influxdb/storage/reads/group_resultset.go

425 lines
8.1 KiB
Go

package reads
import (
"bytes"
"context"
"fmt"
"sort"
"strings"
"github.com/influxdata/platform/logger"
"github.com/influxdata/platform/models"
"github.com/influxdata/platform/storage/reads/datatypes"
"github.com/influxdata/platform/tsdb/cursors"
"github.com/opentracing/opentracing-go"
"go.uber.org/zap"
)
type groupResultSet struct {
ctx context.Context
req *datatypes.ReadRequest
agg *datatypes.Aggregate
mb multiShardCursors
i int
rows []*SeriesRow
keys [][]byte
rgc groupByCursor
km keyMerger
newCursorFn func() (SeriesCursor, error)
nextGroupFn func(c *groupResultSet) GroupCursor
sortFn func(c *groupResultSet) (int, error)
eof bool
}
func NewGroupResultSet(ctx context.Context, req *datatypes.ReadRequest, newCursorFn func() (SeriesCursor, error)) GroupResultSet {
g := &groupResultSet{
ctx: ctx,
req: req,
agg: req.Aggregate,
keys: make([][]byte, len(req.GroupKeys)),
newCursorFn: newCursorFn,
}
g.mb = newMultiShardArrayCursors(ctx, req.TimestampRange.Start, req.TimestampRange.End, !req.Descending, req.PointsLimit)
for i, k := range req.GroupKeys {
g.keys[i] = []byte(k)
}
switch req.Group {
case datatypes.GroupBy:
g.sortFn = groupBySort
g.nextGroupFn = groupByNextGroup
g.rgc = groupByCursor{
ctx: ctx,
mb: g.mb,
agg: req.Aggregate,
vals: make([][]byte, len(req.GroupKeys)),
}
case datatypes.GroupNone:
g.sortFn = groupNoneSort
g.nextGroupFn = groupNoneNextGroup
default:
panic("not implemented")
}
n, err := g.sort()
if n == 0 || err != nil {
return nil
}
return g
}
var nilKey = [...]byte{0xff}
func (g *groupResultSet) Close() {}
func (g *groupResultSet) Next() GroupCursor {
if g.eof {
return nil
}
return g.nextGroupFn(g)
}
func (g *groupResultSet) sort() (int, error) {
log := logger.LoggerFromContext(g.ctx)
if log != nil {
var f func()
log, f = logger.NewOperation(log, "Sort", "group.sort", zap.String("group_type", g.req.Group.String()))
defer f()
}
span := opentracing.SpanFromContext(g.ctx)
if span != nil {
span = opentracing.StartSpan(
"group.sort",
opentracing.ChildOf(span.Context()),
opentracing.Tag{Key: "group_type", Value: g.req.Group.String()})
defer span.Finish()
}
n, err := g.sortFn(g)
if span != nil {
span.SetTag("rows", n)
}
if log != nil {
log.Info("Sort completed", zap.Int("rows", n))
}
return n, err
}
// seriesHasPoints reads the first block of TSM data to verify the series has points for
// the time range of the query.
func (g *groupResultSet) seriesHasPoints(row *SeriesRow) bool {
// TODO(sgc): this is expensive. Storage engine must provide efficient time range queries of series keys.
cur := g.mb.createCursor(*row)
var ts []int64
switch c := cur.(type) {
case cursors.IntegerArrayCursor:
a := c.Next()
ts = a.Timestamps
case cursors.FloatArrayCursor:
a := c.Next()
ts = a.Timestamps
case cursors.UnsignedArrayCursor:
a := c.Next()
ts = a.Timestamps
case cursors.BooleanArrayCursor:
a := c.Next()
ts = a.Timestamps
case cursors.StringArrayCursor:
a := c.Next()
ts = a.Timestamps
case nil:
return false
default:
panic(fmt.Sprintf("unreachable: %T", c))
}
cur.Close()
return len(ts) > 0
}
func groupNoneNextGroup(g *groupResultSet) GroupCursor {
cur, err := g.newCursorFn()
if err != nil {
// TODO(sgc): store error
return nil
} else if cur == nil {
return nil
}
g.eof = true
return &groupNoneCursor{
ctx: g.ctx,
mb: g.mb,
agg: g.agg,
cur: cur,
keys: g.km.get(),
}
}
func groupNoneSort(g *groupResultSet) (int, error) {
cur, err := g.newCursorFn()
if err != nil {
return 0, err
} else if cur == nil {
return 0, nil
}
allTime := g.req.Hints.HintSchemaAllTime()
g.km.clear()
n := 0
row := cur.Next()
for row != nil {
n++
if allTime || g.seriesHasPoints(row) {
g.km.mergeTagKeys(row.Tags)
}
row = cur.Next()
}
cur.Close()
return n, nil
}
func groupByNextGroup(g *groupResultSet) GroupCursor {
next:
row := g.rows[g.i]
for i := range g.keys {
g.rgc.vals[i] = row.Tags.Get(g.keys[i])
}
g.km.clear()
allTime := g.req.Hints.HintSchemaAllTime()
c := 0
rowKey := row.SortKey
j := g.i
for j < len(g.rows) && bytes.Equal(rowKey, g.rows[j].SortKey) {
if allTime || g.seriesHasPoints(g.rows[j]) {
g.km.mergeTagKeys(g.rows[j].Tags)
c++
}
j++
}
g.rgc.reset(g.rows[g.i:j])
g.rgc.keys = g.km.get()
g.i = j
if j == len(g.rows) {
g.eof = true
} else if c == 0 {
// no rows with points
goto next
}
return &g.rgc
}
func groupBySort(g *groupResultSet) (int, error) {
cur, err := g.newCursorFn()
if err != nil {
return 0, err
} else if cur == nil {
return 0, nil
}
var rows []*SeriesRow
vals := make([][]byte, len(g.keys))
tagsBuf := &tagsBuffer{sz: 4096}
row := cur.Next()
for row != nil {
nr := *row
nr.SeriesTags = tagsBuf.copyTags(nr.SeriesTags)
nr.Tags = tagsBuf.copyTags(nr.Tags)
l := 0
for i, k := range g.keys {
vals[i] = nr.Tags.Get(k)
if len(vals[i]) == 0 {
vals[i] = nilKey[:] // if there was no value, ensure it sorts last
}
l += len(vals[i])
}
nr.SortKey = make([]byte, 0, l)
for _, v := range vals {
nr.SortKey = append(nr.SortKey, v...)
}
rows = append(rows, &nr)
row = cur.Next()
}
sort.Slice(rows, func(i, j int) bool {
return bytes.Compare(rows[i].SortKey, rows[j].SortKey) == -1
})
g.rows = rows
cur.Close()
return len(rows), nil
}
type groupNoneCursor struct {
ctx context.Context
mb multiShardCursors
agg *datatypes.Aggregate
cur SeriesCursor
row SeriesRow
keys [][]byte
}
func (c *groupNoneCursor) Tags() models.Tags { return c.row.Tags }
func (c *groupNoneCursor) Keys() [][]byte { return c.keys }
func (c *groupNoneCursor) PartitionKeyVals() [][]byte { return nil }
func (c *groupNoneCursor) Close() { c.cur.Close() }
func (c *groupNoneCursor) Next() bool {
row := c.cur.Next()
if row == nil {
return false
}
c.row = *row
return true
}
func (c *groupNoneCursor) Cursor() cursors.Cursor {
cur := c.mb.createCursor(c.row)
if c.agg != nil {
cur = c.mb.newAggregateCursor(c.ctx, c.agg, cur)
}
return cur
}
type groupByCursor struct {
ctx context.Context
mb multiShardCursors
agg *datatypes.Aggregate
i int
rows []*SeriesRow
keys [][]byte
vals [][]byte
}
func (c *groupByCursor) reset(rows []*SeriesRow) {
c.i = 0
c.rows = rows
}
func (c *groupByCursor) Keys() [][]byte { return c.keys }
func (c *groupByCursor) PartitionKeyVals() [][]byte { return c.vals }
func (c *groupByCursor) Tags() models.Tags { return c.rows[c.i-1].Tags }
func (c *groupByCursor) Close() {}
func (c *groupByCursor) Next() bool {
if c.i < len(c.rows) {
c.i++
return true
}
return false
}
func (c *groupByCursor) Cursor() cursors.Cursor {
cur := c.mb.createCursor(*c.rows[c.i-1])
if c.agg != nil {
cur = c.mb.newAggregateCursor(c.ctx, c.agg, cur)
}
return cur
}
// keyMerger is responsible for determining a merged set of tag keys
type keyMerger struct {
i int
keys [2][][]byte
}
func (km *keyMerger) clear() {
km.i = 0
km.keys[0] = km.keys[0][:0]
}
func (km *keyMerger) get() [][]byte { return km.keys[km.i&1] }
func (km *keyMerger) String() string {
var s []string
for _, k := range km.get() {
s = append(s, string(k))
}
return strings.Join(s, ",")
}
func (km *keyMerger) mergeTagKeys(tags models.Tags) {
keys := km.keys[km.i&1]
i, j := 0, 0
for i < len(keys) && j < len(tags) && bytes.Equal(keys[i], tags[j].Key) {
i++
j++
}
if j == len(tags) {
// no new tags
return
}
km.i = (km.i + 1) & 1
l := len(keys) + len(tags)
if cap(km.keys[km.i]) < l {
km.keys[km.i] = make([][]byte, l)
} else {
km.keys[km.i] = km.keys[km.i][:l]
}
keya := km.keys[km.i]
// back up the pointers
if i > 0 {
i--
j--
}
k := i
copy(keya[:k], keys[:k])
for i < len(keys) && j < len(tags) {
cmp := bytes.Compare(keys[i], tags[j].Key)
if cmp < 0 {
keya[k] = keys[i]
i++
} else if cmp > 0 {
keya[k] = tags[j].Key
j++
} else {
keya[k] = keys[i]
i++
j++
}
k++
}
if i < len(keys) {
k += copy(keya[k:], keys[i:])
}
for j < len(tags) {
keya[k] = tags[j].Key
j++
k++
}
km.keys[km.i] = keya[:k]
}