// Ensure an identifier's segments can be quoted.
func TestQuoteIdent(t *testing.T) {
for i, tt := range []struct {
ident []string
s string
}{
{[]string{``}, `""`},
{[]string{`foo`, `bar`}, `"foo"."bar"`},
{[]string{`foo bar`, `baz`}, `"foo bar"."baz"`},
{[]string{`foo.bar`, `baz`}, `"foo.bar"."baz"`},
} {
if s := influxql.QuoteIdent(tt.ident); tt.s != s {
t.Errorf("%d. %s: mismatch: %s != %s", i, tt.ident, tt.s, s)
}
}
}
// normalizeStatement adds a default database and policy to the measurements in statement.
func (q *QueryExecutor) normalizeStatement(stmt influxql.Statement, defaultDatabase string) (err error) {
// Track prefixes for replacing field names.
prefixes := make(map[string]string)
// Qualify all measurements.
influxql.WalkFunc(stmt, func(n influxql.Node) {
if err != nil {
return
}
switch n := n.(type) {
case *influxql.Measurement:
e := q.normalizeMeasurement(n, defaultDatabase)
if e != nil {
err = e
return
}
prefixes[n.Name] = n.Name
}
})
if err != nil {
return err
}
// Replace all variable references that used measurement prefixes.
influxql.WalkFunc(stmt, func(n influxql.Node) {
switch n := n.(type) {
case *influxql.VarRef:
for k, v := range prefixes {
if strings.HasPrefix(n.Val, k+".") {
n.Val = v + "." + influxql.QuoteIdent(n.Val[len(k)+1:])
}
}
}
})
return
}
// CreateMappers will create a set of mappers that need to be run to execute the map phase of a MapReduceJob.
func (tx *tx) CreateMapReduceJobs(stmt *influxql.SelectStatement, tagKeys []string) ([]*influxql.MapReduceJob, error) {
jobs := []*influxql.MapReduceJob{}
for _, src := range stmt.Sources {
mm, ok := src.(*influxql.Measurement)
if !ok {
return nil, fmt.Errorf("invalid source type: %#v", src)
}
// get the index and the retention policy
rp, err := tx.meta.RetentionPolicy(mm.Database, mm.RetentionPolicy)
if err != nil {
return nil, err
}
m := tx.store.Measurement(mm.Database, mm.Name)
if m == nil {
return nil, ErrMeasurementNotFound(influxql.QuoteIdent([]string{mm.Database, "", mm.Name}...))
}
tx.measurement = m
// Validate the fields and tags asked for exist and keep track of which are in the select vs the where
var selectFields []string
var whereFields []string
var selectTags []string
for _, n := range stmt.NamesInSelect() {
if m.HasField(n) {
selectFields = append(selectFields, n)
continue
}
if !m.HasTagKey(n) {
return nil, fmt.Errorf("unknown field or tag name in select clause: %s", n)
}
selectTags = append(selectTags, n)
tagKeys = append(tagKeys, n)
}
for _, n := range stmt.NamesInWhere() {
if n == "time" {
continue
}
if m.HasField(n) {
whereFields = append(whereFields, n)
continue
}
if !m.HasTagKey(n) {
return nil, fmt.Errorf("unknown field or tag name in where clause: %s", n)
}
}
if len(selectFields) == 0 && len(stmt.FunctionCalls()) == 0 {
return nil, fmt.Errorf("select statement must include at least one field or function call")
}
// Validate that group by is not a field
for _, d := range stmt.Dimensions {
switch e := d.Expr.(type) {
case *influxql.VarRef:
if !m.HasTagKey(e.Val) {
return nil, fmt.Errorf("can not use field in group by clause: %s", e.Val)
}
}
}
// Grab time range from statement.
tmin, tmax := influxql.TimeRange(stmt.Condition)
if tmax.IsZero() {
tmax = tx.now
}
if tmin.IsZero() {
tmin = time.Unix(0, 0)
}
// Find shard groups within time range.
var shardGroups []*meta.ShardGroupInfo
for _, group := range rp.ShardGroups {
if group.Overlaps(tmin, tmax) {
g := group
shardGroups = append(shardGroups, &g)
}
}
if len(shardGroups) == 0 {
return nil, nil
}
// get the group by interval, if there is one
var interval int64
if d, err := stmt.GroupByInterval(); err != nil {
return nil, err
} else {
interval = d.Nanoseconds()
}
// get the sorted unique tag sets for this query.
tagSets, err := m.TagSets(stmt, tagKeys)
if err != nil {
return nil, err
}
for _, t := range tagSets {
// make a job for each tagset
job := &influxql.MapReduceJob{
MeasurementName: m.Name,
TagSet: t,
TMin: tmin.UnixNano(),
TMax: tmax.UnixNano(),
}
// make a mapper for each shard that must be hit. We may need to hit multiple shards within a shard group
var mappers []influxql.Mapper
// create mappers for each shard we need to hit
for _, sg := range shardGroups {
// TODO: implement distributed queries
if len(sg.Shards) != 1 {
return nil, fmt.Errorf("distributed queries aren't supported yet. You have a replication policy with RF < # of servers in cluster")
}
shard := tx.store.Shard(sg.Shards[0].ID)
if shard == nil {
// the store returned nil which means we haven't written any data into this shard yet, so ignore it
continue
}
// get the codec for this measuremnt. If this is nil it just means this measurement was
// never written into this shard, so we can skip it and continue.
codec := shard.FieldCodec(m.Name)
if codec == nil {
continue
}
var mapper influxql.Mapper
mapper = &LocalMapper{
seriesKeys: t.SeriesKeys,
shard: shard,
db: shard.DB(),
job: job,
decoder: codec,
filters: t.Filters,
whereFields: whereFields,
selectFields: selectFields,
selectTags: selectTags,
tmin: tmin.UnixNano(),
tmax: tmax.UnixNano(),
interval: interval,
// multiple mappers may need to be merged together to get the results
// for a raw query. So each mapper will have to read at least the
// limit plus the offset in data points to ensure we've hit our mark
limit: uint64(stmt.Limit) + uint64(stmt.Offset),
}
mappers = append(mappers, mapper)
}
job.Mappers = mappers
jobs = append(jobs, job)
}
}
// always return them in sorted order so the results from running the jobs are returned in a deterministic order
sort.Sort(influxql.MapReduceJobs(jobs))
return jobs, nil
}