Golang parser.QuerySpec类代码示例

func (self *Coordinator) getShardsAndProcessor(querySpec *parser.QuerySpec, writer engine.Processor) ([]*cluster.ShardData, engine.Processor, error) {
	shards, err := self.clusterConfiguration.GetShardsForQuery(querySpec)
	if err != nil {
		return nil, nil, err
	}
	shouldAggregateLocally := shards.ShouldAggregateLocally(querySpec)

	q := querySpec.SelectQuery()
	if q == nil {
		return shards, writer, nil
	}

	if !shouldAggregateLocally {
		// if we should aggregate in the coordinator (i.e. aggregation
		// isn't happening locally at the shard level), create an engine
		shardIds := make([]uint32, len(shards))
		for i, s := range shards {
			shardIds[i] = s.Id()
		}
		writer, err = engine.NewQueryEngine(writer, q, shardIds)
		if err != nil {
			log.Error(err)
			log.Debug("Coordinator processor chain: %s", engine.ProcessorChain(writer))
		}
		return shards, writer, err
	}

	// if we have a query with limit, then create an engine, or we can
	// make the passthrough limit aware
	writer = engine.NewPassthroughEngineWithLimit(writer, 100, q.Limit)
	return shards, writer, nil
}

func (self *ShardData) QueryResponseBufferSize(querySpec *parser.QuerySpec, batchPointSize int) int {
	groupByTime := querySpec.GetGroupByInterval()
	if groupByTime == nil {
		// If the group by time is nil, we shouldn't have to use a buffer since the shards should be queried sequentially.
		// However, set this to something high just to be safe.
		log.Debug("BUFFER SIZE: 1000")
		return 1000
	}

	tickCount := int(self.shardNanoseconds / uint64(*groupByTime))
	if tickCount < 10 {
		tickCount = 100
	} else if tickCount > 1000 {
		// cap this because each response should have up to this number of points in it.
		tickCount = tickCount / batchPointSize

		// but make sure it's at least 1k
		if tickCount < 1000 {
			tickCount = 1000
		}
	}
	columnCount := querySpec.GetGroupByColumnCount()
	if columnCount > 1 {
		// we don't really know the cardinality for any column up front. This is a just a multiplier so we'll see how this goes.
		// each response can have many points, so having a buffer of the ticks * 100 should be safe, but we'll see.
		tickCount = tickCount * 100
	}
	log.Debug("BUFFER SIZE: %d", tickCount)
	return tickCount
}

func (self *Coordinator) runListSeriesQuery(querySpec *parser.QuerySpec, p engine.Processor) error {
	allSeries := self.clusterConfiguration.MetaStore.GetSeriesForDatabase(querySpec.Database())
	matchingSeries := allSeries
	if q := querySpec.Query().GetListSeriesQuery(); q.HasRegex() {
		matchingSeries = nil
		regex := q.GetRegex()
		for _, s := range allSeries {
			if !regex.MatchString(s) {
				continue
			}
			matchingSeries = append(matchingSeries, s)
		}
	}
	name := "list_series_result"
	fields := []string{"name"}
	points := make([]*protocol.Point, len(matchingSeries), len(matchingSeries))

	for i, s := range matchingSeries {
		fieldValues := []*protocol.FieldValue{{StringValue: proto.String(s)}}
		points[i] = &protocol.Point{Values: fieldValues}
	}

	seriesResult := &protocol.Series{Name: &name, Fields: fields, Points: points}
	_, err := p.Yield(seriesResult)
	return err
}

func (self *ClusterConfiguration) getShardsToMatchQuery(querySpec *parser.QuerySpec) ([]*ShardData, error) {
	self.shardLock.RLock()
	defer self.shardLock.RUnlock()
	seriesNames, fromRegex := querySpec.TableNamesAndRegex()
	db := querySpec.Database()
	if fromRegex != nil {
		seriesNames = self.MetaStore.GetSeriesForDatabaseAndRegex(db, fromRegex)
	}
	uniqueShards := make(map[uint32]*ShardData)
	for _, name := range seriesNames {
		if fs := self.MetaStore.GetFieldsForSeries(db, name); len(fs) == 0 {
			return nil, fmt.Errorf("Couldn't find series: %s", name)
		}
		space := self.getShardSpaceToMatchSeriesName(db, name)
		if space == nil {
			continue
		}
		for _, shard := range space.shards {
			uniqueShards[shard.id] = shard
		}
	}
	shards := make([]*ShardData, 0, len(uniqueShards))
	for _, shard := range uniqueShards {
		shards = append(shards, shard)
	}
	SortShardsByTimeDescending(shards)
	return shards, nil
}

func (self *LevelDbShard) executeListSeriesQuery(querySpec *parser.QuerySpec, processor cluster.QueryProcessor) error {
	it := self.db.NewIterator(self.readOptions)
	defer it.Close()

	database := querySpec.Database()
	seekKey := append(DATABASE_SERIES_INDEX_PREFIX, []byte(querySpec.Database()+"~")...)
	it.Seek(seekKey)
	dbNameStart := len(DATABASE_SERIES_INDEX_PREFIX)
	for it = it; it.Valid(); it.Next() {
		key := it.Key()
		if len(key) < dbNameStart || !bytes.Equal(key[:dbNameStart], DATABASE_SERIES_INDEX_PREFIX) {
			break
		}
		dbSeries := string(key[dbNameStart:])
		parts := strings.Split(dbSeries, "~")
		if len(parts) > 1 {
			if parts[0] != database {
				break
			}
			name := parts[1]
			shouldContinue := processor.YieldPoint(&name, nil, nil)
			if !shouldContinue {
				return nil
			}
		}
	}
	return nil
}

func (self *Shard) Query(querySpec *parser.QuerySpec, processor cluster.QueryProcessor) error {
	if querySpec.IsListSeriesQuery() {
		return self.executeListSeriesQuery(querySpec, processor)
	} else if querySpec.IsDeleteFromSeriesQuery() {
		return self.executeDeleteQuery(querySpec, processor)
	}

	seriesAndColumns := querySpec.SelectQuery().GetReferencedColumns()

	if !self.hasReadAccess(querySpec) {
		return errors.New("User does not have access to one or more of the series requested.")
	}

	for series, columns := range seriesAndColumns {
		if regex, ok := series.GetCompiledRegex(); ok {
			seriesNames := self.metaStore.GetSeriesForDatabaseAndRegex(querySpec.Database(), regex)
			for _, name := range seriesNames {
				if !querySpec.HasReadAccess(name) {
					continue
				}
				err := self.executeQueryForSeries(querySpec, name, columns, processor)
				if err != nil {
					return err
				}
			}
		} else {
			err := self.executeQueryForSeries(querySpec, series.Name, columns, processor)
			if err != nil {
				return err
			}
		}
	}
	return nil
}

func (self *ClusterConfiguration) GetShardsForQuery(querySpec *parser.QuerySpec) Shards {
	shards := self.getShardsToMatchQuery(querySpec)
	shards = self.getShardRange(querySpec, shards)
	if querySpec.IsAscending() {
		SortShardsByTimeAscending(shards)
	}
	return shards
}

func (self *CoordinatorImpl) runDeleteQuery(querySpec *parser.QuerySpec, seriesWriter SeriesWriter) error {
	user := querySpec.User()
	db := querySpec.Database()
	if ok, err := self.permissions.AuthorizeDeleteQuery(user, db); !ok {
		return err
	}
	querySpec.RunAgainstAllServersInShard = true
	return self.runQuerySpec(querySpec, seriesWriter)
}

func (self *CoordinatorImpl) runQuerySpec(querySpec *parser.QuerySpec, seriesWriter SeriesWriter) error {
	shards, processor, seriesClosed, err := self.getShardsAndProcessor(querySpec, seriesWriter)
	if err != nil {
		return err
	}

	if len(shards) == 0 {
		return fmt.Errorf("Couldn't look up columns")
	}

	defer func() {
		if processor != nil {
			processor.Close()
			<-seriesClosed
		} else {
			seriesWriter.Close()
		}
	}()

	shardConcurrentLimit := self.config.ConcurrentShardQueryLimit
	if self.shouldQuerySequentially(shards, querySpec) {
		log.Debug("Querying shards sequentially")
		shardConcurrentLimit = 1
	}
	log.Debug("Shard concurrent limit: %d", shardConcurrentLimit)

	errors := make(chan error, shardConcurrentLimit)
	for i := 0; i < shardConcurrentLimit; i++ {
		errors <- nil
	}
	responseChannels := make(chan (<-chan *protocol.Response), shardConcurrentLimit)

	go self.readFromResponseChannels(processor, seriesWriter, querySpec.IsExplainQuery(), errors, responseChannels)

	err = self.queryShards(querySpec, shards, errors, responseChannels)

	// make sure we read the rest of the errors and responses
	for _err := range errors {
		if err == nil {
			err = _err
		}
	}

	for responsechan := range responseChannels {
		for response := range responsechan {
			if response.GetType() != endStreamResponse {
				continue
			}
			if response.ErrorMessage != nil && err == nil {
				err = common.NewQueryError(common.InvalidArgument, *response.ErrorMessage)
			}
			break
		}
	}
	return err
}

func (self *Permissions) CheckQueryPermissions(user common.User, db string, querySpec *parser.QuerySpec) (ok bool, err common.AuthorizationError) {
	switch querySpec.Query().Type() {
	case parser.Delete:
		return self.AuthorizeDeleteQuery(user, db)
	case parser.Select:
		return self.AuthorizeSelectQuery(user, db, querySpec)
	default:
		return true, ""
	}
}

func (self *Shard) hasReadAccess(querySpec *parser.QuerySpec) bool {
	for series := range querySpec.SeriesValuesAndColumns() {
		if _, isRegex := series.GetCompiledRegex(); !isRegex {
			if !querySpec.HasReadAccess(series.Name) {
				return false
			}
		}
	}
	return true
}

func (self *LevelDbShard) executeDropSeriesQuery(querySpec *parser.QuerySpec, processor cluster.QueryProcessor) error {
	database := querySpec.Database()
	series := querySpec.Query().DropSeriesQuery.GetTableName()
	err := self.dropSeries(database, series)
	if err != nil {
		return err
	}
	self.compact()
	return nil
}

func (self *Coordinator) expandRegex(spec *parser.QuerySpec) {
	q := spec.SelectQuery()
	if q == nil {
		return
	}

	f := func(r *regexp.Regexp) []string {
		return self.clusterConfiguration.MetaStore.GetSeriesForDatabaseAndRegex(spec.Database(), r)
	}

	parser.RewriteMergeQuery(q, f)
}

func (self *ClusterConfiguration) GetShardsForQuery(querySpec *parser.QuerySpec) (Shards, error) {
	shards, err := self.getShardsToMatchQuery(querySpec)
	if err != nil {
		return nil, err
	}
	log.Debug("Querying %d shards for query", len(shards))
	shards = self.getShardRange(querySpec, shards)
	if querySpec.IsAscending() {
		SortShardsByTimeAscending(shards)
	}
	return shards, nil
}

func (self *Shard) executeQueryForSeries(querySpec *parser.QuerySpec, name string, columns []string, processor engine.Processor) error {
	if querySpec.IsSinglePointQuery() {
		log.Debug("Running single query for series %s", name)
		return self.executeSinglePointQuery(querySpec, name, columns, processor)
	}
	var pi *PointIterator
	var err error
	columns, pi, err = self.getPointIteratorForSeries(querySpec, name, columns)
	if err != nil {
		return err
	}
	defer pi.Close()

	query := querySpec.SelectQuery()
	aliases := query.GetTableAliases(name)

	seriesOutgoing := &protocol.Series{Name: protocol.String(name), Fields: columns, Points: make([]*protocol.Point, 0, self.pointBatchSize)}
	for pi.Valid() {
		p := pi.Point()
		seriesOutgoing.Points = append(seriesOutgoing.Points, p)
		if len(seriesOutgoing.Points) >= self.pointBatchSize {
			ok, err := yieldToProcessor(seriesOutgoing, processor, aliases)
			if !ok || err != nil {
				log.Debug("Stopping processing.")
				if err != nil {
					log.Error("Error while processing data: %v", err)
					return err
				}
				return nil
			}
			seriesOutgoing = &protocol.Series{Name: protocol.String(name), Fields: columns, Points: make([]*protocol.Point, 0, self.pointBatchSize)}
		}

		pi.Next()
	}

	if err := pi.Error(); err != nil {
		return err
	}

	//Yield remaining data
	if ok, err := yieldToProcessor(seriesOutgoing, processor, aliases); !ok || err != nil {
		log.Debug("Stopping processing remaining points...")
		if err != nil {
			log.Error("Error while processing data: %v", err)
			return err
		}
	}

	log.Debug("Finished running query %s", query.GetQueryString())
	return nil
}