| 1 |
package inmem |
| 2 |
|
| 3 |
import ( |
| 4 |
"bytes" |
| 5 |
"fmt" |
| 6 |
"regexp" |
| 7 |
"sort" |
| 8 |
"sync" |
| 9 |
"unsafe" |
| 10 |
|
| 11 |
"github.com/influxdata/influxdb/models" |
| 12 |
"github.com/influxdata/influxdb/pkg/bytesutil" |
| 13 |
"github.com/influxdata/influxdb/pkg/radix" |
| 14 |
"github.com/influxdata/influxdb/query" |
| 15 |
"github.com/influxdata/influxdb/tsdb" |
| 16 |
"github.com/influxdata/influxql" |
| 17 |
) |
| 18 |
|
| 19 |
// Measurement represents a collection of time series in a database. It also |
| 20 |
// contains in memory structures for indexing tags. Exported functions are |
| 21 |
// goroutine safe while un-exported functions assume the caller will use the |
| 22 |
// appropriate locks. |
| 23 |
type measurement struct { |
| 24 |
Database string |
| 25 |
Name string `json:"name,omitempty"` |
| 26 |
NameBytes []byte // cached version as []byte |
| 27 |
|
| 28 |
mu sync.RWMutex |
| 29 |
fieldNames map[string]struct{} |
| 30 |
|
| 31 |
// in-memory index fields |
| 32 |
seriesByID map[uint64]*series // lookup table for series by their id |
| 33 |
seriesByTagKeyValue map[string]*tagKeyValue // map from tag key to value to sorted set of series ids |
| 34 |
|
| 35 |
// lazyily created sorted series IDs |
| 36 |
sortedSeriesIDs seriesIDs // sorted list of series IDs in this measurement |
| 37 |
|
| 38 |
// Indicates whether the seriesByTagKeyValueMap needs to be rebuilt as it contains deleted series |
| 39 |
// that waste memory. |
| 40 |
dirty bool |
| 41 |
} |
| 42 |
|
| 43 |
// newMeasurement allocates and initializes a new Measurement. |
| 44 |
func newMeasurement(database, name string) *measurement { |
| 45 |
return &measurement{ |
| 46 |
Database: database, |
| 47 |
Name: name, |
| 48 |
NameBytes: []byte(name), |
| 49 |
|
| 50 |
fieldNames: make(map[string]struct{}), |
| 51 |
seriesByID: make(map[uint64]*series), |
| 52 |
seriesByTagKeyValue: make(map[string]*tagKeyValue), |
| 53 |
} |
| 54 |
} |
| 55 |
|
| 56 |
// bytes estimates the memory footprint of this measurement, in bytes. |
| 57 |
func (m *measurement) bytes() int { |
| 58 |
var b int |
| 59 |
m.mu.RLock() |
| 60 |
b += int(unsafe.Sizeof(m.Database)) + len(m.Database) |
| 61 |
b += int(unsafe.Sizeof(m.Name)) + len(m.Name) |
| 62 |
if m.NameBytes != nil { |
| 63 |
b += int(unsafe.Sizeof(m.NameBytes)) + len(m.NameBytes) |
| 64 |
} |
| 65 |
b += 24 // 24 bytes for m.mu RWMutex |
| 66 |
b += int(unsafe.Sizeof(m.fieldNames)) |
| 67 |
for fieldName := range m.fieldNames { |
| 68 |
b += int(unsafe.Sizeof(fieldName)) + len(fieldName) |
| 69 |
} |
| 70 |
b += int(unsafe.Sizeof(m.seriesByID)) |
| 71 |
for k, v := range m.seriesByID { |
| 72 |
b += int(unsafe.Sizeof(k)) |
| 73 |
b += int(unsafe.Sizeof(v)) |
| 74 |
// Do not count footprint of each series, to avoid double-counting in Index.bytes(). |
| 75 |
} |
| 76 |
b += int(unsafe.Sizeof(m.seriesByTagKeyValue)) |
| 77 |
for k, v := range m.seriesByTagKeyValue { |
| 78 |
b += int(unsafe.Sizeof(k)) + len(k) |
| 79 |
b += int(unsafe.Sizeof(v)) + v.bytes() |
| 80 |
} |
| 81 |
b += int(unsafe.Sizeof(m.sortedSeriesIDs)) |
| 82 |
for _, seriesID := range m.sortedSeriesIDs { |
| 83 |
b += int(unsafe.Sizeof(seriesID)) |
| 84 |
} |
| 85 |
b += int(unsafe.Sizeof(m.dirty)) |
| 86 |
m.mu.RUnlock() |
| 87 |
return b |
| 88 |
} |
| 89 |
|
| 90 |
// Authorized determines if this Measurement is authorized to be read, according |
| 91 |
// to the provided Authorizer. A measurement is authorized to be read if at |
| 92 |
// least one undeleted series from the measurement is authorized to be read. |
| 93 |
func (m *measurement) Authorized(auth query.Authorizer) bool { |
| 94 |
// Note(edd): the cost of this check scales linearly with the number of series |
| 95 |
// belonging to a measurement, which means it may become expensive when there |
| 96 |
// are large numbers of series on a measurement. |
| 97 |
// |
| 98 |
// In the future we might want to push the set of series down into the |
| 99 |
// authorizer, but that will require an API change. |
| 100 |
for _, s := range m.SeriesByIDMap() { |
| 101 |
if s != nil && s.Deleted() { |
| 102 |
continue |
| 103 |
} |
| 104 |
|
| 105 |
if query.AuthorizerIsOpen(auth) || auth.AuthorizeSeriesRead(m.Database, m.NameBytes, s.Tags) { |
| 106 |
return true |
| 107 |
} |
| 108 |
} |
| 109 |
return false |
| 110 |
} |
| 111 |
|
| 112 |
func (m *measurement) HasField(name string) bool { |
| 113 |
m.mu.RLock() |
| 114 |
_, hasField := m.fieldNames[name] |
| 115 |
m.mu.RUnlock() |
| 116 |
return hasField |
| 117 |
} |
| 118 |
|
| 119 |
// SeriesByID returns a series by identifier. |
| 120 |
func (m *measurement) SeriesByID(id uint64) *series { |
| 121 |
m.mu.RLock() |
| 122 |
defer m.mu.RUnlock() |
| 123 |
return m.seriesByID[id] |
| 124 |
} |
| 125 |
|
| 126 |
// SeriesByIDMap returns the internal seriesByID map. |
| 127 |
func (m *measurement) SeriesByIDMap() map[uint64]*series { |
| 128 |
m.mu.RLock() |
| 129 |
defer m.mu.RUnlock() |
| 130 |
return m.seriesByID |
| 131 |
} |
| 132 |
|
| 133 |
// SeriesByIDSlice returns a list of series by identifiers. |
| 134 |
func (m *measurement) SeriesByIDSlice(ids []uint64) []*series { |
| 135 |
m.mu.RLock() |
| 136 |
defer m.mu.RUnlock() |
| 137 |
a := make([]*series, len(ids)) |
| 138 |
for i, id := range ids { |
| 139 |
a[i] = m.seriesByID[id] |
| 140 |
} |
| 141 |
return a |
| 142 |
} |
| 143 |
|
| 144 |
// AppendSeriesKeysByID appends keys for a list of series ids to a buffer. |
| 145 |
func (m *measurement) AppendSeriesKeysByID(dst []string, ids []uint64) []string { |
| 146 |
m.mu.RLock() |
| 147 |
defer m.mu.RUnlock() |
| 148 |
for _, id := range ids { |
| 149 |
if s := m.seriesByID[id]; s != nil && !s.Deleted() { |
| 150 |
dst = append(dst, s.Key) |
| 151 |
} |
| 152 |
} |
| 153 |
return dst |
| 154 |
} |
| 155 |
|
| 156 |
// SeriesKeysByID returns the a list of keys for a set of ids. |
| 157 |
func (m *measurement) SeriesKeysByID(ids seriesIDs) [][]byte { |
| 158 |
m.mu.RLock() |
| 159 |
defer m.mu.RUnlock() |
| 160 |
keys := make([][]byte, 0, len(ids)) |
| 161 |
for _, id := range ids { |
| 162 |
s := m.seriesByID[id] |
| 163 |
if s == nil || s.Deleted() { |
| 164 |
continue |
| 165 |
} |
| 166 |
keys = append(keys, []byte(s.Key)) |
| 167 |
} |
| 168 |
|
| 169 |
if !bytesutil.IsSorted(keys) { |
| 170 |
bytesutil.Sort(keys) |
| 171 |
} |
| 172 |
|
| 173 |
return keys |
| 174 |
} |
| 175 |
|
| 176 |
// SeriesKeys returns the keys of every series in this measurement |
| 177 |
func (m *measurement) SeriesKeys() [][]byte { |
| 178 |
m.mu.RLock() |
| 179 |
defer m.mu.RUnlock() |
| 180 |
keys := make([][]byte, 0, len(m.seriesByID)) |
| 181 |
for _, s := range m.seriesByID { |
| 182 |
if s.Deleted() { |
| 183 |
continue |
| 184 |
} |
| 185 |
keys = append(keys, []byte(s.Key)) |
| 186 |
} |
| 187 |
|
| 188 |
if !bytesutil.IsSorted(keys) { |
| 189 |
bytesutil.Sort(keys) |
| 190 |
} |
| 191 |
|
| 192 |
return keys |
| 193 |
} |
| 194 |
|
| 195 |
func (m *measurement) SeriesIDs() seriesIDs { |
| 196 |
m.mu.RLock() |
| 197 |
if len(m.sortedSeriesIDs) == len(m.seriesByID) { |
| 198 |
s := m.sortedSeriesIDs |
| 199 |
m.mu.RUnlock() |
| 200 |
return s |
| 201 |
} |
| 202 |
m.mu.RUnlock() |
| 203 |
|
| 204 |
m.mu.Lock() |
| 205 |
if len(m.sortedSeriesIDs) == len(m.seriesByID) { |
| 206 |
s := m.sortedSeriesIDs |
| 207 |
m.mu.Unlock() |
| 208 |
return s |
| 209 |
} |
| 210 |
|
| 211 |
m.sortedSeriesIDs = m.sortedSeriesIDs[:0] |
| 212 |
if cap(m.sortedSeriesIDs) < len(m.seriesByID) { |
| 213 |
m.sortedSeriesIDs = make(seriesIDs, 0, len(m.seriesByID)) |
| 214 |
} |
| 215 |
|
| 216 |
for k, v := range m.seriesByID { |
| 217 |
if v.Deleted() { |
| 218 |
continue |
| 219 |
} |
| 220 |
m.sortedSeriesIDs = append(m.sortedSeriesIDs, k) |
| 221 |
} |
| 222 |
sort.Sort(m.sortedSeriesIDs) |
| 223 |
s := m.sortedSeriesIDs |
| 224 |
m.mu.Unlock() |
| 225 |
return s |
| 226 |
} |
| 227 |
|
| 228 |
// HasTagKey returns true if at least one series in this measurement has written a value for the passed in tag key |
| 229 |
func (m *measurement) HasTagKey(k string) bool { |
| 230 |
m.mu.RLock() |
| 231 |
defer m.mu.RUnlock() |
| 232 |
_, hasTag := m.seriesByTagKeyValue[k] |
| 233 |
return hasTag |
| 234 |
} |
| 235 |
|
| 236 |
func (m *measurement) HasTagKeyValue(k, v []byte) bool { |
| 237 |
m.mu.RLock() |
| 238 |
defer m.mu.RUnlock() |
| 239 |
return m.seriesByTagKeyValue[string(k)].Contains(string(v)) |
| 240 |
} |
| 241 |
|
| 242 |
// HasSeries returns true if there is at least 1 series under this measurement. |
| 243 |
func (m *measurement) HasSeries() bool { |
| 244 |
m.mu.RLock() |
| 245 |
defer m.mu.RUnlock() |
| 246 |
return len(m.seriesByID) > 0 |
| 247 |
} |
| 248 |
|
| 249 |
// CardinalityBytes returns the number of values associated with the given tag key. |
| 250 |
func (m *measurement) CardinalityBytes(key []byte) int { |
| 251 |
m.mu.RLock() |
| 252 |
defer m.mu.RUnlock() |
| 253 |
return m.seriesByTagKeyValue[string(key)].Cardinality() |
| 254 |
} |
| 255 |
|
| 256 |
// AddSeries adds a series to the measurement's index. |
| 257 |
// It returns true if the series was added successfully or false if the series was already present. |
| 258 |
func (m *measurement) AddSeries(s *series) bool { |
| 259 |
if s == nil { |
| 260 |
return false |
| 261 |
} |
| 262 |
|
| 263 |
m.mu.RLock() |
| 264 |
if m.seriesByID[s.ID] != nil { |
| 265 |
m.mu.RUnlock() |
| 266 |
return false |
| 267 |
} |
| 268 |
m.mu.RUnlock() |
| 269 |
|
| 270 |
m.mu.Lock() |
| 271 |
defer m.mu.Unlock() |
| 272 |
|
| 273 |
if m.seriesByID[s.ID] != nil { |
| 274 |
return false |
| 275 |
} |
| 276 |
|
| 277 |
m.seriesByID[s.ID] = s |
| 278 |
|
| 279 |
if len(m.seriesByID) == 1 || (len(m.sortedSeriesIDs) == len(m.seriesByID)-1 && s.ID > m.sortedSeriesIDs[len(m.sortedSeriesIDs)-1]) { |
| 280 |
m.sortedSeriesIDs = append(m.sortedSeriesIDs, s.ID) |
| 281 |
} |
| 282 |
|
| 283 |
// add this series id to the tag index on the measurement |
| 284 |
for _, t := range s.Tags { |
| 285 |
valueMap := m.seriesByTagKeyValue[string(t.Key)] |
| 286 |
if valueMap == nil { |
| 287 |
valueMap = newTagKeyValue() |
| 288 |
m.seriesByTagKeyValue[string(t.Key)] = valueMap |
| 289 |
} |
| 290 |
valueMap.InsertSeriesIDByte(t.Value, s.ID) |
| 291 |
} |
| 292 |
|
| 293 |
return true |
| 294 |
} |
| 295 |
|
| 296 |
// DropSeries removes a series from the measurement's index. |
| 297 |
func (m *measurement) DropSeries(series *series) { |
| 298 |
seriesID := series.ID |
| 299 |
m.mu.Lock() |
| 300 |
defer m.mu.Unlock() |
| 301 |
|
| 302 |
// Existence check before delete here to clean up the caching/indexing only when needed |
| 303 |
if _, ok := m.seriesByID[seriesID]; !ok { |
| 304 |
return |
| 305 |
} |
| 306 |
delete(m.seriesByID, seriesID) |
| 307 |
|
| 308 |
// clear our lazily sorted set of ids |
| 309 |
m.sortedSeriesIDs = m.sortedSeriesIDs[:0] |
| 310 |
|
| 311 |
// Mark that this measurements tagValue map has stale entries that need to be rebuilt. |
| 312 |
m.dirty = true |
| 313 |
} |
| 314 |
|
| 315 |
func (m *measurement) Rebuild() *measurement { |
| 316 |
m.mu.RLock() |
| 317 |
|
| 318 |
// Nothing needs to be rebuilt. |
| 319 |
if !m.dirty { |
| 320 |
m.mu.RUnlock() |
| 321 |
return m |
| 322 |
} |
| 323 |
|
| 324 |
// Create a new measurement from the state of the existing measurement |
| 325 |
nm := newMeasurement(m.Database, string(m.NameBytes)) |
| 326 |
nm.fieldNames = m.fieldNames |
| 327 |
m.mu.RUnlock() |
| 328 |
|
| 329 |
// Re-add each series to allow the measurement indexes to get re-created. If there were |
| 330 |
// deletes, the existing measurement may have references to deleted series that need to be |
| 331 |
// expunged. Note: we're NOT using SeriesIDs which returns the series in sorted order because |
| 332 |
// we need to do this under a write lock to prevent races. The series are added in sorted |
| 333 |
// order to prevent resorting them again after they are all re-added. |
| 334 |
m.mu.Lock() |
| 335 |
defer m.mu.Unlock() |
| 336 |
|
| 337 |
for k, v := range m.seriesByID { |
| 338 |
if v.Deleted() { |
| 339 |
continue |
| 340 |
} |
| 341 |
m.sortedSeriesIDs = append(m.sortedSeriesIDs, k) |
| 342 |
} |
| 343 |
sort.Sort(m.sortedSeriesIDs) |
| 344 |
|
| 345 |
for _, id := range m.sortedSeriesIDs { |
| 346 |
if s := m.seriesByID[id]; s != nil { |
| 347 |
// Explicitly set the new measurement on the series. |
| 348 |
s.Measurement = nm |
| 349 |
nm.AddSeries(s) |
| 350 |
} |
| 351 |
} |
| 352 |
return nm |
| 353 |
} |
| 354 |
|
| 355 |
// filters walks the where clause of a select statement and returns a map with all series ids |
| 356 |
// matching the where clause and any filter expression that should be applied to each |
| 357 |
func (m *measurement) filters(condition influxql.Expr) ([]uint64, map[uint64]influxql.Expr, error) { |
| 358 |
if condition == nil { |
| 359 |
return m.SeriesIDs(), nil, nil |
| 360 |
} |
| 361 |
return m.WalkWhereForSeriesIds(condition) |
| 362 |
} |
| 363 |
|
| 364 |
// TagSets returns the unique tag sets that exist for the given tag keys. This is used to determine |
| 365 |
// what composite series will be created by a group by. i.e. "group by region" should return: |
| 366 |
// {"region":"uswest"}, {"region":"useast"} |
| 367 |
// or region, service returns |
| 368 |
// {"region": "uswest", "service": "redis"}, {"region": "uswest", "service": "mysql"}, etc... |
| 369 |
// This will also populate the TagSet objects with the series IDs that match each tagset and any |
| 370 |
// influx filter expression that goes with the series |
| 371 |
// TODO: this shouldn't be exported. However, until tx.go and the engine get refactored into tsdb, we need it. |
| 372 |
func (m *measurement) TagSets(shardSeriesIDs *tsdb.SeriesIDSet, opt query.IteratorOptions) ([]*query.TagSet, error) { |
| 373 |
// get the unique set of series ids and the filters that should be applied to each |
| 374 |
ids, filters, err := m.filters(opt.Condition) |
| 375 |
if err != nil { |
| 376 |
return nil, err |
| 377 |
} |
| 378 |
|
| 379 |
var dims []string |
| 380 |
if len(opt.Dimensions) > 0 { |
| 381 |
dims = make([]string, len(opt.Dimensions)) |
| 382 |
copy(dims, opt.Dimensions) |
| 383 |
sort.Strings(dims) |
| 384 |
} |
| 385 |
|
| 386 |
m.mu.RLock() |
| 387 |
// For every series, get the tag values for the requested tag keys i.e. dimensions. This is the |
| 388 |
// TagSet for that series. Series with the same TagSet are then grouped together, because for the |
| 389 |
// purpose of GROUP BY they are part of the same composite series. |
| 390 |
tagSets := make(map[string]*query.TagSet, 64) |
| 391 |
var seriesN int |
| 392 |
for _, id := range ids { |
| 393 |
// Abort if the query was killed |
| 394 |
select { |
| 395 |
case <-opt.InterruptCh: |
| 396 |
m.mu.RUnlock() |
| 397 |
return nil, query.ErrQueryInterrupted |
| 398 |
default: |
| 399 |
} |
| 400 |
|
| 401 |
if opt.MaxSeriesN > 0 && seriesN > opt.MaxSeriesN { |
| 402 |
m.mu.RUnlock() |
| 403 |
return nil, fmt.Errorf("max-select-series limit exceeded: (%d/%d)", seriesN, opt.MaxSeriesN) |
| 404 |
} |
| 405 |
|
| 406 |
s := m.seriesByID[id] |
| 407 |
if s == nil || s.Deleted() || !shardSeriesIDs.Contains(id) { |
| 408 |
continue |
| 409 |
} |
| 410 |
|
| 411 |
if opt.Authorizer != nil && !opt.Authorizer.AuthorizeSeriesRead(m.Database, m.NameBytes, s.Tags) { |
| 412 |
continue |
| 413 |
} |
| 414 |
|
| 415 |
var tagsAsKey []byte |
| 416 |
if len(dims) > 0 { |
| 417 |
tagsAsKey = tsdb.MakeTagsKey(dims, s.Tags) |
| 418 |
} |
| 419 |
|
| 420 |
tagSet := tagSets[string(tagsAsKey)] |
| 421 |
if tagSet == nil { |
| 422 |
// This TagSet is new, create a new entry for it. |
| 423 |
tagSet = &query.TagSet{ |
| 424 |
Tags: nil, |
| 425 |
Key: tagsAsKey, |
| 426 |
} |
| 427 |
tagSets[string(tagsAsKey)] = tagSet |
| 428 |
} |
| 429 |
// Associate the series and filter with the Tagset. |
| 430 |
tagSet.AddFilter(s.Key, filters[id]) |
| 431 |
seriesN++ |
| 432 |
} |
| 433 |
// Release the lock while we sort all the tags |
| 434 |
m.mu.RUnlock() |
| 435 |
|
| 436 |
// Sort the series in each tag set. |
| 437 |
for _, t := range tagSets { |
| 438 |
// Abort if the query was killed |
| 439 |
select { |
| 440 |
case <-opt.InterruptCh: |
| 441 |
return nil, query.ErrQueryInterrupted |
| 442 |
default: |
| 443 |
} |
| 444 |
|
| 445 |
sort.Sort(t) |
| 446 |
} |
| 447 |
|
| 448 |
// The TagSets have been created, as a map of TagSets. Just send |
| 449 |
// the values back as a slice, sorting for consistency. |
| 450 |
sortedTagsSets := make([]*query.TagSet, 0, len(tagSets)) |
| 451 |
for _, v := range tagSets { |
| 452 |
sortedTagsSets = append(sortedTagsSets, v) |
| 453 |
} |
| 454 |
sort.Sort(byTagKey(sortedTagsSets)) |
| 455 |
|
| 456 |
return sortedTagsSets, nil |
| 457 |
} |
| 458 |
|
| 459 |
// intersectSeriesFilters performs an intersection for two sets of ids and filter expressions. |
| 460 |
func intersectSeriesFilters(lids, rids seriesIDs, lfilters, rfilters FilterExprs) (seriesIDs, FilterExprs) { |
| 461 |
// We only want to allocate a slice and map of the smaller size. |
| 462 |
var ids []uint64 |
| 463 |
if len(lids) > len(rids) { |
| 464 |
ids = make([]uint64, 0, len(rids)) |
| 465 |
} else { |
| 466 |
ids = make([]uint64, 0, len(lids)) |
| 467 |
} |
| 468 |
|
| 469 |
var filters FilterExprs |
| 470 |
if len(lfilters) > len(rfilters) { |
| 471 |
filters = make(FilterExprs, len(rfilters)) |
| 472 |
} else { |
| 473 |
filters = make(FilterExprs, len(lfilters)) |
| 474 |
} |
| 475 |
|
| 476 |
// They're in sorted order so advance the counter as needed. |
| 477 |
// This is, don't run comparisons against lower values that we've already passed. |
| 478 |
for len(lids) > 0 && len(rids) > 0 { |
| 479 |
lid, rid := lids[0], rids[0] |
| 480 |
if lid == rid { |
| 481 |
ids = append(ids, lid) |
| 482 |
|
| 483 |
var expr influxql.Expr |
| 484 |
lfilter := lfilters[lid] |
| 485 |
rfilter := rfilters[rid] |
| 486 |
|
| 487 |
if lfilter != nil && rfilter != nil { |
| 488 |
be := &influxql.BinaryExpr{ |
| 489 |
Op: influxql.AND, |
| 490 |
LHS: lfilter, |
| 491 |
RHS: rfilter, |
| 492 |
} |
| 493 |
expr = influxql.Reduce(be, nil) |
| 494 |
} else if lfilter != nil { |
| 495 |
expr = lfilter |
| 496 |
} else if rfilter != nil { |
| 497 |
expr = rfilter |
| 498 |
} |
| 499 |
|
| 500 |
if expr != nil { |
| 501 |
filters[lid] = expr |
| 502 |
} |
| 503 |
lids, rids = lids[1:], rids[1:] |
| 504 |
} else if lid < rid { |
| 505 |
lids = lids[1:] |
| 506 |
} else { |
| 507 |
rids = rids[1:] |
| 508 |
} |
| 509 |
} |
| 510 |
return ids, filters |
| 511 |
} |
| 512 |
|
| 513 |
// unionSeriesFilters performs a union for two sets of ids and filter expressions. |
| 514 |
func unionSeriesFilters(lids, rids seriesIDs, lfilters, rfilters FilterExprs) (seriesIDs, FilterExprs) { |
| 515 |
ids := make([]uint64, 0, len(lids)+len(rids)) |
| 516 |
|
| 517 |
// Setup the filters with the smallest size since we will discard filters |
| 518 |
// that do not have a match on the other side. |
| 519 |
var filters FilterExprs |
| 520 |
if len(lfilters) < len(rfilters) { |
| 521 |
filters = make(FilterExprs, len(lfilters)) |
| 522 |
} else { |
| 523 |
filters = make(FilterExprs, len(rfilters)) |
| 524 |
} |
| 525 |
|
| 526 |
for len(lids) > 0 && len(rids) > 0 { |
| 527 |
lid, rid := lids[0], rids[0] |
| 528 |
if lid == rid { |
| 529 |
ids = append(ids, lid) |
| 530 |
|
| 531 |
// If one side does not have a filter, then the series has been |
| 532 |
// included on one side of the OR with no condition. Eliminate the |
| 533 |
// filter in this case. |
| 534 |
var expr influxql.Expr |
| 535 |
lfilter := lfilters[lid] |
| 536 |
rfilter := rfilters[rid] |
| 537 |
if lfilter != nil && rfilter != nil { |
| 538 |
be := &influxql.BinaryExpr{ |
| 539 |
Op: influxql.OR, |
| 540 |
LHS: lfilter, |
| 541 |
RHS: rfilter, |
| 542 |
} |
| 543 |
expr = influxql.Reduce(be, nil) |
| 544 |
} |
| 545 |
|
| 546 |
if expr != nil { |
| 547 |
filters[lid] = expr |
| 548 |
} |
| 549 |
lids, rids = lids[1:], rids[1:] |
| 550 |
} else if lid < rid { |
| 551 |
ids = append(ids, lid) |
| 552 |
|
| 553 |
filter := lfilters[lid] |
| 554 |
if filter != nil { |
| 555 |
filters[lid] = filter |
| 556 |
} |
| 557 |
lids = lids[1:] |
| 558 |
} else { |
| 559 |
ids = append(ids, rid) |
| 560 |
|
| 561 |
filter := rfilters[rid] |
| 562 |
if filter != nil { |
| 563 |
filters[rid] = filter |
| 564 |
} |
| 565 |
rids = rids[1:] |
| 566 |
} |
| 567 |
} |
| 568 |
|
| 569 |
// Now append the remainder. |
| 570 |
if len(lids) > 0 { |
| 571 |
for i := 0; i < len(lids); i++ { |
| 572 |
ids = append(ids, lids[i]) |
| 573 |
|
| 574 |
filter := lfilters[lids[i]] |
| 575 |
if filter != nil { |
| 576 |
filters[lids[i]] = filter |
| 577 |
} |
| 578 |
} |
| 579 |
} else if len(rids) > 0 { |
| 580 |
for i := 0; i < len(rids); i++ { |
| 581 |
ids = append(ids, rids[i]) |
| 582 |
|
| 583 |
filter := rfilters[rids[i]] |
| 584 |
if filter != nil { |
| 585 |
filters[rids[i]] = filter |
| 586 |
} |
| 587 |
} |
| 588 |
} |
| 589 |
return ids, filters |
| 590 |
} |
| 591 |
|
| 592 |
// SeriesIDsByTagKey returns a list of all series for a tag key. |
| 593 |
func (m *measurement) SeriesIDsByTagKey(key []byte) seriesIDs { |
| 594 |
tagVals := m.seriesByTagKeyValue[string(key)] |
| 595 |
if tagVals == nil { |
| 596 |
return nil |
| 597 |
} |
| 598 |
|
| 599 |
var ids seriesIDs |
| 600 |
tagVals.RangeAll(func(_ string, a seriesIDs) { |
| 601 |
ids = append(ids, a...) |
| 602 |
}) |
| 603 |
sort.Sort(ids) |
| 604 |
return ids |
| 605 |
} |
| 606 |
|
| 607 |
// SeriesIDsByTagValue returns a list of all series for a tag value. |
| 608 |
func (m *measurement) SeriesIDsByTagValue(key, value []byte) seriesIDs { |
| 609 |
tagVals := m.seriesByTagKeyValue[string(key)] |
| 610 |
if tagVals == nil { |
| 611 |
return nil |
| 612 |
} |
| 613 |
return tagVals.Load(string(value)) |
| 614 |
} |
| 615 |
|
| 616 |
// IDsForExpr returns the series IDs that are candidates to match the given expression. |
| 617 |
func (m *measurement) IDsForExpr(n *influxql.BinaryExpr) seriesIDs { |
| 618 |
ids, _, _ := m.idsForExpr(n) |
| 619 |
return ids |
| 620 |
} |
| 621 |
|
| 622 |
// idsForExpr returns a collection of series ids and a filter expression that should |
| 623 |
// be used to filter points from those series. |
| 624 |
func (m *measurement) idsForExpr(n *influxql.BinaryExpr) (seriesIDs, influxql.Expr, error) { |
| 625 |
// If this binary expression has another binary expression, then this |
| 626 |
// is some expression math and we should just pass it to the underlying query. |
| 627 |
if _, ok := n.LHS.(*influxql.BinaryExpr); ok { |
| 628 |
return m.SeriesIDs(), n, nil |
| 629 |
} else if _, ok := n.RHS.(*influxql.BinaryExpr); ok { |
| 630 |
return m.SeriesIDs(), n, nil |
| 631 |
} |
| 632 |
|
| 633 |
// Retrieve the variable reference from the correct side of the expression. |
| 634 |
name, ok := n.LHS.(*influxql.VarRef) |
| 635 |
value := n.RHS |
| 636 |
if !ok { |
| 637 |
name, ok = n.RHS.(*influxql.VarRef) |
| 638 |
if !ok { |
| 639 |
// This is an expression we do not know how to evaluate. Let the |
| 640 |
// query engine take care of this. |
| 641 |
return m.SeriesIDs(), n, nil |
| 642 |
} |
| 643 |
value = n.LHS |
| 644 |
} |
| 645 |
|
| 646 |
// For fields, return all series IDs from this measurement and return |
| 647 |
// the expression passed in, as the filter. |
| 648 |
if name.Val != "_name" && ((name.Type == influxql.Unknown && m.HasField(name.Val)) || name.Type == influxql.AnyField || (name.Type != influxql.Tag && name.Type != influxql.Unknown)) { |
| 649 |
return m.SeriesIDs(), n, nil |
| 650 |
} else if value, ok := value.(*influxql.VarRef); ok { |
| 651 |
// Check if the RHS is a variable and if it is a field. |
| 652 |
if value.Val != "_name" && ((value.Type == influxql.Unknown && m.HasField(value.Val)) || name.Type == influxql.AnyField || (value.Type != influxql.Tag && value.Type != influxql.Unknown)) { |
| 653 |
return m.SeriesIDs(), n, nil |
| 654 |
} |
| 655 |
} |
| 656 |
|
| 657 |
// Retrieve list of series with this tag key. |
| 658 |
tagVals := m.seriesByTagKeyValue[name.Val] |
| 659 |
|
| 660 |
// if we're looking for series with a specific tag value |
| 661 |
if str, ok := value.(*influxql.StringLiteral); ok { |
| 662 |
var ids seriesIDs |
| 663 |
|
| 664 |
// Special handling for "_name" to match measurement name. |
| 665 |
if name.Val == "_name" { |
| 666 |
if (n.Op == influxql.EQ && str.Val == m.Name) || (n.Op == influxql.NEQ && str.Val != m.Name) { |
| 667 |
return m.SeriesIDs(), nil, nil |
| 668 |
} |
| 669 |
return nil, nil, nil |
| 670 |
} |
| 671 |
|
| 672 |
if n.Op == influxql.EQ { |
| 673 |
if str.Val != "" { |
| 674 |
// return series that have a tag of specific value. |
| 675 |
ids = tagVals.Load(str.Val) |
| 676 |
} else { |
| 677 |
// Make a copy of all series ids and mark the ones we need to evict. |
| 678 |
sIDs := newEvictSeriesIDs(m.SeriesIDs()) |
| 679 |
|
| 680 |
// Go through each slice and mark the values we find as zero so |
| 681 |
// they can be removed later. |
| 682 |
tagVals.RangeAll(func(_ string, a seriesIDs) { |
| 683 |
sIDs.mark(a) |
| 684 |
}) |
| 685 |
|
| 686 |
// Make a new slice with only the remaining ids. |
| 687 |
ids = sIDs.evict() |
| 688 |
} |
| 689 |
} else if n.Op == influxql.NEQ { |
| 690 |
if str.Val != "" { |
| 691 |
ids = m.SeriesIDs() |
| 692 |
if vals := tagVals.Load(str.Val); len(vals) > 0 { |
| 693 |
ids = ids.Reject(vals) |
| 694 |
} |
| 695 |
} else { |
| 696 |
tagVals.RangeAll(func(_ string, a seriesIDs) { |
| 697 |
ids = append(ids, a...) |
| 698 |
}) |
| 699 |
sort.Sort(ids) |
| 700 |
} |
| 701 |
} |
| 702 |
return ids, nil, nil |
| 703 |
} |
| 704 |
|
| 705 |
// if we're looking for series with a tag value that matches a regex |
| 706 |
if re, ok := value.(*influxql.RegexLiteral); ok { |
| 707 |
var ids seriesIDs |
| 708 |
|
| 709 |
// Special handling for "_name" to match measurement name. |
| 710 |
if name.Val == "_name" { |
| 711 |
match := re.Val.MatchString(m.Name) |
| 712 |
if (n.Op == influxql.EQREGEX && match) || (n.Op == influxql.NEQREGEX && !match) { |
| 713 |
return m.SeriesIDs(), &influxql.BooleanLiteral{Val: true}, nil |
| 714 |
} |
| 715 |
return nil, nil, nil |
| 716 |
} |
| 717 |
|
| 718 |
// Check if we match the empty string to see if we should include series |
| 719 |
// that are missing the tag. |
| 720 |
empty := re.Val.MatchString("") |
| 721 |
|
| 722 |
// Gather the series that match the regex. If we should include the empty string, |
| 723 |
// start with the list of all series and reject series that don't match our condition. |
| 724 |
// If we should not include the empty string, include series that match our condition. |
| 725 |
if empty && n.Op == influxql.EQREGEX { |
| 726 |
// See comments above for EQ with a StringLiteral. |
| 727 |
sIDs := newEvictSeriesIDs(m.SeriesIDs()) |
| 728 |
tagVals.RangeAll(func(k string, a seriesIDs) { |
| 729 |
if !re.Val.MatchString(k) { |
| 730 |
sIDs.mark(a) |
| 731 |
} |
| 732 |
}) |
| 733 |
ids = sIDs.evict() |
| 734 |
} else if empty && n.Op == influxql.NEQREGEX { |
| 735 |
ids = make(seriesIDs, 0, len(m.SeriesIDs())) |
| 736 |
tagVals.RangeAll(func(k string, a seriesIDs) { |
| 737 |
if !re.Val.MatchString(k) { |
| 738 |
ids = append(ids, a...) |
| 739 |
} |
| 740 |
}) |
| 741 |
sort.Sort(ids) |
| 742 |
} else if !empty && n.Op == influxql.EQREGEX { |
| 743 |
ids = make(seriesIDs, 0, len(m.SeriesIDs())) |
| 744 |
tagVals.RangeAll(func(k string, a seriesIDs) { |
| 745 |
if re.Val.MatchString(k) { |
| 746 |
ids = append(ids, a...) |
| 747 |
} |
| 748 |
}) |
| 749 |
sort.Sort(ids) |
| 750 |
} else if !empty && n.Op == influxql.NEQREGEX { |
| 751 |
// See comments above for EQ with a StringLiteral. |
| 752 |
sIDs := newEvictSeriesIDs(m.SeriesIDs()) |
| 753 |
tagVals.RangeAll(func(k string, a seriesIDs) { |
| 754 |
if re.Val.MatchString(k) { |
| 755 |
sIDs.mark(a) |
| 756 |
} |
| 757 |
}) |
| 758 |
ids = sIDs.evict() |
| 759 |
} |
| 760 |
return ids, nil, nil |
| 761 |
} |
| 762 |
|
| 763 |
// compare tag values |
| 764 |
if ref, ok := value.(*influxql.VarRef); ok { |
| 765 |
var ids seriesIDs |
| 766 |
|
| 767 |
if n.Op == influxql.NEQ { |
| 768 |
ids = m.SeriesIDs() |
| 769 |
} |
| 770 |
|
| 771 |
rhsTagVals := m.seriesByTagKeyValue[ref.Val] |
| 772 |
tagVals.RangeAll(func(k string, a seriesIDs) { |
| 773 |
tags := a.Intersect(rhsTagVals.Load(k)) |
| 774 |
if n.Op == influxql.EQ { |
| 775 |
ids = ids.Union(tags) |
| 776 |
} else if n.Op == influxql.NEQ { |
| 777 |
ids = ids.Reject(tags) |
| 778 |
} |
| 779 |
}) |
| 780 |
return ids, nil, nil |
| 781 |
} |
| 782 |
|
| 783 |
// We do not know how to evaluate this expression so pass it |
| 784 |
// on to the query engine. |
| 785 |
return m.SeriesIDs(), n, nil |
| 786 |
} |
| 787 |
|
| 788 |
// FilterExprs represents a map of series IDs to filter expressions. |
| 789 |
type FilterExprs map[uint64]influxql.Expr |
| 790 |
|
| 791 |
// DeleteBoolLiteralTrues deletes all elements whose filter expression is a boolean literal true. |
| 792 |
func (fe FilterExprs) DeleteBoolLiteralTrues() { |
| 793 |
for id, expr := range fe { |
| 794 |
if e, ok := expr.(*influxql.BooleanLiteral); ok && e.Val { |
| 795 |
delete(fe, id) |
| 796 |
} |
| 797 |
} |
| 798 |
} |
| 799 |
|
| 800 |
// Len returns the number of elements. |
| 801 |
func (fe FilterExprs) Len() int { |
| 802 |
if fe == nil { |
| 803 |
return 0 |
| 804 |
} |
| 805 |
return len(fe) |
| 806 |
} |
| 807 |
|
| 808 |
// WalkWhereForSeriesIds recursively walks the WHERE clause and returns an ordered set of series IDs and |
| 809 |
// a map from those series IDs to filter expressions that should be used to limit points returned in |
| 810 |
// the final query result. |
| 811 |
func (m *measurement) WalkWhereForSeriesIds(expr influxql.Expr) (seriesIDs, FilterExprs, error) { |
| 812 |
switch n := expr.(type) { |
| 813 |
case *influxql.BinaryExpr: |
| 814 |
switch n.Op { |
| 815 |
case influxql.EQ, influxql.NEQ, influxql.LT, influxql.LTE, influxql.GT, influxql.GTE, influxql.EQREGEX, influxql.NEQREGEX: |
| 816 |
// Get the series IDs and filter expression for the tag or field comparison. |
| 817 |
ids, expr, err := m.idsForExpr(n) |
| 818 |
if err != nil { |
| 819 |
return nil, nil, err |
| 820 |
} |
| 821 |
|
| 822 |
if len(ids) == 0 { |
| 823 |
return ids, nil, nil |
| 824 |
} |
| 825 |
|
| 826 |
// If the expression is a boolean literal that is true, ignore it. |
| 827 |
if b, ok := expr.(*influxql.BooleanLiteral); ok && b.Val { |
| 828 |
expr = nil |
| 829 |
} |
| 830 |
|
| 831 |
var filters FilterExprs |
| 832 |
if expr != nil { |
| 833 |
filters = make(FilterExprs, len(ids)) |
| 834 |
for _, id := range ids { |
| 835 |
filters[id] = expr |
| 836 |
} |
| 837 |
} |
| 838 |
|
| 839 |
return ids, filters, nil |
| 840 |
case influxql.AND, influxql.OR: |
| 841 |
// Get the series IDs and filter expressions for the LHS. |
| 842 |
lids, lfilters, err := m.WalkWhereForSeriesIds(n.LHS) |
| 843 |
if err != nil { |
| 844 |
return nil, nil, err |
| 845 |
} |
| 846 |
|
| 847 |
// Get the series IDs and filter expressions for the RHS. |
| 848 |
rids, rfilters, err := m.WalkWhereForSeriesIds(n.RHS) |
| 849 |
if err != nil { |
| 850 |
return nil, nil, err |
| 851 |
} |
| 852 |
|
| 853 |
// Combine the series IDs from the LHS and RHS. |
| 854 |
if n.Op == influxql.AND { |
| 855 |
ids, filters := intersectSeriesFilters(lids, rids, lfilters, rfilters) |
| 856 |
return ids, filters, nil |
| 857 |
} else { |
| 858 |
ids, filters := unionSeriesFilters(lids, rids, lfilters, rfilters) |
| 859 |
return ids, filters, nil |
| 860 |
} |
| 861 |
} |
| 862 |
|
| 863 |
ids, _, err := m.idsForExpr(n) |
| 864 |
return ids, nil, err |
| 865 |
case *influxql.ParenExpr: |
| 866 |
// walk down the tree |
| 867 |
return m.WalkWhereForSeriesIds(n.Expr) |
| 868 |
case *influxql.BooleanLiteral: |
| 869 |
if n.Val { |
| 870 |
return m.SeriesIDs(), nil, nil |
| 871 |
} |
| 872 |
return nil, nil, nil |
| 873 |
default: |
| 874 |
return nil, nil, nil |
| 875 |
} |
| 876 |
} |
| 877 |
|
| 878 |
// SeriesIDsAllOrByExpr walks an expressions for matching series IDs |
| 879 |
// or, if no expressions is given, returns all series IDs for the measurement. |
| 880 |
func (m *measurement) SeriesIDsAllOrByExpr(expr influxql.Expr) (seriesIDs, error) { |
| 881 |
// If no expression given or the measurement has no series, |
| 882 |
// we can take just return the ids or nil accordingly. |
| 883 |
if expr == nil { |
| 884 |
return m.SeriesIDs(), nil |
| 885 |
} |
| 886 |
|
| 887 |
m.mu.RLock() |
| 888 |
l := len(m.seriesByID) |
| 889 |
m.mu.RUnlock() |
| 890 |
if l == 0 { |
| 891 |
return nil, nil |
| 892 |
} |
| 893 |
|
| 894 |
// Get series IDs that match the WHERE clause. |
| 895 |
ids, _, err := m.WalkWhereForSeriesIds(expr) |
| 896 |
if err != nil { |
| 897 |
return nil, err |
| 898 |
} |
| 899 |
|
| 900 |
return ids, nil |
| 901 |
} |
| 902 |
|
| 903 |
// tagKeysByExpr extracts the tag keys wanted by the expression. |
| 904 |
func (m *measurement) TagKeysByExpr(expr influxql.Expr) (map[string]struct{}, error) { |
| 905 |
if expr == nil { |
| 906 |
set := make(map[string]struct{}) |
| 907 |
for _, key := range m.TagKeys() { |
| 908 |
set[key] = struct{}{} |
| 909 |
} |
| 910 |
return set, nil |
| 911 |
} |
| 912 |
|
| 913 |
switch e := expr.(type) { |
| 914 |
case *influxql.BinaryExpr: |
| 915 |
switch e.Op { |
| 916 |
case influxql.EQ, influxql.NEQ, influxql.EQREGEX, influxql.NEQREGEX: |
| 917 |
tag, ok := e.LHS.(*influxql.VarRef) |
| 918 |
if !ok { |
| 919 |
return nil, fmt.Errorf("left side of '%s' must be a tag key", e.Op.String()) |
| 920 |
} else if tag.Val != "_tagKey" { |
| 921 |
return nil, nil |
| 922 |
} |
| 923 |
|
| 924 |
if influxql.IsRegexOp(e.Op) { |
| 925 |
re, ok := e.RHS.(*influxql.RegexLiteral) |
| 926 |
if !ok { |
| 927 |
return nil, fmt.Errorf("right side of '%s' must be a regular expression", e.Op.String()) |
| 928 |
} |
| 929 |
return m.tagKeysByFilter(e.Op, "", re.Val), nil |
| 930 |
} |
| 931 |
|
| 932 |
s, ok := e.RHS.(*influxql.StringLiteral) |
| 933 |
if !ok { |
| 934 |
return nil, fmt.Errorf("right side of '%s' must be a tag value string", e.Op.String()) |
| 935 |
} |
| 936 |
return m.tagKeysByFilter(e.Op, s.Val, nil), nil |
| 937 |
|
| 938 |
case influxql.AND, influxql.OR: |
| 939 |
lhs, err := m.TagKeysByExpr(e.LHS) |
| 940 |
if err != nil { |
| 941 |
return nil, err |
| 942 |
} |
| 943 |
|
| 944 |
rhs, err := m.TagKeysByExpr(e.RHS) |
| 945 |
if err != nil { |
| 946 |
return nil, err |
| 947 |
} |
| 948 |
|
| 949 |
if lhs != nil && rhs != nil { |
| 950 |
if e.Op == influxql.OR { |
| 951 |
return stringSet(lhs).union(rhs), nil |
| 952 |
} |
| 953 |
return stringSet(lhs).intersect(rhs), nil |
| 954 |
} else if lhs != nil { |
| 955 |
return lhs, nil |
| 956 |
} else if rhs != nil { |
| 957 |
return rhs, nil |
| 958 |
} |
| 959 |
return nil, nil |
| 960 |
default: |
| 961 |
return nil, fmt.Errorf("invalid operator") |
| 962 |
} |
| 963 |
|
| 964 |
case *influxql.ParenExpr: |
| 965 |
return m.TagKeysByExpr(e.Expr) |
| 966 |
} |
| 967 |
|
| 968 |
return nil, fmt.Errorf("%#v", expr) |
| 969 |
} |
| 970 |
|
| 971 |
// tagKeysByFilter will filter the tag keys for the measurement. |
| 972 |
func (m *measurement) tagKeysByFilter(op influxql.Token, val string, regex *regexp.Regexp) stringSet { |
| 973 |
ss := newStringSet() |
| 974 |
for _, key := range m.TagKeys() { |
| 975 |
var matched bool |
| 976 |
switch op { |
| 977 |
case influxql.EQ: |
| 978 |
matched = key == val |
| 979 |
case influxql.NEQ: |
| 980 |
matched = key != val |
| 981 |
case influxql.EQREGEX: |
| 982 |
matched = regex.MatchString(key) |
| 983 |
case influxql.NEQREGEX: |
| 984 |
matched = !regex.MatchString(key) |
| 985 |
} |
| 986 |
|
| 987 |
if !matched { |
| 988 |
continue |
| 989 |
} |
| 990 |
ss.add(key) |
| 991 |
} |
| 992 |
return ss |
| 993 |
} |
| 994 |
|
| 995 |
// Measurements represents a list of *Measurement. |
| 996 |
type measurements []*measurement |
| 997 |
|
| 998 |
// Len implements sort.Interface. |
| 999 |
func (a measurements) Len() int { return len(a) } |
| 1000 |
|
| 1001 |
// Less implements sort.Interface. |
| 1002 |
func (a measurements) Less(i, j int) bool { return a[i].Name < a[j].Name } |
| 1003 |
|
| 1004 |
// Swap implements sort.Interface. |
| 1005 |
func (a measurements) Swap(i, j int) { a[i], a[j] = a[j], a[i] } |
| 1006 |
|
| 1007 |
// series belong to a Measurement and represent unique time series in a database. |
| 1008 |
type series struct { |
| 1009 |
mu sync.RWMutex |
| 1010 |
deleted bool |
| 1011 |
|
| 1012 |
// immutable |
| 1013 |
ID uint64 |
| 1014 |
Measurement *measurement |
| 1015 |
Key string |
| 1016 |
Tags models.Tags |
| 1017 |
} |
| 1018 |
|
| 1019 |
// newSeries returns an initialized series struct |
| 1020 |
func newSeries(id uint64, m *measurement, key string, tags models.Tags) *series { |
| 1021 |
return &series{ |
| 1022 |
ID: id, |
| 1023 |
Measurement: m, |
| 1024 |
Key: key, |
| 1025 |
Tags: tags, |
| 1026 |
} |
| 1027 |
} |
| 1028 |
|
| 1029 |
// bytes estimates the memory footprint of this series, in bytes. |
| 1030 |
func (s *series) bytes() int { |
| 1031 |
var b int |
| 1032 |
s.mu.RLock() |
| 1033 |
b += 24 // RWMutex uses 24 bytes |
| 1034 |
b += int(unsafe.Sizeof(s.deleted) + unsafe.Sizeof(s.ID)) |
| 1035 |
// Do not count s.Measurement to prevent double-counting in Index.Bytes. |
| 1036 |
b += int(unsafe.Sizeof(s.Key)) + len(s.Key) |
| 1037 |
for _, tag := range s.Tags { |
| 1038 |
b += int(unsafe.Sizeof(tag)) + len(tag.Key) + len(tag.Value) |
| 1039 |
} |
| 1040 |
b += int(unsafe.Sizeof(s.Tags)) |
| 1041 |
s.mu.RUnlock() |
| 1042 |
return b |
| 1043 |
} |
| 1044 |
|
| 1045 |
// Delete marks this series as deleted. A deleted series should not be returned for queries. |
| 1046 |
func (s *series) Delete() { |
| 1047 |
s.mu.Lock() |
| 1048 |
s.deleted = true |
| 1049 |
s.mu.Unlock() |
| 1050 |
} |
| 1051 |
|
| 1052 |
// Deleted indicates if this was previously deleted. |
| 1053 |
func (s *series) Deleted() bool { |
| 1054 |
s.mu.RLock() |
| 1055 |
v := s.deleted |
| 1056 |
s.mu.RUnlock() |
| 1057 |
return v |
| 1058 |
} |
| 1059 |
|
| 1060 |
// TagKeyValue provides goroutine-safe concurrent access to the set of series |
| 1061 |
// ids mapping to a set of tag values. |
| 1062 |
type tagKeyValue struct { |
| 1063 |
mu sync.RWMutex |
| 1064 |
entries map[string]*tagKeyValueEntry |
| 1065 |
} |
| 1066 |
|
| 1067 |
// bytes estimates the memory footprint of this tagKeyValue, in bytes. |
| 1068 |
func (t *tagKeyValue) bytes() int { |
| 1069 |
var b int |
| 1070 |
t.mu.RLock() |
| 1071 |
b += 24 // RWMutex is 24 bytes |
| 1072 |
b += int(unsafe.Sizeof(t.entries)) |
| 1073 |
for k, v := range t.entries { |
| 1074 |
b += int(unsafe.Sizeof(k)) + len(k) |
| 1075 |
b += len(v.m) * 8 // uint64 |
| 1076 |
b += len(v.a) * 8 // uint64 |
| 1077 |
b += int(unsafe.Sizeof(v) + unsafe.Sizeof(*v)) |
| 1078 |
} |
| 1079 |
t.mu.RUnlock() |
| 1080 |
return b |
| 1081 |
} |
| 1082 |
|
| 1083 |
// NewTagKeyValue initialises a new TagKeyValue. |
| 1084 |
func newTagKeyValue() *tagKeyValue { |
| 1085 |
return &tagKeyValue{entries: make(map[string]*tagKeyValueEntry)} |
| 1086 |
} |
| 1087 |
|
| 1088 |
// Cardinality returns the number of values in the TagKeyValue. |
| 1089 |
func (t *tagKeyValue) Cardinality() int { |
| 1090 |
if t == nil { |
| 1091 |
return 0 |
| 1092 |
} |
| 1093 |
|
| 1094 |
t.mu.RLock() |
| 1095 |
defer t.mu.RUnlock() |
| 1096 |
return len(t.entries) |
| 1097 |
} |
| 1098 |
|
| 1099 |
// Contains returns true if the TagKeyValue contains value. |
| 1100 |
func (t *tagKeyValue) Contains(value string) bool { |
| 1101 |
if t == nil { |
| 1102 |
return false |
| 1103 |
} |
| 1104 |
|
| 1105 |
t.mu.RLock() |
| 1106 |
defer t.mu.RUnlock() |
| 1107 |
_, ok := t.entries[value] |
| 1108 |
return ok |
| 1109 |
} |
| 1110 |
|
| 1111 |
// InsertSeriesIDByte adds a series id to the tag key value. |
| 1112 |
func (t *tagKeyValue) InsertSeriesIDByte(value []byte, id uint64) { |
| 1113 |
t.mu.Lock() |
| 1114 |
entry := t.entries[string(value)] |
| 1115 |
if entry == nil { |
| 1116 |
entry = newTagKeyValueEntry() |
| 1117 |
t.entries[string(value)] = entry |
| 1118 |
} |
| 1119 |
entry.m[id] = struct{}{} |
| 1120 |
t.mu.Unlock() |
| 1121 |
} |
| 1122 |
|
| 1123 |
// Load returns the SeriesIDs for the provided tag value. |
| 1124 |
func (t *tagKeyValue) Load(value string) seriesIDs { |
| 1125 |
if t == nil { |
| 1126 |
return nil |
| 1127 |
} |
| 1128 |
|
| 1129 |
t.mu.RLock() |
| 1130 |
entry := t.entries[value] |
| 1131 |
ids, changed := entry.ids() |
| 1132 |
t.mu.RUnlock() |
| 1133 |
|
| 1134 |
if changed { |
| 1135 |
t.mu.Lock() |
| 1136 |
entry.setIDs(ids) |
| 1137 |
t.mu.Unlock() |
| 1138 |
} |
| 1139 |
return ids |
| 1140 |
} |
| 1141 |
|
| 1142 |
// TagKeyValue is a no-op. |
| 1143 |
// |
| 1144 |
// If f returns false then iteration over any remaining keys or values will cease. |
| 1145 |
func (t *tagKeyValue) Range(f func(tagValue string, a seriesIDs) bool) { |
| 1146 |
if t == nil { |
| 1147 |
return |
| 1148 |
} |
| 1149 |
|
| 1150 |
t.mu.RLock() |
| 1151 |
for tagValue, entry := range t.entries { |
| 1152 |
ids, changed := entry.ids() |
| 1153 |
if changed { |
| 1154 |
t.mu.RUnlock() |
| 1155 |
t.mu.Lock() |
| 1156 |
entry.setIDs(ids) |
| 1157 |
t.mu.Unlock() |
| 1158 |
t.mu.RLock() |
| 1159 |
} |
| 1160 |
|
| 1161 |
if !f(tagValue, ids) { |
| 1162 |
t.mu.RUnlock() |
| 1163 |
return |
| 1164 |
} |
| 1165 |
} |
| 1166 |
t.mu.RUnlock() |
| 1167 |
} |
| 1168 |
|
| 1169 |
// RangeAll calls f sequentially on each key and value. A call to RangeAll on a |
| 1170 |
// nil TagKeyValue is a no-op. |
| 1171 |
func (t *tagKeyValue) RangeAll(f func(k string, a seriesIDs)) { |
| 1172 |
t.Range(func(k string, a seriesIDs) bool { |
| 1173 |
f(k, a) |
| 1174 |
return true |
| 1175 |
}) |
| 1176 |
} |
| 1177 |
|
| 1178 |
type tagKeyValueEntry struct { |
| 1179 |
m map[uint64]struct{} // series id set |
| 1180 |
a seriesIDs // lazily sorted list of series. |
| 1181 |
} |
| 1182 |
|
| 1183 |
func newTagKeyValueEntry() *tagKeyValueEntry { |
| 1184 |
return &tagKeyValueEntry{m: make(map[uint64]struct{})} |
| 1185 |
} |
| 1186 |
|
| 1187 |
func (e *tagKeyValueEntry) ids() (_ seriesIDs, changed bool) { |
| 1188 |
if e == nil { |
| 1189 |
return nil, false |
| 1190 |
} else if len(e.a) == len(e.m) { |
| 1191 |
return e.a, false |
| 1192 |
} |
| 1193 |
|
| 1194 |
a := make(seriesIDs, 0, len(e.m)) |
| 1195 |
for id := range e.m { |
| 1196 |
a = append(a, id) |
| 1197 |
} |
| 1198 |
radix.SortUint64s(a) |
| 1199 |
|
| 1200 |
return a, true |
| 1201 |
} |
| 1202 |
|
| 1203 |
func (e *tagKeyValueEntry) setIDs(a seriesIDs) { |
| 1204 |
if e == nil { |
| 1205 |
return |
| 1206 |
} |
| 1207 |
e.a = a |
| 1208 |
} |
| 1209 |
|
| 1210 |
// SeriesIDs is a convenience type for sorting, checking equality, and doing |
| 1211 |
// union and intersection of collections of series ids. |
| 1212 |
type seriesIDs []uint64 |
| 1213 |
|
| 1214 |
// Len implements sort.Interface. |
| 1215 |
func (a seriesIDs) Len() int { return len(a) } |
| 1216 |
|
| 1217 |
// Less implements sort.Interface. |
| 1218 |
func (a seriesIDs) Less(i, j int) bool { return a[i] < a[j] } |
| 1219 |
|
| 1220 |
// Swap implements sort.Interface. |
| 1221 |
func (a seriesIDs) Swap(i, j int) { a[i], a[j] = a[j], a[i] } |
| 1222 |
|
| 1223 |
// Equals assumes that both are sorted. |
| 1224 |
func (a seriesIDs) Equals(other seriesIDs) bool { |
| 1225 |
if len(a) != len(other) { |
| 1226 |
return false |
| 1227 |
} |
| 1228 |
for i, s := range other { |
| 1229 |
if a[i] != s { |
| 1230 |
return false |
| 1231 |
} |
| 1232 |
} |
| 1233 |
return true |
| 1234 |
} |
| 1235 |
|
| 1236 |
// Intersect returns a new collection of series ids in sorted order that is the intersection of the two. |
| 1237 |
// The two collections must already be sorted. |
| 1238 |
func (a seriesIDs) Intersect(other seriesIDs) seriesIDs { |
| 1239 |
l := a |
| 1240 |
r := other |
| 1241 |
|
| 1242 |
// we want to iterate through the shortest one and stop |
| 1243 |
if len(other) < len(a) { |
| 1244 |
l = other |
| 1245 |
r = a |
| 1246 |
} |
| 1247 |
|
| 1248 |
// they're in sorted order so advance the counter as needed. |
| 1249 |
// That is, don't run comparisons against lower values that we've already passed |
| 1250 |
var i, j int |
| 1251 |
|
| 1252 |
ids := make([]uint64, 0, len(l)) |
| 1253 |
for i < len(l) && j < len(r) { |
| 1254 |
if l[i] == r[j] { |
| 1255 |
ids = append(ids, l[i]) |
| 1256 |
i++ |
| 1257 |
j++ |
| 1258 |
} else if l[i] < r[j] { |
| 1259 |
i++ |
| 1260 |
} else { |
| 1261 |
j++ |
| 1262 |
} |
| 1263 |
} |
| 1264 |
|
| 1265 |
return seriesIDs(ids) |
| 1266 |
} |
| 1267 |
|
| 1268 |
// Union returns a new collection of series ids in sorted order that is the union of the two. |
| 1269 |
// The two collections must already be sorted. |
| 1270 |
func (a seriesIDs) Union(other seriesIDs) seriesIDs { |
| 1271 |
l := a |
| 1272 |
r := other |
| 1273 |
ids := make([]uint64, 0, len(l)+len(r)) |
| 1274 |
var i, j int |
| 1275 |
for i < len(l) && j < len(r) { |
| 1276 |
if l[i] == r[j] { |
| 1277 |
ids = append(ids, l[i]) |
| 1278 |
i++ |
| 1279 |
j++ |
| 1280 |
} else if l[i] < r[j] { |
| 1281 |
ids = append(ids, l[i]) |
| 1282 |
i++ |
| 1283 |
} else { |
| 1284 |
ids = append(ids, r[j]) |
| 1285 |
j++ |
| 1286 |
} |
| 1287 |
} |
| 1288 |
|
| 1289 |
// now append the remainder |
| 1290 |
if i < len(l) { |
| 1291 |
ids = append(ids, l[i:]...) |
| 1292 |
} else if j < len(r) { |
| 1293 |
ids = append(ids, r[j:]...) |
| 1294 |
} |
| 1295 |
|
| 1296 |
return ids |
| 1297 |
} |
| 1298 |
|
| 1299 |
// Reject returns a new collection of series ids in sorted order with the passed in set removed from the original. |
| 1300 |
// This is useful for the NOT operator. The two collections must already be sorted. |
| 1301 |
func (a seriesIDs) Reject(other seriesIDs) seriesIDs { |
| 1302 |
l := a |
| 1303 |
r := other |
| 1304 |
var i, j int |
| 1305 |
|
| 1306 |
ids := make([]uint64, 0, len(l)) |
| 1307 |
for i < len(l) && j < len(r) { |
| 1308 |
if l[i] == r[j] { |
| 1309 |
i++ |
| 1310 |
j++ |
| 1311 |
} else if l[i] < r[j] { |
| 1312 |
ids = append(ids, l[i]) |
| 1313 |
i++ |
| 1314 |
} else { |
| 1315 |
j++ |
| 1316 |
} |
| 1317 |
} |
| 1318 |
|
| 1319 |
// Append the remainder |
| 1320 |
if i < len(l) { |
| 1321 |
ids = append(ids, l[i:]...) |
| 1322 |
} |
| 1323 |
|
| 1324 |
return seriesIDs(ids) |
| 1325 |
} |
| 1326 |
|
| 1327 |
// seriesID is a series id that may or may not have been evicted from the |
| 1328 |
// current id list. |
| 1329 |
type seriesID struct { |
| 1330 |
val uint64 |
| 1331 |
evict bool |
| 1332 |
} |
| 1333 |
|
| 1334 |
// evictSeriesIDs is a slice of SeriesIDs with an extra field to mark if the |
| 1335 |
// field should be evicted or not. |
| 1336 |
type evictSeriesIDs struct { |
| 1337 |
ids []seriesID |
| 1338 |
sz int |
| 1339 |
} |
| 1340 |
|
| 1341 |
// newEvictSeriesIDs copies the ids into a new slice that can be used for |
| 1342 |
// evicting series from the slice. |
| 1343 |
func newEvictSeriesIDs(ids []uint64) evictSeriesIDs { |
| 1344 |
a := make([]seriesID, len(ids)) |
| 1345 |
for i, id := range ids { |
| 1346 |
a[i].val = id |
| 1347 |
} |
| 1348 |
return evictSeriesIDs{ |
| 1349 |
ids: a, |
| 1350 |
sz: len(a), |
| 1351 |
} |
| 1352 |
} |
| 1353 |
|
| 1354 |
// mark marks all of the ids in the sorted slice to be evicted from the list of |
| 1355 |
// series ids. If an id to be evicted does not exist, it just gets ignored. |
| 1356 |
func (a *evictSeriesIDs) mark(ids []uint64) { |
| 1357 |
sIDs := a.ids |
| 1358 |
for _, id := range ids { |
| 1359 |
if len(sIDs) == 0 { |
| 1360 |
break |
| 1361 |
} |
| 1362 |
|
| 1363 |
// Perform a binary search of the remaining slice if |
| 1364 |
// the first element does not match the value we're |
| 1365 |
// looking for. |
| 1366 |
i := 0 |
| 1367 |
if sIDs[0].val < id { |
| 1368 |
i = sort.Search(len(sIDs), func(i int) bool { |
| 1369 |
return sIDs[i].val >= id |
| 1370 |
}) |
| 1371 |
} |
| 1372 |
|
| 1373 |
if i >= len(sIDs) { |
| 1374 |
break |
| 1375 |
} else if sIDs[i].val == id { |
| 1376 |
if !sIDs[i].evict { |
| 1377 |
sIDs[i].evict = true |
| 1378 |
a.sz-- |
| 1379 |
} |
| 1380 |
// Skip over this series since it has been evicted and won't be |
| 1381 |
// encountered again. |
| 1382 |
i++ |
| 1383 |
} |
| 1384 |
sIDs = sIDs[i:] |
| 1385 |
} |
| 1386 |
} |
| 1387 |
|
| 1388 |
// evict creates a new slice with only the series that have not been evicted. |
| 1389 |
func (a *evictSeriesIDs) evict() (ids seriesIDs) { |
| 1390 |
if a.sz == 0 { |
| 1391 |
return ids |
| 1392 |
} |
| 1393 |
|
| 1394 |
// Make a new slice with only the remaining ids. |
| 1395 |
ids = make([]uint64, 0, a.sz) |
| 1396 |
for _, id := range a.ids { |
| 1397 |
if id.evict { |
| 1398 |
continue |
| 1399 |
} |
| 1400 |
ids = append(ids, id.val) |
| 1401 |
} |
| 1402 |
return ids |
| 1403 |
} |
| 1404 |
|
| 1405 |
// TagFilter represents a tag filter when looking up other tags or measurements. |
| 1406 |
type TagFilter struct { |
| 1407 |
Op influxql.Token |
| 1408 |
Key string |
| 1409 |
Value string |
| 1410 |
Regex *regexp.Regexp |
| 1411 |
} |
| 1412 |
|
| 1413 |
// TagKeys returns a list of the measurement's tag names, in sorted order. |
| 1414 |
func (m *measurement) TagKeys() []string { |
| 1415 |
m.mu.RLock() |
| 1416 |
keys := make([]string, 0, len(m.seriesByTagKeyValue)) |
| 1417 |
for k := range m.seriesByTagKeyValue { |
| 1418 |
keys = append(keys, k) |
| 1419 |
} |
| 1420 |
m.mu.RUnlock() |
| 1421 |
sort.Strings(keys) |
| 1422 |
return keys |
| 1423 |
} |
| 1424 |
|
| 1425 |
// TagValues returns all the values for the given tag key, in an arbitrary order. |
| 1426 |
func (m *measurement) TagValues(auth query.Authorizer, key string) []string { |
| 1427 |
m.mu.RLock() |
| 1428 |
defer m.mu.RUnlock() |
| 1429 |
values := make([]string, 0, m.seriesByTagKeyValue[key].Cardinality()) |
| 1430 |
|
| 1431 |
m.seriesByTagKeyValue[key].RangeAll(func(k string, a seriesIDs) { |
| 1432 |
if query.AuthorizerIsOpen(auth) { |
| 1433 |
values = append(values, k) |
| 1434 |
} else { |
| 1435 |
for _, sid := range a { |
| 1436 |
s := m.seriesByID[sid] |
| 1437 |
if s == nil { |
| 1438 |
continue |
| 1439 |
} |
| 1440 |
if auth.AuthorizeSeriesRead(m.Database, m.NameBytes, s.Tags) { |
| 1441 |
values = append(values, k) |
| 1442 |
return |
| 1443 |
} |
| 1444 |
} |
| 1445 |
} |
| 1446 |
}) |
| 1447 |
return values |
| 1448 |
} |
| 1449 |
|
| 1450 |
// SetFieldName adds the field name to the measurement. |
| 1451 |
func (m *measurement) SetFieldName(name string) { |
| 1452 |
m.mu.RLock() |
| 1453 |
_, ok := m.fieldNames[name] |
| 1454 |
m.mu.RUnlock() |
| 1455 |
|
| 1456 |
if ok { |
| 1457 |
return |
| 1458 |
} |
| 1459 |
|
| 1460 |
m.mu.Lock() |
| 1461 |
m.fieldNames[name] = struct{}{} |
| 1462 |
m.mu.Unlock() |
| 1463 |
} |
| 1464 |
|
| 1465 |
// SeriesByTagKeyValue returns the TagKeyValue for the provided tag key. |
| 1466 |
func (m *measurement) SeriesByTagKeyValue(key string) *tagKeyValue { |
| 1467 |
m.mu.RLock() |
| 1468 |
defer m.mu.RUnlock() |
| 1469 |
return m.seriesByTagKeyValue[key] |
| 1470 |
} |
| 1471 |
|
| 1472 |
// stringSet represents a set of strings. |
| 1473 |
type stringSet map[string]struct{} |
| 1474 |
|
| 1475 |
// newStringSet returns an empty stringSet. |
| 1476 |
func newStringSet() stringSet { |
| 1477 |
return make(map[string]struct{}) |
| 1478 |
} |
| 1479 |
|
| 1480 |
// add adds strings to the set. |
| 1481 |
func (s stringSet) add(ss ...string) { |
| 1482 |
for _, n := range ss { |
| 1483 |
s[n] = struct{}{} |
| 1484 |
} |
| 1485 |
} |
| 1486 |
|
| 1487 |
// list returns the current elements in the set, in sorted order. |
| 1488 |
func (s stringSet) list() []string { |
| 1489 |
l := make([]string, 0, len(s)) |
| 1490 |
for k := range s { |
| 1491 |
l = append(l, k) |
| 1492 |
} |
| 1493 |
sort.Strings(l) |
| 1494 |
return l |
| 1495 |
} |
| 1496 |
|
| 1497 |
// union returns the union of this set and another. |
| 1498 |
func (s stringSet) union(o stringSet) stringSet { |
| 1499 |
ns := newStringSet() |
| 1500 |
for k := range s { |
| 1501 |
ns[k] = struct{}{} |
| 1502 |
} |
| 1503 |
for k := range o { |
| 1504 |
ns[k] = struct{}{} |
| 1505 |
} |
| 1506 |
return ns |
| 1507 |
} |
| 1508 |
|
| 1509 |
// intersect returns the intersection of this set and another. |
| 1510 |
func (s stringSet) intersect(o stringSet) stringSet { |
| 1511 |
shorter, longer := s, o |
| 1512 |
if len(longer) < len(shorter) { |
| 1513 |
shorter, longer = longer, shorter |
| 1514 |
} |
| 1515 |
|
| 1516 |
ns := newStringSet() |
| 1517 |
for k := range shorter { |
| 1518 |
if _, ok := longer[k]; ok { |
| 1519 |
ns[k] = struct{}{} |
| 1520 |
} |
| 1521 |
} |
| 1522 |
return ns |
| 1523 |
} |
| 1524 |
|
| 1525 |
type byTagKey []*query.TagSet |
| 1526 |
|
| 1527 |
func (t byTagKey) Len() int { return len(t) } |
| 1528 |
func (t byTagKey) Less(i, j int) bool { return bytes.Compare(t[i].Key, t[j].Key) < 0 } |
| 1529 |
func (t byTagKey) Swap(i, j int) { t[i], t[j] = t[j], t[i] } |