-
Notifications
You must be signed in to change notification settings - Fork 493
/
window.go
470 lines (430 loc) · 10.9 KB
/
window.go
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
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
package kapacitor
import (
"errors"
"fmt"
"time"
"github.com/influxdata/kapacitor/edge"
"github.com/influxdata/kapacitor/models"
"github.com/influxdata/kapacitor/pipeline"
)
type WindowNode struct {
node
w *pipeline.WindowNode
}
// Create a new WindowNode, which windows data for a period of time and emits the window.
func newWindowNode(et *ExecutingTask, n *pipeline.WindowNode, d NodeDiagnostic) (*WindowNode, error) {
if n.Period == 0 && n.PeriodCount == 0 {
return nil, errors.New("window node must have either a non zero period or non zero period count")
}
wn := &WindowNode{
w: n,
node: node{Node: n, et: et, diag: d},
}
wn.node.runF = wn.runWindow
return wn, nil
}
func (n *WindowNode) runWindow([]byte) (err error) {
consumer := edge.NewGroupedConsumer(n.ins[0], n)
n.statMap.Set(statCardinalityGauge, consumer.CardinalityVar())
err = consumer.Consume()
return
}
func (n *WindowNode) NewGroup(group edge.GroupInfo, first edge.PointMeta) (edge.Receiver, error) {
r, err := n.newWindow(group, first)
if err != nil {
return nil, err
}
return edge.NewReceiverFromForwardReceiverWithStats(
n.outs,
edge.NewTimedForwardReceiver(n.timer, r),
), nil
}
func (n *WindowNode) DeleteGroup(group models.GroupID) {
// Nothing to do
}
func (n *WindowNode) newWindow(group edge.GroupInfo, first edge.PointMeta) (edge.ForwardReceiver, error) {
switch {
case n.w.Period != 0:
return newWindowByTime(
first.Name(),
first.Time(),
group,
n.w.Period,
n.w.Every,
n.w.AlignFlag,
n.w.FillPeriodFlag,
n.diag,
), nil
case n.w.PeriodCount != 0:
return newWindowByCount(
first.Name(),
group,
int(n.w.PeriodCount),
int(n.w.EveryCount),
n.w.FillPeriodFlag,
n.diag,
), nil
default:
return nil, errors.New("unreachable code, window node should have a non-zero period or period count")
}
}
type windowByTime struct {
name string
group edge.GroupInfo
nextEmit time.Time
buf *windowTimeBuffer
align,
fillPeriod bool
period time.Duration
every time.Duration
diag NodeDiagnostic
}
func newWindowByTime(
name string,
t time.Time,
group edge.GroupInfo,
period,
every time.Duration,
align,
fillPeriod bool,
d NodeDiagnostic,
) *windowByTime {
// Determine nextEmit time.
var nextEmit time.Time
if fillPeriod {
nextEmit = t.Add(period)
if align {
firstPeriod := nextEmit
// Needs to be aligned with Every and be greater than now+Period
nextEmit = nextEmit.Truncate(every)
if !nextEmit.After(firstPeriod) {
// This means we will drop the first few points
nextEmit = nextEmit.Add(every)
}
}
} else {
nextEmit = t.Add(every)
if align {
nextEmit = nextEmit.Truncate(every)
}
}
return &windowByTime{
name: name,
group: group,
nextEmit: nextEmit,
buf: &windowTimeBuffer{diag: d},
align: align,
fillPeriod: fillPeriod,
period: period,
every: every,
diag: d,
}
}
func (w *windowByTime) BeginBatch(edge.BeginBatchMessage) (edge.Message, error) {
return nil, errors.New("window does not support batch data")
}
func (w *windowByTime) BatchPoint(edge.BatchPointMessage) (edge.Message, error) {
return nil, errors.New("window does not support batch data")
}
func (w *windowByTime) EndBatch(edge.EndBatchMessage) (edge.Message, error) {
return nil, errors.New("window does not support batch data")
}
func (w *windowByTime) Barrier(b edge.BarrierMessage) (msg edge.Message, err error) {
if w.every == 0 {
// Since we are emitting every point we can use a right aligned window (oldest, now]
if !b.Time().Before(w.nextEmit) {
// purge old points
oldest := b.Time().Add(-1 * w.period)
w.buf.purge(oldest, false)
// get current batch
msg = w.batch(b.Time())
// Next emit time is now
w.nextEmit = b.Time()
}
} else {
// Since more points can arrive with the same time we need to use a left aligned window [oldest, now).
if !b.Time().Before(w.nextEmit) {
// purge old points
oldest := w.nextEmit.Add(-1 * w.period)
w.buf.purge(oldest, true)
// get current batch
msg = w.batch(w.nextEmit)
// Determine next emit time.
// This is dependent on the current time not the last time we emitted.
w.nextEmit = b.Time().Add(w.every)
if w.align {
w.nextEmit = w.nextEmit.Truncate(w.every)
}
}
}
return
}
func (w *windowByTime) DeleteGroup(d edge.DeleteGroupMessage) (edge.Message, error) {
return d, nil
}
func (w *windowByTime) Done() {}
func (w *windowByTime) Point(p edge.PointMessage) (msg edge.Message, err error) {
if w.every == 0 {
// Insert point before.
w.buf.insert(p)
// Since we are emitting every point we can use a right aligned window (oldest, now]
if !p.Time().Before(w.nextEmit) {
// purge old points
oldest := p.Time().Add(-1 * w.period)
w.buf.purge(oldest, false)
// get current batch
msg = w.batch(p.Time())
// Next emit time is now
w.nextEmit = p.Time()
}
} else {
// Since more points can arrive with the same time we need to use a left aligned window [oldest, now).
if !p.Time().Before(w.nextEmit) {
// purge old points
oldest := w.nextEmit.Add(-1 * w.period)
w.buf.purge(oldest, true)
// get current batch
msg = w.batch(w.nextEmit)
// Determine next emit time.
// This is dependent on the current time not the last time we emitted.
w.nextEmit = p.Time().Add(w.every)
if w.align {
w.nextEmit = w.nextEmit.Truncate(w.every)
}
}
// Insert point after.
w.buf.insert(p)
}
return
}
// batch returns the current window buffer as a batch message.
// TODO(nathanielc): A possible optimization could be to not buffer the data at all if we know that we do not have overlapping windows.
func (w *windowByTime) batch(tmax time.Time) edge.BufferedBatchMessage {
points := w.buf.points()
return edge.NewBufferedBatchMessage(
edge.NewBeginBatchMessage(
w.name,
w.group.Tags,
w.group.Dimensions.ByName,
tmax,
len(points),
),
points,
edge.NewEndBatchMessage(),
)
}
// implements a purpose built ring buffer for the window of points
type windowTimeBuffer struct {
window []edge.PointMessage
start int
stop int
size int
diag NodeDiagnostic
}
// Insert a single point into the buffer.
func (b *windowTimeBuffer) insert(p edge.PointMessage) {
if b.size == cap(b.window) {
//Increase our buffer
c := 2 * (b.size + 1)
w := make([]edge.PointMessage, b.size+1, c)
if b.size == 0 {
//do nothing
} else if b.stop > b.start {
n := copy(w, b.window[b.start:b.stop])
if n != b.size {
panic(fmt.Sprintf("did not copy all the data: copied: %d size: %d start: %d stop: %d\n", n, b.size, b.start, b.stop))
}
} else {
n := 0
n += copy(w, b.window[b.start:])
n += copy(w[b.size-b.start:], b.window[:b.stop])
if n != b.size {
panic(fmt.Sprintf("did not copy all the data: copied: %d size: %d start: %d stop: %d\n", n, b.size, b.start, b.stop))
}
}
b.window = w
b.start = 0
b.stop = b.size
}
// Check if we need to wrap around
if len(b.window) == cap(b.window) && b.stop == len(b.window) {
b.stop = 0
}
// Insert point
if b.stop == len(b.window) {
b.window = append(b.window, p)
} else {
b.window[b.stop] = p
}
b.size++
b.stop++
}
// Purge expired data from the window.
func (b *windowTimeBuffer) purge(oldest time.Time, inclusive bool) {
include := func(t time.Time) bool {
if inclusive {
return !t.Before(oldest)
}
return t.After(oldest)
}
l := len(b.window)
if l == 0 {
return
}
if b.start < b.stop {
for ; b.start < b.stop; b.start++ {
if include(b.window[b.start].Time()) {
break
}
}
b.size = b.stop - b.start
} else {
if include(b.window[l-1].Time()) {
for ; b.start < l; b.start++ {
if include(b.window[b.start].Time()) {
break
}
}
b.size = l - b.start + b.stop
} else {
for b.start = 0; b.start < b.stop; b.start++ {
if include(b.window[b.start].Time()) {
break
}
}
b.size = b.stop - b.start
}
}
}
// Returns a copy of the current buffer.
// TODO(nathanielc): Optimize this function use buffered vs unbuffered batch messages.
func (b *windowTimeBuffer) points() []edge.BatchPointMessage {
if b.size == 0 {
return nil
}
points := make([]edge.BatchPointMessage, b.size)
if b.stop > b.start {
for i, p := range b.window[b.start:b.stop] {
points[i] = edge.BatchPointFromPoint(p)
}
} else {
j := 0
l := len(b.window)
for i := b.start; i < l; i++ {
p := b.window[i]
points[j] = edge.BatchPointFromPoint(p)
j++
}
for i := 0; i < b.stop; i++ {
p := b.window[i]
points[j] = edge.BatchPointFromPoint(p)
j++
}
}
return points
}
type windowByCount struct {
name string
group edge.GroupInfo
buf []edge.BatchPointMessage
start int
stop int
period int
every int
nextEmit int
size int
count int
diag NodeDiagnostic
}
func newWindowByCount(
name string,
group edge.GroupInfo,
period,
every int,
fillPeriod bool,
d NodeDiagnostic,
) *windowByCount {
// Determine the first nextEmit index
nextEmit := every
if fillPeriod {
nextEmit = period
}
return &windowByCount{
name: name,
group: group,
buf: make([]edge.BatchPointMessage, period),
period: period,
every: every,
nextEmit: nextEmit,
diag: d,
}
}
func (w *windowByCount) BeginBatch(edge.BeginBatchMessage) (edge.Message, error) {
return nil, errors.New("window does not support batch data")
}
func (w *windowByCount) BatchPoint(edge.BatchPointMessage) (edge.Message, error) {
return nil, errors.New("window does not support batch data")
}
func (w *windowByCount) EndBatch(edge.EndBatchMessage) (edge.Message, error) {
return nil, errors.New("window does not support batch data")
}
func (w *windowByCount) Barrier(b edge.BarrierMessage) (edge.Message, error) {
//TODO(nathanielc): Implement barrier messages to flush window
return b, nil
}
func (w *windowByCount) DeleteGroup(d edge.DeleteGroupMessage) (edge.Message, error) {
return d, nil
}
func (w *windowByCount) Done() {}
func (w *windowByCount) Point(p edge.PointMessage) (msg edge.Message, err error) {
w.buf[w.stop] = edge.BatchPointFromPoint(p)
w.stop = (w.stop + 1) % w.period
if w.size == w.period {
w.start = (w.start + 1) % w.period
} else {
w.size++
}
w.count++
//Check if its time to emit
if w.count == w.nextEmit {
w.nextEmit += w.every
msg = w.batch()
}
return
}
func (w *windowByCount) batch() edge.BufferedBatchMessage {
points := w.points()
return edge.NewBufferedBatchMessage(
edge.NewBeginBatchMessage(
w.name,
w.group.Tags,
w.group.Dimensions.ByName,
points[len(points)-1].Time(),
len(points),
),
points,
edge.NewEndBatchMessage(),
)
}
// Returns a copy of the current buffer.
func (w *windowByCount) points() []edge.BatchPointMessage {
if w.size == 0 {
return nil
}
points := make([]edge.BatchPointMessage, w.size)
if w.stop > w.start {
copy(points, w.buf[w.start:w.stop])
} else {
j := 0
l := len(w.buf)
for i := w.start; i < l; i++ {
points[j] = w.buf[i]
j++
}
for i := 0; i < w.stop; i++ {
points[j] = w.buf[i]
j++
}
}
return points
}