Vec<Tuple>, InQueueMetricsUpdateByTask)> {
match &self.join {
Some((join_subtask, right_collect_subtask)) => self.run_join(
context,
&self.left_collect_subtask,
right_collect_subtask,
join_subtask,
),
None => self
.run_left_collect(context)
.map(|(tuple, metrics_collect)| {
(
vec![tuple],
InQueueMetricsUpdateByTask::new(
metrics_collect,
None, // single collect subtask does not use window yet
),
)
}),
}
}
/// JOIN takes tuples from left or right at a time.
///
/// Left or right is determined randomly and if first candidate does not have tuple to collect, then the other is selected.
fn run_join(
&self,
context: &TaskContext,
left_collect_subtask: &CollectSubtask,
right_collect_subtask: &CollectSubtask,
join_subtask: &JoinSubtask,
) -> Option<(Vec<Tuple>, InQueueMetricsUpdateByTask)> {
self.join_dir_candidates().into_iter().find_map(|dir| {
let collect_subtask = match dir {
JoinDir::Left => left_collect_subtask,
JoinDir::Right => right_collect_subtask,
};
self.run_join_core(context, collect_subtask, join_subtask, dir)
})
}
fn join_dir_candidates(&self) -> [JoinDir; 2] {
let first = [JoinDir::Left, JoinDir::Right]
.choose(&mut *self.rng.lock().expect("rng lock poisoned"))
.expect("not empty");
if first == &JoinDir::Left {
[JoinDir::Left, JoinDir::Right]
} else {
[JoinDir::Right, JoinDir::Left]
}
}
fn run_join_core(
&self,
context: &TaskContext,
collect_subtask: &CollectSubtask,
join_subtask: &JoinSubtask,
join_dir: JoinDir,
) -> Option<(Vec<Tuple>, InQueueMetricsUpdateByTask)> {
collect_subtask
.run(context)
.map(|(tuple, metrics_collect)| {
let (tuples, metrics_join) = join_subtask.run(&self.expr_resolver, tuple, join_dir);
let metrics = InQueueMetricsUpdateByTask::new(metrics_collect, Some(metrics_join));
(tuples, metrics)
})
}
fn run_left_collect(
&self,
context: &TaskContext,
) -> Option<(Tuple, InQueueMetricsUpdateByCollect)> {
self.left_collect_subtask.run(context)
}
pub(in crate::stream_engine::autonomous_executor) fn get_aggr_window_mut(
&self,
) -> Option<MutexGuard<AggrWindow>> {
self.group_aggr_window_subtask
.as_ref()
.map(|subtask| subtask.get_window_mut())
}
pub(in crate::stream_engine::autonomous_executor) fn get_join_window_mut(
&self,
) -> Option<MutexGuard<JoinWindow>> {
self.join
.as_ref()
.map(|(subtask, _)| subtask.get_window_mut())
}
}