lib/logstorage: work-in-progress

app/vlselect/logsql/logsql.go

@@ -2,13 +2,17 @@ package logsql
 
 import (
 	"context"
+	"fmt"
+	"math"
 	"net/http"
+	"time"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/app/vlstorage"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httpserver"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/httputils"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logstorage"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
 )
 
 // ProcessQueryRequest handles /select/logsql/query request.
@@ -19,23 +23,46 @@ func ProcessQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Req
 		httpserver.Errorf(w, r, "%s", err)
 		return
 	}
-	limit, err := httputils.GetInt(r, "limit")
-	if err != nil {
-		httpserver.Errorf(w, r, "%s", err)
-		return
-	}
 
+	// Parse query
 	qStr := r.FormValue("query")
 	q, err := logstorage.ParseQuery(qStr)
 	if err != nil {
 		httpserver.Errorf(w, r, "cannot parse query [%s]: %s", qStr, err)
 		return
 	}
-	w.Header().Set("Content-Type", "application/stream+json; charset=utf-8")
 
+	// Parse optional start and end args
+	start, okStart, err := getTimeNsec(r, "start")
+	if err != nil {
+		httpserver.Errorf(w, r, "%s", err)
+		return
+	}
+	end, okEnd, err := getTimeNsec(r, "end")
+	if err != nil {
+		httpserver.Errorf(w, r, "%s", err)
+		return
+	}
+	if okStart || okEnd {
+		if !okStart {
+			start = math.MinInt64
+		}
+		if !okEnd {
+			end = math.MaxInt64
+		}
+		q.AddTimeFilter(start, end)
+	}
+
+	// Parse limit query arg
+	limit, err := httputils.GetInt(r, "limit")
+	if err != nil {
+		httpserver.Errorf(w, r, "%s", err)
+		return
+	}
 	if limit > 0 {
 		q.AddPipeLimit(uint64(limit))
 	}
+	q.Optimize()
 
 	tenantIDs := []logstorage.TenantID{tenantID}
 
@@ -54,6 +81,7 @@ func ProcessQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Req
 		blockResultPool.Put(bb)
 	}
 
+	w.Header().Set("Content-Type", "application/stream+json; charset=utf-8")
 	err = vlstorage.RunQuery(ctx, tenantIDs, q, writeBlock)
 
 	bw.FlushIgnoreErrors()
@@ -66,3 +94,16 @@ func ProcessQueryRequest(ctx context.Context, w http.ResponseWriter, r *http.Req
 }
 
 var blockResultPool bytesutil.ByteBufferPool
+
+func getTimeNsec(r *http.Request, argName string) (int64, bool, error) {
+	s := r.FormValue(argName)
+	if s == "" {
+		return 0, false, nil
+	}
+	currentTimestamp := float64(time.Now().UnixNano()) / 1e9
+	secs, err := promutils.ParseTimeAt(s, currentTimestamp)
+	if err != nil {
+		return 0, false, fmt.Errorf("cannot parse %s=%s: %w", argName, s, err)
+	}
+	return int64(secs * 1e9), true, nil
+}

docs/VictoriaLogs/CHANGELOG.md

@@ -19,6 +19,10 @@ according to [these docs](https://docs.victoriametrics.com/VictoriaLogs/QuickSta
 
 ## tip
 
+* FEATURE: add support for optional `start` and `end` query args to [HTTP querying API](https://docs.victoriametrics.com/victorialogs/querying/#http-api), which can be used for limiting the time range for [LogsQL query](https://docs.victoriametrics.com/victorialogs/logsql/).
+* FEATURE: add ability to return the first `N` results from [`sort` pipe](#https://docs.victoriametrics.com/victorialogs/logsql/#sort-pipe). This is useful when `N` biggest or `N` smallest values must be returned from large amounts of logs.
+* FEATURE: add [`quantile`](https://docs.victoriametrics.com/victorialogs/logsql/#quantile-stats) and [`median`](https://docs.victoriametrics.com/victorialogs/logsql/#median-stats) [stats functions](https://docs.victoriametrics.com/victorialogs/logsql/#stats-pipe).
+
 ## [v0.6.1](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v0.6.1-victorialogs)
 
 Released at 2024-05-14

docs/VictoriaLogs/LogsQL.md

@@ -1128,6 +1128,7 @@ By default rows are selected in arbitrary order because of performance reasons,
 
 See also:
 
+- [`sort` pipe](#sort-pipe)
 - [`offset` pipe](#offset-pipe)
 
 ### offset pipe
@@ -1147,6 +1148,7 @@ Rows can be sorted with [`sort` pipe](#sort-pipe).
 See also:
 
 - [`limit` pipe](#limit-pipe)
+- [`sort` pipe](#sort-pipe)
 
 ### rename pipe
 
@@ -1198,11 +1200,31 @@ The reverse order can be applied globally via `desc` keyword after `by(...)` cla
 _time:5m | sort by (foo, bar) desc
 ```
 
+Sorting of big number of logs can consume a lot of CPU time and memory. Sometimes it is enough to return the first `N` entries with the biggest
+or the smallest values. This can be done by adding `limit N` to the end of `sort ...` pipe.
+Such a query consumes lower amounts of memory when sorting big number of logs, since it keeps in memory only `N` log entries.
+For example, the following query returns top 10 log entries with the biggest values
+for the `request_duration` [field](https://docs.victoriametrics.com/VictoriaLogs/keyConcepts.html#data-model) during the last hour:
+
+```logsql
+_time:1h | sort by (request_duration desc) limit 10
+```
+
+If the first `N` sorted results must be skipped, then `offset N` can be added to `sort` pipe. For example,
+the following query skips the first 10 logs with the biggest `request_duration` [field](https://docs.victoriametrics.com/VictoriaLogs/keyConcepts.html#data-model),
+and then returns the next 20 sorted logs for the last 5 minutes:
+
+```logsql
+_time:1h | sort by (request_duration desc) offset 10 limit 20
+```
+
 Note that sorting of big number of logs can be slow and can consume a lot of additional memory.
 It is recommended limiting the number of logs before sorting with the following approaches:
 
+- Adding `limit N` to the end of `sort ...` pipe.
 - Reducing the selected time range with [time filter](#time-filter).
 - Using more specific [filters](#filters), so they select less logs.
+- Limiting the number of selected [fields](https://docs.victoriametrics.com/VictoriaLogs/keyConcepts.html#data-model) via [`fields` pipe](#fields-pipe).
 
 See also:
 
@@ -1371,8 +1393,11 @@ LogsQL supports the following functions for [`stats` pipe](#stats-pipe):
 - [`count_empty`](#count_empty-stats) calculates the number logs with empty [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 - [`count_uniq`](#count_uniq-stats) calculates the number of unique non-empty values for the given [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 - [`max`](#max-stats) calcualtes the maximum value over the given numeric [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
+- [`median`](#median-stats) calcualtes the [median](https://en.wikipedia.org/wiki/Median) value over the given numeric [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 - [`min`](#min-stats) calculates the minumum value over the given numeric [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
+- [`quantile`](#quantile-stats) calculates the given quantile for the given numeric [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 - [`sum`](#sum-stats) calculates the sum for the given numeric [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
+- [`sum_len`](#sum_len-stats) calculates the sum of lengths for the given [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 - [`uniq_values`](#uniq_values-stats) returns unique non-empty values for the given [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 - [`values`](#values-stats) returns all the values for the given [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
 
@@ -1391,6 +1416,8 @@ _time:5m | stats avg(duration) avg_duration
 
 See also:
 
+- [`median`](#median-stats)
+- [`quantile`](#quantile-stats)
 - [`min`](#min-stats)
 - [`max`](#max-stats)
 - [`sum`](#sum-stats)
@@ -1492,10 +1519,28 @@ _time:5m | stats max(duration) max_duration
 See also:
 
 - [`min`](#min-stats)
+- [`quantile`](#quantile-stats)
 - [`avg`](#avg-stats)
 - [`sum`](#sum-stats)
 - [`count`](#count-stats)
 
+### median stats
+
+`median(field1, ..., fieldN)` [stats pipe](#stats-pipe) calculates the [median](https://en.wikipedia.org/wiki/Median) value across
+the give numeric [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
+
+For example, the following query return median for the `duration` [field](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model)
+over logs for the last 5 minutes:
+
+```logsql
+_time:5m | stats median(duration) median_duration
+```
+
+See also:
+
+- [`quantile`](#quantile-stats)
+- [`avg`](#avg-stats)
+
 ### min stats
 
 `min(field1, ..., fieldN)` [stats pipe](#stats-pipe) calculates the minimum value across
@@ -1512,10 +1557,34 @@ _time:5m | stats min(duration) min_duration
 See also:
 
 - [`max`](#max-stats)
+- [`quantile`](#quantile-stats)
 - [`avg`](#avg-stats)
 - [`sum`](#sum-stats)
 - [`count`](#count-stats)
 
+### quantile stats
+
+`quantile(phi, field1, ..., fieldN)` [stats pipe](#stats-pipe) calculates `phi` [percentile](https://en.wikipedia.org/wiki/Percentile) over numeric values
+for the given [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model). The `phi` must be in the range `0 ... 1`, where `0` means `0th` percentile,
+while `1` means `100th` percentile.
+
+For example, the following query calculates `50th`, `90th` and `99th` percentiles for the `request_duration_seconds` [field](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model)
+over logs for the last 5 minutes:
+
+```logsql
+_time:5m | stats
+  quantile(0.5, request_duration_seconds) p50,
+  quantile(0.9, request_duration_seconds) p90,
+  quantile(0.99, request_duration_seconds) p99
+```
+
+See also:
+
+- [`min`](#min-stats)
+- [`max`](#max-stats)
+- [`median`](#median-stats)
+- [`avg`](#avg-stats)
+
 ### sum stats
 
 `sum(field1, ..., fieldN)` [stats pipe](#stats-pipe) calculates the sum of numeric values across
@@ -1535,6 +1604,22 @@ See also:
 - [`max`](#max-stats)
 - [`min`](#min-stats)
 
+### sum_len stats
+
+`sum_len(field1, ..., fieldN)` [stats pipe](#stats-pipe) calculates the sum of lengths of all the values
+for the given [log fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#data-model).
+
+For example, the following query returns the sum of lengths of [`_msg` fields](https://docs.victoriametrics.com/victorialogs/keyconcepts/#message-field)
+across all the logs for the last 5 minutes:
+
+```logsql
+_time:5m | stats sum_len(_msg) messages_len
+```
+
+See also:
+
+- [`count`](#count-stats)
+
 ### uniq_values stats
 
 `uniq_values(field1, ..., fieldN)` [stats pipe](#stats-pipe) returns the unique non-empty values across
@@ -1631,8 +1716,6 @@ Stats over the selected logs can be calculated via [`stats` pipe](#stats-pipe).
 LogsQL will support calculating the following additional stats based on the [log fields](https://docs.victoriametrics.com/VictoriaLogs/keyConcepts.html#data-model)
 and fields created by [transformations](#transformations):
 
-- The median and [percentile](https://en.wikipedia.org/wiki/Percentile) for the given field.
-
 It will be possible specifying an optional condition [filter](#post-filters) when calculating the stats.
 For example, `sum(response_size) if (is_admin:true)` calculates the total response size for admins only.
 

docs/VictoriaLogs/querying/README.md

@@ -50,6 +50,13 @@ By default the `/select/logsql/query` returns all the log entries matching the g
   ```sh
   curl http://localhost:9428/select/logsql/query -d 'query=error' -d 'limit=10'
   ```
+- By adding [`limit` pipe](https://docs.victoriametrics.com/victorialogs/logsql/#limit-pipe) to the query. For example:
+  ```sh
+  curl http://localhost:9428/select/logsql/query -d 'query=error | limit 10'
+  ```
+- By adding [`_time` filter](https://docs.victoriametrics.com/victorialogs/logsql/#time-filter). The time range for the query can be specified via optional
+  `start` and `end` query ars formatted according to [these docs](https://docs.victoriametrics.com/single-server-victoriametrics/#timestamp-formats).
+- By adding other [filters](https://docs.victoriametrics.com/victorialogs/logsql/#filters) to the query.
 
 The `/select/logsql/query` endpoint returns [a stream of JSON lines](https://jsonlines.org/),
 where each line contains JSON-encoded log entry in the form `{field1="value1",...,fieldN="valueN"}`.

lib/logstorage/block_result.go

@@ -1738,6 +1738,55 @@ func (c *blockResultColumn) getMinValue() float64 {
 	}
 }
 
+func (c *blockResultColumn) sumLenValues(br *blockResult) uint64 {
+	if c.isConst {
+		v := c.encodedValues[0]
+		return uint64(len(v)) * uint64(len(br.timestamps))
+	}
+	if c.isTime {
+		return uint64(len(time.RFC3339Nano)) * uint64(len(br.timestamps))
+	}
+
+	switch c.valueType {
+	case valueTypeString:
+		return c.sumLenStringValues(br)
+	case valueTypeDict:
+		n := uint64(0)
+		dictValues := c.dictValues
+		for _, v := range c.encodedValues {
+			idx := v[0]
+			v := dictValues[idx]
+			n += uint64(len(v))
+		}
+		return n
+	case valueTypeUint8:
+		return c.sumLenStringValues(br)
+	case valueTypeUint16:
+		return c.sumLenStringValues(br)
+	case valueTypeUint32:
+		return c.sumLenStringValues(br)
+	case valueTypeUint64:
+		return c.sumLenStringValues(br)
+	case valueTypeFloat64:
+		return c.sumLenStringValues(br)
+	case valueTypeIPv4:
+		return c.sumLenStringValues(br)
+	case valueTypeTimestampISO8601:
+		return uint64(len(iso8601Timestamp)) * uint64(len(br.timestamps))
+	default:
+		logger.Panicf("BUG: unknown valueType=%d", c.valueType)
+		return 0
+	}
+}
+
+func (c *blockResultColumn) sumLenStringValues(br *blockResult) uint64 {
+	n := uint64(0)
+	for _, v := range c.getValues(br) {
+		n += uint64(len(v))
+	}
+	return n
+}
+
 func (c *blockResultColumn) sumValues(br *blockResult) (float64, int) {
 	if c.isConst {
 		v := c.encodedValues[0]

lib/logstorage/parser.go

@@ -206,6 +206,29 @@ func (q *Query) String() string {
 	return s
 }
 
+// AddTimeFilter adds global filter _time:[start ... end] to q.
+func (q *Query) AddTimeFilter(start, end int64) {
+	startStr := marshalTimestampRFC3339Nano(nil, start)
+	endStr := marshalTimestampRFC3339Nano(nil, end)
+	ft := &filterTime{
+		minTimestamp: start,
+		maxTimestamp: end,
+		stringRepr:   fmt.Sprintf("[%s, %s]", startStr, endStr),
+	}
+
+	fa, ok := q.f.(*filterAnd)
+	if ok {
+		filters := make([]filter, len(fa.filters)+1)
+		filters[0] = ft
+		copy(filters[1:], fa.filters)
+		fa.filters = filters
+	} else {
+		q.f = &filterAnd{
+			filters: []filter{ft, q.f},
+		}
+	}
+}
+
 // AddPipeLimit adds `| limit n` pipe to q.
 //
 // See https://docs.victoriametrics.com/victorialogs/logsql/#limit-pipe
@@ -215,6 +238,56 @@ func (q *Query) AddPipeLimit(n uint64) {
 	})
 }
 
+// Optimize tries optimizing the query.
+func (q *Query) Optimize() {
+	q.pipes = optimizeSortOffsetPipes(q.pipes)
+	q.pipes = optimizeSortLimitPipes(q.pipes)
+}
+
+func optimizeSortOffsetPipes(pipes []pipe) []pipe {
+	// Merge 'sort ... | offset ...' into 'sort ... offset ...'
+	i := 1
+	for i < len(pipes) {
+		po, ok := pipes[i].(*pipeOffset)
+		if !ok {
+			i++
+			continue
+		}
+		ps, ok := pipes[i-1].(*pipeSort)
+		if !ok {
+			i++
+			continue
+		}
+		if ps.offset == 0 && ps.limit == 0 {
+			ps.offset = po.n
+		}
+		pipes = append(pipes[:i], pipes[i+1:]...)
+	}
+	return pipes
+}
+
+func optimizeSortLimitPipes(pipes []pipe) []pipe {
+	// Merge 'sort ... | limit ...' into 'sort ... limit ...'
+	i := 1
+	for i < len(pipes) {
+		pl, ok := pipes[i].(*pipeLimit)
+		if !ok {
+			i++
+			continue
+		}
+		ps, ok := pipes[i-1].(*pipeSort)
+		if !ok {
+			i++
+			continue
+		}
+		if ps.limit == 0 || pl.n < ps.limit {
+			ps.limit = pl.n
+		}
+		pipes = append(pipes[:i], pipes[i+1:]...)
+	}
+	return pipes
+}
+
 func (q *Query) getNeededColumns() ([]string, []string) {
 	neededFields := newFieldsSet()
 	neededFields.add("*")

lib/logstorage/parser_test.go

@@ -926,6 +926,27 @@ func TestParseQuerySuccess(t *testing.T) {
 	f(`* | stats by(x) values() limit 1_000 AS y`, `* | stats by (x) values(*) limit 1000 as y`)
 	f(`* | stats by(x) values(a,*,b) y`, `* | stats by (x) values(*) as y`)
 
+	// stats pipe sum_len
+	f(`* | stats Sum_len(foo) bar`, `* | stats sum_len(foo) as bar`)
+	f(`* | stats BY(x, y, ) SUM_Len(foo,bar,) bar`, `* | stats by (x, y) sum_len(foo, bar) as bar`)
+	f(`* | stats sum_len() x`, `* | stats sum_len(*) as x`)
+	f(`* | stats sum_len(*) x`, `* | stats sum_len(*) as x`)
+	f(`* | stats sum_len(foo,*,bar) x`, `* | stats sum_len(*) as x`)
+
+	// stats pipe quantile
+	f(`* | stats quantile(0, foo) bar`, `* | stats quantile(0, foo) as bar`)
+	f(`* | stats quantile(1, foo) bar`, `* | stats quantile(1, foo) as bar`)
+	f(`* | stats quantile(0.5, a, b, c) bar`, `* | stats quantile(0.5, a, b, c) as bar`)
+	f(`* | stats quantile(0.99, *) bar`, `* | stats quantile(0.99, *) as bar`)
+	f(`* | stats quantile(0.99, a, *, b) bar`, `* | stats quantile(0.99, *) as bar`)
+
+	// stats pipe median
+	f(`* | stats Median(foo) bar`, `* | stats median(foo) as bar`)
+	f(`* | stats BY(x, y, ) MEDIAN(foo,bar,) bar`, `* | stats by (x, y) median(foo, bar) as bar`)
+	f(`* | stats median() x`, `* | stats median(*) as x`)
+	f(`* | stats median(*) x`, `* | stats median(*) as x`)
+	f(`* | stats median(foo,*,bar) x`, `* | stats median(*) as x`)
+
 	// stats pipe multiple funcs
 	f(`* | stats count() "foo.bar:baz", count_uniq(a) bar`, `* | stats count(*) as "foo.bar:baz", count_uniq(a) as bar`)
 	f(`* | stats by (x, y) count(*) foo, count_uniq(a,b) bar`, `* | stats by (x, y) count(*) as foo, count_uniq(a, b) as bar`)
@@ -953,6 +974,15 @@ func TestParseQuerySuccess(t *testing.T) {
 	f(`* | sort bY (foo)`, `* | sort by (foo)`)
 	f(`* | sORt bY (_time, _stream DEsc, host)`, `* | sort by (_time, _stream desc, host)`)
 	f(`* | sort bY (foo desc, bar,) desc`, `* | sort by (foo desc, bar) desc`)
+	f(`* | sort limit 10`, `* | sort limit 10`)
+	f(`* | sort offset 20 limit 10`, `* | sort offset 20 limit 10`)
+	f(`* | sort desc limit 10`, `* | sort desc limit 10`)
+	f(`* | sort desc offset 20 limit 10`, `* | sort desc offset 20 limit 10`)
+	f(`* | sort by (foo desc, bar) limit 10`, `* | sort by (foo desc, bar) limit 10`)
+	f(`* | sort by (foo desc, bar) oFFset 20 limit 10`, `* | sort by (foo desc, bar) offset 20 limit 10`)
+	f(`* | sort by (foo desc, bar) desc limit 10`, `* | sort by (foo desc, bar) desc limit 10`)
+	f(`* | sort by (foo desc, bar) desc OFFSET 30 limit 10`, `* | sort by (foo desc, bar) desc offset 30 limit 10`)
+	f(`* | sort by (foo desc, bar) desc limit 10 OFFSET 30`, `* | sort by (foo desc, bar) desc offset 30 limit 10`)
 
 	// uniq pipe
 	f(`* | uniq`, `* | uniq`)
@@ -1275,6 +1305,18 @@ func TestParseQueryFailure(t *testing.T) {
 	f(`foo | stats values(a) limit 0.5`)
 	f(`foo | stats values(a) limit -1`)
 
+	// invalid stats sum_len
+	f(`foo | stats sum_len`)
+	f(`foo | stats sum_len()`)
+
+	// invalid stats quantile
+	f(`foo | stats quantile`)
+	f(`foo | stats quantile() foo`)
+	f(`foo | stats quantile(bar, baz) foo`)
+	f(`foo | stats quantile(0.5) foo`)
+	f(`foo | stats quantile(-1, x) foo`)
+	f(`foo | stats quantile(10, x) foo`)
+
 	// invalid stats grouping fields
 	f(`foo | stats by(foo:bar) count() baz`)
 	f(`foo | stats by(foo:/bar) count() baz`)
@@ -1297,6 +1339,16 @@ func TestParseQueryFailure(t *testing.T) {
 	f(`foo | sort by(baz`)
 	f(`foo | sort by(baz,`)
 	f(`foo | sort by(bar) foo`)
+	f(`foo | sort by(bar) limit`)
+	f(`foo | sort by(bar) limit foo`)
+	f(`foo | sort by(bar) limit -1234`)
+	f(`foo | sort by(bar) limit 12.34`)
+	f(`foo | sort by(bar) limit 10 limit 20`)
+	f(`foo | sort by(bar) offset`)
+	f(`foo | sort by(bar) offset limit`)
+	f(`foo | sort by(bar) offset -1234`)
+	f(`foo | sort by(bar) offset 12.34`)
+	f(`foo | sort by(bar) offset 10 offset 20`)
 
 	// invalid uniq pipe
 	f(`foo | uniq bar`)

lib/logstorage/pipe.go

@@ -65,7 +65,7 @@ func parsePipes(lex *lexer) ([]pipe, error) {
 	var pipes []pipe
 	for !lex.isKeyword(")", "") {
 		if !lex.isKeyword("|") {
-			return nil, fmt.Errorf("expecting '|'")
+			return nil, fmt.Errorf("expecting '|'; got %q", lex.token)
 		}
 		if !lex.mustNextToken() {
 			return nil, fmt.Errorf("missing token after '|'")

lib/logstorage/pipe_sort.go

@@ -25,6 +25,14 @@ type pipeSort struct {
 
 	// whether to apply descending order
 	isDesc bool
+
+	// how many results to skip
+	offset uint64
+
+	// how many results to return
+	//
+	// if zero, then all the results are returned
+	limit uint64
 }
 
 func (ps *pipeSort) String() string {
@@ -39,6 +47,12 @@ func (ps *pipeSort) String() string {
 	if ps.isDesc {
 		s += " desc"
 	}
+	if ps.offset > 0 {
+		s += fmt.Sprintf(" offset %d", ps.offset)
+	}
+	if ps.limit > 0 {
+		s += fmt.Sprintf(" limit %d", ps.limit)
+	}
 	return s
 }
 
@@ -55,6 +69,13 @@ func (ps *pipeSort) updateNeededFields(neededFields, unneededFields fieldsSet) {
 }
 
 func (ps *pipeSort) newPipeProcessor(workersCount int, stopCh <-chan struct{}, cancel func(), ppBase pipeProcessor) pipeProcessor {
+	if ps.limit > 0 {
+		return newPipeTopkProcessor(ps, workersCount, stopCh, cancel, ppBase)
+	}
+	return newPipeSortProcessor(ps, workersCount, stopCh, cancel, ppBase)
+}
+
+func newPipeSortProcessor(ps *pipeSort, workersCount int, stopCh <-chan struct{}, cancel func(), ppBase pipeProcessor) pipeProcessor {
 	maxStateSize := int64(float64(memory.Allowed()) * 0.2)
 
 	shards := make([]pipeSortProcessorShard, workersCount)
@@ -117,6 +138,9 @@ type pipeSortProcessorShardNopad struct {
 	// stateSizeBudget is the remaining budget for the whole state size for the shard.
 	// The per-shard budget is provided in chunks from the parent pipeSortProcessor.
 	stateSizeBudget int
+
+	// columnValues is used as temporary buffer at pipeSortProcessorShard.writeBlock
+	columnValues [][]string
 }
 
 // sortBlock represents a block of logs for sorting.
@@ -176,16 +200,23 @@ func (shard *pipeSortProcessorShard) writeBlock(br *blockResult) {
 	if len(byFields) == 0 {
 		// Sort by all the columns
 
+		columnValues := shard.columnValues[:0]
+		for _, c := range cs {
+			columnValues = append(columnValues, c.getValues(br))
+		}
+		shard.columnValues = columnValues
+
 		// Generate byColumns
 		var rc resultColumn
+
 		bb := bbPool.Get()
-		for i := range br.timestamps {
+		for rowIdx := range br.timestamps {
-			// JSON-encode all the columns per each row into a single string
+			// Marshal all the columns per each row into a single string
 			// and sort rows by the resulting string.
 			bb.B = bb.B[:0]
-			for _, c := range cs {
+			for i, values := range columnValues {
-				v := c.getValueAtRow(br, i)
+				v := values[rowIdx]
-				bb.B = marshalJSONKeyValue(bb.B, c.name, v)
+				bb.B = marshalJSONKeyValue(bb.B, cs[i].name, v)
 				bb.B = append(bb.B, ',')
 			}
 			rc.addValue(bytesutil.ToUnsafeString(bb.B))
@@ -358,10 +389,8 @@ func (psp *pipeSortProcessor) flush() error {
 		return fmt.Errorf("cannot calculate [%s], since it requires more than %dMB of memory", psp.ps.String(), psp.maxStateSize/(1<<20))
 	}
 
-	select {
+	if needStop(psp.stopCh) {
-	case <-psp.stopCh:
 		return nil
-	default:
 	}
 
 	// Sort every shard in parallel
@@ -377,17 +406,15 @@ func (psp *pipeSortProcessor) flush() error {
 	}
 	wg.Wait()
 
-	select {
+	if needStop(psp.stopCh) {
-	case <-psp.stopCh:
 		return nil
-	default:
 	}
 
 	// Merge sorted results across shards
 	sh := pipeSortProcessorShardsHeap(make([]*pipeSortProcessorShard, 0, len(shards)))
 	for i := range shards {
 		shard := &shards[i]
-		if shard.Len() > 0 {
+		if len(shard.rowRefs) > 0 {
 			sh = append(sh, shard)
 		}
 	}
@@ -400,49 +427,43 @@ func (psp *pipeSortProcessor) flush() error {
 	wctx := &pipeSortWriteContext{
 		psp: psp,
 	}
-	var shardNext *pipeSortProcessorShard
+	shardNextIdx := 0
 
 	for len(sh) > 1 {
 		shard := sh[0]
-		wctx.writeRow(shard, shard.rowRefNext)
+		wctx.writeNextRow(shard)
-		shard.rowRefNext++
 
 		if shard.rowRefNext >= len(shard.rowRefs) {
 			_ = heap.Pop(&sh)
-			shardNext = nil
+			shardNextIdx = 0
 
-			select {
+			if needStop(psp.stopCh) {
-			case <-psp.stopCh:
 				return nil
-			default:
 			}
 
 			continue
 		}
 
-		if shardNext == nil {
+		if shardNextIdx == 0 {
-			shardNext = sh[1]
+			shardNextIdx = 1
-			if len(sh) > 2 && sortBlockLess(sh[2], sh[2].rowRefNext, shardNext, shardNext.rowRefNext) {
+			if len(sh) > 2 && sh.Less(2, 1) {
-				shardNext = sh[2]
+				shardNextIdx = 2
 			}
 		}
 
-		if sortBlockLess(shardNext, shardNext.rowRefNext, shard, shard.rowRefNext) {
+		if sh.Less(shardNextIdx, 0) {
 			heap.Fix(&sh, 0)
-			shardNext = nil
+			shardNextIdx = 0
 
-			select {
+			if needStop(psp.stopCh) {
-			case <-psp.stopCh:
 				return nil
-			default:
 			}
 		}
 	}
 	if len(sh) == 1 {
 		shard := sh[0]
 		for shard.rowRefNext < len(shard.rowRefs) {
-			wctx.writeRow(shard, shard.rowRefNext)
+			wctx.writeNextRow(shard)
-			shard.rowRefNext++
 		}
 	}
 	wctx.flush()
@@ -455,14 +476,25 @@ type pipeSortWriteContext struct {
 	rcs []resultColumn
 	br  blockResult
 
-	valuesLen int
+	rowsWritten uint64
+	valuesLen   int
 }
 
-func (wctx *pipeSortWriteContext) writeRow(shard *pipeSortProcessorShard, rowIdx int) {
+func (wctx *pipeSortWriteContext) writeNextRow(shard *pipeSortProcessorShard) {
+	ps := shard.ps
+
+	rowIdx := shard.rowRefNext
+	shard.rowRefNext++
+
+	wctx.rowsWritten++
+	if wctx.rowsWritten <= ps.offset {
+		return
+	}
+
 	rr := shard.rowRefs[rowIdx]
 	b := &shard.blocks[rr.blockIdx]
 
-	byFields := shard.ps.byFields
+	byFields := ps.byFields
 	rcs := wctx.rcs
 
 	areEqualColumns := len(rcs) == len(byFields)+len(b.otherColumns)
@@ -671,7 +703,36 @@ func parsePipeSort(lex *lexer) (*pipeSort, error) {
 		ps.isDesc = true
 	}
 
-	return &ps, nil
+	for {
+		switch {
+		case lex.isKeyword("offset"):
+			lex.nextToken()
+			s := lex.token
+			n, ok := tryParseUint64(s)
+			lex.nextToken()
+			if !ok {
+				return nil, fmt.Errorf("cannot parse 'offset %s'", s)
+			}
+			if ps.offset > 0 {
+				return nil, fmt.Errorf("duplicate 'offset'; the previous one is %d; the new one is %s", ps.offset, s)
+			}
+			ps.offset = n
+		case lex.isKeyword("limit"):
+			lex.nextToken()
+			s := lex.token
+			n, ok := tryParseUint64(s)
+			lex.nextToken()
+			if !ok {
+				return nil, fmt.Errorf("cannot parse 'limit %s'", s)
+			}
+			if ps.limit > 0 {
+				return nil, fmt.Errorf("duplicate 'limit'; the previous one is %d; the new one is %s", ps.limit, s)
+			}
+			ps.limit = n
+		default:
+			return &ps, nil
+		}
+	}
 }
 
 // bySortField represents 'by (...)' part of the pipeSort.
@@ -725,13 +786,6 @@ func parseBySortFields(lex *lexer) ([]*bySortField, error) {
 	}
 }
 
-func marshalJSONKeyValue(dst []byte, k, v string) []byte {
-	dst = strconv.AppendQuote(dst, k)
-	dst = append(dst, ':')
-	dst = strconv.AppendQuote(dst, v)
-	return dst
-}
-
 func tryParseInt64(s string) (int64, bool) {
 	if len(s) == 0 {
 		return 0, false
@@ -756,3 +810,10 @@ func tryParseInt64(s string) (int64, bool) {
 	}
 	return -int64(u64), true
 }
+
+func marshalJSONKeyValue(dst []byte, k, v string) []byte {
+	dst = strconv.AppendQuote(dst, k)
+	dst = append(dst, ':')
+	dst = strconv.AppendQuote(dst, v)
+	return dst
+}

lib/logstorage/pipe_stats.go

@@ -345,10 +345,8 @@ func (psp *pipeStatsProcessor) flush() error {
 		for key, psg := range shard.m {
 			// shard.m may be quite big, so this loop can take a lot of time and CPU.
 			// Stop processing data as soon as stopCh is closed without wasting additional CPU time.
-			select {
+			if needStop(psp.stopCh) {
-			case <-psp.stopCh:
 				return nil
-			default:
 			}
 
 			spgBase := m[key]
@@ -388,10 +386,8 @@ func (psp *pipeStatsProcessor) flush() error {
 	for key, psg := range m {
 		// m may be quite big, so this loop can take a lot of time and CPU.
 		// Stop processing data as soon as stopCh is closed without wasting additional CPU time.
-		select {
+		if needStop(psp.stopCh) {
-		case <-psp.stopCh:
 			return nil
-		default:
 		}
 
 		// Unmarshal values for byFields from key.
@@ -534,6 +530,24 @@ func parseStatsFunc(lex *lexer) (statsFunc, string, error) {
 			return nil, "", fmt.Errorf("cannot parse 'values' func: %w", err)
 		}
 		sf = svs
+	case lex.isKeyword("sum_len"):
+		sss, err := parseStatsSumLen(lex)
+		if err != nil {
+			return nil, "", fmt.Errorf("cannot parse 'sum_len' func: %w", err)
+		}
+		sf = sss
+	case lex.isKeyword("quantile"):
+		sqs, err := parseStatsQuantile(lex)
+		if err != nil {
+			return nil, "", fmt.Errorf("cannot parse 'quantile' func: %w", err)
+		}
+		sf = sqs
+	case lex.isKeyword("median"):
+		sms, err := parseStatsMedian(lex)
+		if err != nil {
+			return nil, "", fmt.Errorf("cannot parse 'median' func: %w", err)
+		}
+		sf = sms
 	default:
 		return nil, "", fmt.Errorf("unknown stats func %q", lex.token)
 	}

lib/logstorage/pipe_topk.go

@@ -0,0 +1,553 @@
+package logstorage
+
+import (
+	"container/heap"
+	"fmt"
+	"strings"
+	"sync"
+	"sync/atomic"
+	"unsafe"
+
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/stringsutil"
+)
+
+func newPipeTopkProcessor(ps *pipeSort, workersCount int, stopCh <-chan struct{}, cancel func(), ppBase pipeProcessor) pipeProcessor {
+	maxStateSize := int64(float64(memory.Allowed()) * 0.2)
+
+	shards := make([]pipeTopkProcessorShard, workersCount)
+	for i := range shards {
+		shard := &shards[i]
+		shard.ps = ps
+		shard.stateSizeBudget = stateSizeBudgetChunk
+		maxStateSize -= stateSizeBudgetChunk
+	}
+
+	ptp := &pipeTopkProcessor{
+		ps:     ps,
+		stopCh: stopCh,
+		cancel: cancel,
+		ppBase: ppBase,
+
+		shards: shards,
+
+		maxStateSize: maxStateSize,
+	}
+	ptp.stateSizeBudget.Store(maxStateSize)
+
+	return ptp
+}
+
+type pipeTopkProcessor struct {
+	ps     *pipeSort
+	stopCh <-chan struct{}
+	cancel func()
+	ppBase pipeProcessor
+
+	shards []pipeTopkProcessorShard
+
+	maxStateSize    int64
+	stateSizeBudget atomic.Int64
+}
+
+type pipeTopkProcessorShard struct {
+	pipeTopkProcessorShardNopad
+
+	// The padding prevents false sharing on widespread platforms with 128 mod (cache line size) = 0 .
+	_ [128 - unsafe.Sizeof(pipeTopkProcessorShardNopad{})%128]byte
+}
+
+type pipeTopkProcessorShardNopad struct {
+	// ps points to the parent pipeSort.
+	ps *pipeSort
+
+	// rows contains rows tracked by the given shard.
+	rows []*pipeTopkRow
+
+	// rowNext points to the next index at rows during merge shards phase
+	rowNext int
+
+	// tmpRow is used as a temporary row when determining whether the next ingested row must be stored in the shard.
+	tmpRow pipeTopkRow
+
+	// these are aux fields for determining whether the next row must be stored in rows.
+	byColumnValues    [][]string
+	otherColumnValues []pipeTopkOtherColumn
+	byColumns         []string
+	otherColumns      []Field
+
+	// stateSizeBudget is the remaining budget for the whole state size for the shard.
+	// The per-shard budget is provided in chunks from the parent pipeTopkProcessor.
+	stateSizeBudget int
+}
+
+type pipeTopkRow struct {
+	byColumns    []string
+	otherColumns []Field
+}
+
+type pipeTopkOtherColumn struct {
+	name   string
+	values []string
+}
+
+func (r *pipeTopkRow) clone() *pipeTopkRow {
+	byColumnsCopy := make([]string, len(r.byColumns))
+	for i := range byColumnsCopy {
+		byColumnsCopy[i] = strings.Clone(r.byColumns[i])
+	}
+
+	otherColumnsCopy := make([]Field, len(r.otherColumns))
+	for i := range otherColumnsCopy {
+		src := &r.otherColumns[i]
+		dst := &otherColumnsCopy[i]
+		dst.Name = strings.Clone(src.Name)
+		dst.Value = strings.Clone(src.Value)
+	}
+
+	return &pipeTopkRow{
+		byColumns:    byColumnsCopy,
+		otherColumns: otherColumnsCopy,
+	}
+}
+
+func (r *pipeTopkRow) sizeBytes() int {
+	n := int(unsafe.Sizeof(*r))
+
+	for _, v := range r.byColumns {
+		n += len(v)
+	}
+	n += len(r.byColumns) * int(unsafe.Sizeof(r.byColumns[0]))
+
+	for _, f := range r.otherColumns {
+		n += len(f.Name) + len(f.Value)
+	}
+	n += len(r.otherColumns) * int(unsafe.Sizeof(r.otherColumns[0]))
+
+	return n
+}
+
+func (shard *pipeTopkProcessorShard) Len() int {
+	return len(shard.rows)
+}
+
+func (shard *pipeTopkProcessorShard) Swap(i, j int) {
+	rows := shard.rows
+	rows[i], rows[j] = rows[j], rows[i]
+}
+
+func (shard *pipeTopkProcessorShard) Less(i, j int) bool {
+	rows := shard.rows
+
+	// This is max heap
+	return topkLess(shard.ps, rows[j], rows[i])
+}
+
+func (shard *pipeTopkProcessorShard) Push(x any) {
+	r := x.(*pipeTopkRow)
+	shard.rows = append(shard.rows, r)
+}
+
+func (shard *pipeTopkProcessorShard) Pop() any {
+	rows := shard.rows
+	x := rows[len(rows)-1]
+	rows[len(rows)-1] = nil
+	shard.rows = rows[:len(rows)-1]
+	return x
+}
+
+// writeBlock writes br to shard.
+func (shard *pipeTopkProcessorShard) writeBlock(br *blockResult) {
+	cs := br.getColumns()
+
+	byFields := shard.ps.byFields
+	if len(byFields) == 0 {
+		// Sort by all the fields
+
+		byColumnValues := shard.byColumnValues[:0]
+		for _, c := range cs {
+			byColumnValues = append(byColumnValues, c.getValues(br))
+		}
+		shard.byColumnValues = byColumnValues
+
+		byColumns := shard.byColumns[:0]
+		otherColumns := shard.otherColumns[:0]
+		bb := bbPool.Get()
+		for rowIdx := range br.timestamps {
+			byColumns = byColumns[:0]
+			bb.B = bb.B[:0]
+			for i, values := range byColumnValues {
+				v := values[rowIdx]
+				bb.B = marshalJSONKeyValue(bb.B, cs[i].name, v)
+				bb.B = append(bb.B, ',')
+			}
+			byColumns = append(byColumns, bytesutil.ToUnsafeString(bb.B))
+
+			otherColumns = otherColumns[:0]
+			for i, values := range byColumnValues {
+				otherColumns = append(otherColumns, Field{
+					Name:  cs[i].name,
+					Value: values[rowIdx],
+				})
+			}
+
+			shard.addRow(byColumns, otherColumns)
+		}
+		bbPool.Put(bb)
+		shard.byColumns = byColumns
+		shard.otherColumns = otherColumns
+	} else {
+		// Sort by byFields
+
+		byColumnValues := shard.byColumnValues[:0]
+		for _, bf := range byFields {
+			c := br.getColumnByName(bf.name)
+			byColumnValues = append(byColumnValues, c.getValues(br))
+		}
+		shard.byColumnValues = byColumnValues
+
+		otherColumnValues := shard.otherColumnValues[:0]
+		for _, c := range cs {
+			isByField := false
+			for _, bf := range byFields {
+				if bf.name == c.name {
+					isByField = true
+					break
+				}
+			}
+			if !isByField {
+				otherColumnValues = append(otherColumnValues, pipeTopkOtherColumn{
+					name:   c.name,
+					values: c.getValues(br),
+				})
+			}
+		}
+		shard.otherColumnValues = otherColumnValues
+
+		// add rows to shard
+		byColumns := shard.byColumns[:0]
+		otherColumns := shard.otherColumns[:0]
+		for rowIdx := range br.timestamps {
+			byColumns = byColumns[:0]
+			for _, values := range byColumnValues {
+				byColumns = append(byColumns, values[rowIdx])
+			}
+
+			otherColumns = otherColumns[:0]
+			for _, ocv := range otherColumnValues {
+				otherColumns = append(otherColumns, Field{
+					Name:  ocv.name,
+					Value: ocv.values[rowIdx],
+				})
+			}
+
+			shard.addRow(byColumns, otherColumns)
+		}
+		shard.byColumns = byColumns
+		shard.otherColumns = otherColumns
+	}
+}
+
+func (shard *pipeTopkProcessorShard) addRow(byColumns []string, otherColumns []Field) {
+	r := &shard.tmpRow
+	r.byColumns = byColumns
+	r.otherColumns = otherColumns
+
+	rows := shard.rows
+	if len(rows) > 0 && !topkLess(shard.ps, r, rows[0]) {
+		// Fast path - nothing to add.
+		return
+	}
+
+	// Slow path - add r to shard.rows.
+	r = r.clone()
+	shard.stateSizeBudget -= r.sizeBytes()
+	if uint64(len(rows)) < shard.ps.limit {
+		heap.Push(shard, r)
+		shard.stateSizeBudget -= int(unsafe.Sizeof(r))
+	} else {
+		shard.stateSizeBudget += rows[0].sizeBytes()
+		rows[0] = r
+		heap.Fix(shard, 0)
+	}
+}
+
+func (shard *pipeTopkProcessorShard) sortRows(stopCh <-chan struct{}) {
+	rows := shard.rows
+	for i := len(rows) - 1; i > 0; i-- {
+		x := heap.Pop(shard)
+		rows[i] = x.(*pipeTopkRow)
+
+		if needStop(stopCh) {
+			return
+		}
+	}
+	shard.rows = rows
+}
+
+func (ptp *pipeTopkProcessor) writeBlock(workerID uint, br *blockResult) {
+	if len(br.timestamps) == 0 {
+		return
+	}
+
+	shard := &ptp.shards[workerID]
+
+	for shard.stateSizeBudget < 0 {
+		// steal some budget for the state size from the global budget.
+		remaining := ptp.stateSizeBudget.Add(-stateSizeBudgetChunk)
+		if remaining < 0 {
+			// The state size is too big. Stop processing data in order to avoid OOM crash.
+			if remaining+stateSizeBudgetChunk >= 0 {
+				// Notify worker goroutines to stop calling writeBlock() in order to save CPU time.
+				ptp.cancel()
+			}
+			return
+		}
+		shard.stateSizeBudget += stateSizeBudgetChunk
+	}
+
+	shard.writeBlock(br)
+}
+
+func (ptp *pipeTopkProcessor) flush() error {
+	if n := ptp.stateSizeBudget.Load(); n <= 0 {
+		return fmt.Errorf("cannot calculate [%s], since it requires more than %dMB of memory", ptp.ps.String(), ptp.maxStateSize/(1<<20))
+	}
+
+	if needStop(ptp.stopCh) {
+		return nil
+	}
+
+	// Sort every shard in parallel
+	var wg sync.WaitGroup
+	shards := ptp.shards
+	for i := range shards {
+		wg.Add(1)
+		go func(shard *pipeTopkProcessorShard) {
+			shard.sortRows(ptp.stopCh)
+			wg.Done()
+		}(&shards[i])
+	}
+	wg.Wait()
+
+	if needStop(ptp.stopCh) {
+		return nil
+	}
+
+	// Merge sorted results across shards
+	sh := pipeTopkProcessorShardsHeap(make([]*pipeTopkProcessorShard, 0, len(shards)))
+	for i := range shards {
+		shard := &shards[i]
+		if len(shard.rows) > 0 {
+			sh = append(sh, shard)
+		}
+	}
+	if len(sh) == 0 {
+		return nil
+	}
+
+	heap.Init(&sh)
+
+	wctx := &pipeTopkWriteContext{
+		ptp: ptp,
+	}
+	shardNextIdx := 0
+
+	for len(sh) > 1 {
+		shard := sh[0]
+		if !wctx.writeNextRow(shard) {
+			break
+		}
+
+		if shard.rowNext >= len(shard.rows) {
+			_ = heap.Pop(&sh)
+			shardNextIdx = 0
+
+			if needStop(ptp.stopCh) {
+				return nil
+			}
+
+			continue
+		}
+
+		if shardNextIdx == 0 {
+			shardNextIdx = 1
+			if len(sh) > 2 && sh.Less(2, 1) {
+				shardNextIdx = 2
+			}
+		}
+
+		if sh.Less(shardNextIdx, 0) {
+			heap.Fix(&sh, 0)
+			shardNextIdx = 0
+
+			if needStop(ptp.stopCh) {
+				return nil
+			}
+		}
+	}
+	if len(sh) == 1 {
+		shard := sh[0]
+		for shard.rowNext < len(shard.rows) {
+			if !wctx.writeNextRow(shard) {
+				break
+			}
+		}
+	}
+	wctx.flush()
+
+	return nil
+}
+
+type pipeTopkWriteContext struct {
+	ptp *pipeTopkProcessor
+	rcs []resultColumn
+	br  blockResult
+
+	rowsWritten uint64
+	valuesLen   int
+}
+
+func (wctx *pipeTopkWriteContext) writeNextRow(shard *pipeTopkProcessorShard) bool {
+	ps := shard.ps
+
+	rowIdx := shard.rowNext
+	shard.rowNext++
+
+	wctx.rowsWritten++
+	if wctx.rowsWritten <= ps.offset {
+		return true
+	}
+	if wctx.rowsWritten > ps.offset+ps.limit {
+		return false
+	}
+
+	r := shard.rows[rowIdx]
+
+	byFields := ps.byFields
+	rcs := wctx.rcs
+
+	areEqualColumns := len(rcs) == len(byFields)+len(r.otherColumns)
+	if areEqualColumns {
+		for i, c := range r.otherColumns {
+			if rcs[len(byFields)+i].name != c.Name {
+				areEqualColumns = false
+				break
+			}
+		}
+	}
+	if !areEqualColumns {
+		// send the current block to bbBase and construct a block with new set of columns
+		wctx.flush()
+
+		rcs = wctx.rcs[:0]
+		for _, bf := range byFields {
+			rcs = append(rcs, resultColumn{
+				name: bf.name,
+			})
+		}
+		for _, c := range r.otherColumns {
+			rcs = append(rcs, resultColumn{
+				name: c.Name,
+			})
+		}
+		wctx.rcs = rcs
+	}
+
+	byColumns := r.byColumns
+	for i := range byFields {
+		v := byColumns[i]
+		rcs[i].addValue(v)
+		wctx.valuesLen += len(v)
+	}
+
+	for i, c := range r.otherColumns {
+		v := c.Value
+		rcs[len(byFields)+i].addValue(v)
+		wctx.valuesLen += len(v)
+	}
+
+	if wctx.valuesLen >= 1_000_000 {
+		wctx.flush()
+	}
+
+	return true
+}
+
+func (wctx *pipeTopkWriteContext) flush() {
+	rcs := wctx.rcs
+	br := &wctx.br
+
+	wctx.valuesLen = 0
+
+	if len(rcs) == 0 {
+		return
+	}
+
+	// Flush rcs to ppBase
+	br.setResultColumns(rcs)
+	wctx.ptp.ppBase.writeBlock(0, br)
+	br.reset()
+	for i := range rcs {
+		rcs[i].resetKeepName()
+	}
+}
+
+type pipeTopkProcessorShardsHeap []*pipeTopkProcessorShard
+
+func (sh *pipeTopkProcessorShardsHeap) Len() int {
+	return len(*sh)
+}
+
+func (sh *pipeTopkProcessorShardsHeap) Swap(i, j int) {
+	a := *sh
+	a[i], a[j] = a[j], a[i]
+}
+
+func (sh *pipeTopkProcessorShardsHeap) Less(i, j int) bool {
+	a := *sh
+	shardA := a[i]
+	shardB := a[j]
+	return topkLess(shardA.ps, shardA.rows[shardA.rowNext], shardB.rows[shardB.rowNext])
+}
+
+func (sh *pipeTopkProcessorShardsHeap) Push(x any) {
+	shard := x.(*pipeTopkProcessorShard)
+	*sh = append(*sh, shard)
+}
+
+func (sh *pipeTopkProcessorShardsHeap) Pop() any {
+	a := *sh
+	x := a[len(a)-1]
+	a[len(a)-1] = nil
+	*sh = a[:len(a)-1]
+	return x
+}
+
+func topkLess(ps *pipeSort, a, b *pipeTopkRow) bool {
+	byFields := ps.byFields
+
+	csA := a.byColumns
+	csB := b.byColumns
+
+	for k := range csA {
+		isDesc := ps.isDesc
+		if len(byFields) > 0 && byFields[k].isDesc {
+			isDesc = !isDesc
+		}
+
+		vA := csA[k]
+		vB := csB[k]
+
+		if vA == vB {
+			continue
+		}
+
+		if isDesc {
+			return stringsutil.LessNatural(vB, vA)
+		}
+		return stringsutil.LessNatural(vA, vB)
+	}
+	return false
+}

lib/logstorage/pipe_uniq.go

@@ -209,10 +209,8 @@ func (pup *pipeUniqProcessor) flush() error {
 	m := shards[0].m
 	shards = shards[1:]
 	for i := range shards {
-		select {
+		if needStop(pup.stopCh) {
-		case <-pup.stopCh:
 			return nil
-		default:
 		}
 
 		for k := range shards[i].m {
@@ -229,10 +227,8 @@ func (pup *pipeUniqProcessor) flush() error {
 
 	if len(byFields) == 0 {
 		for k := range m {
-			select {
+			if needStop(pup.stopCh) {
-			case <-pup.stopCh:
 				return nil
-			default:
 			}
 
 			rowFields = rowFields[:0]
@@ -259,10 +255,8 @@ func (pup *pipeUniqProcessor) flush() error {
 		}
 	} else {
 		for k := range m {
-			select {
+			if needStop(pup.stopCh) {
-			case <-pup.stopCh:
 				return nil
-			default:
 			}
 
 			rowFields = rowFields[:0]

lib/logstorage/stats_max.go

@@ -86,7 +86,7 @@ func (smp *statsMaxProcessor) mergeState(sfp statsProcessor) {
 }
 
 func (smp *statsMaxProcessor) finalizeStats() string {
-	return strconv.FormatFloat(smp.max, 'g', -1, 64)
+	return strconv.FormatFloat(smp.max, 'f', -1, 64)
 }
 
 func parseStatsMax(lex *lexer) (*statsMax, error) {

lib/logstorage/stats_median.go

@@ -0,0 +1,65 @@
+package logstorage
+
+import (
+	"slices"
+	"unsafe"
+)
+
+type statsMedian struct {
+	fields       []string
+	containsStar bool
+}
+
+func (sm *statsMedian) String() string {
+	return "median(" + fieldNamesString(sm.fields) + ")"
+}
+
+func (sm *statsMedian) neededFields() []string {
+	return sm.fields
+}
+
+func (sm *statsMedian) newStatsProcessor() (statsProcessor, int) {
+	smp := &statsMedianProcessor{
+		sqp: &statsQuantileProcessor{
+			sq: &statsQuantile{
+				fields:       sm.fields,
+				containsStar: sm.containsStar,
+				phi:          0.5,
+			},
+		},
+	}
+	return smp, int(unsafe.Sizeof(*smp)) + int(unsafe.Sizeof(*smp.sqp)) + int(unsafe.Sizeof(*smp.sqp.sq))
+}
+
+type statsMedianProcessor struct {
+	sqp *statsQuantileProcessor
+}
+
+func (smp *statsMedianProcessor) updateStatsForAllRows(br *blockResult) int {
+	return smp.sqp.updateStatsForAllRows(br)
+}
+
+func (smp *statsMedianProcessor) updateStatsForRow(br *blockResult, rowIdx int) int {
+	return smp.sqp.updateStatsForRow(br, rowIdx)
+}
+
+func (smp *statsMedianProcessor) mergeState(sfp statsProcessor) {
+	src := sfp.(*statsMedianProcessor)
+	smp.sqp.mergeState(src.sqp)
+}
+
+func (smp *statsMedianProcessor) finalizeStats() string {
+	return smp.sqp.finalizeStats()
+}
+
+func parseStatsMedian(lex *lexer) (*statsMedian, error) {
+	fields, err := parseFieldNamesForStatsFunc(lex, "median")
+	if err != nil {
+		return nil, err
+	}
+	sm := &statsMedian{
+		fields:       fields,
+		containsStar: slices.Contains(fields, "*"),
+	}
+	return sm, nil
+}

lib/logstorage/stats_min.go

@@ -86,7 +86,7 @@ func (smp *statsMinProcessor) mergeState(sfp statsProcessor) {
 }
 
 func (smp *statsMinProcessor) finalizeStats() string {
-	return strconv.FormatFloat(smp.min, 'g', -1, 64)
+	return strconv.FormatFloat(smp.min, 'f', -1, 64)
 }
 
 func parseStatsMin(lex *lexer) (*statsMin, error) {

lib/logstorage/stats_quantile.go

@@ -0,0 +1,215 @@
+package logstorage
+
+import (
+	"fmt"
+	"math"
+	"slices"
+	"strconv"
+	"unsafe"
+
+	"github.com/valyala/fastrand"
+)
+
+type statsQuantile struct {
+	fields       []string
+	containsStar bool
+
+	phi float64
+}
+
+func (sq *statsQuantile) String() string {
+	return fmt.Sprintf("quantile(%g, %s)", sq.phi, fieldNamesString(sq.fields))
+}
+
+func (sq *statsQuantile) neededFields() []string {
+	return sq.fields
+}
+
+func (sq *statsQuantile) newStatsProcessor() (statsProcessor, int) {
+	sqp := &statsQuantileProcessor{
+		sq: sq,
+	}
+	return sqp, int(unsafe.Sizeof(*sqp))
+}
+
+type statsQuantileProcessor struct {
+	sq *statsQuantile
+
+	h histogram
+}
+
+func (sqp *statsQuantileProcessor) updateStatsForAllRows(br *blockResult) int {
+	h := &sqp.h
+	stateSizeIncrease := 0
+
+	if sqp.sq.containsStar {
+		for _, c := range br.getColumns() {
+			for _, v := range c.getValues(br) {
+				f, ok := tryParseFloat64(v)
+				if ok {
+					stateSizeIncrease += h.update(f)
+				}
+			}
+		}
+	} else {
+		for _, field := range sqp.sq.fields {
+			c := br.getColumnByName(field)
+			for _, v := range c.getValues(br) {
+				f, ok := tryParseFloat64(v)
+				if ok {
+					stateSizeIncrease += h.update(f)
+				}
+			}
+		}
+	}
+
+	return stateSizeIncrease
+}
+
+func (sqp *statsQuantileProcessor) updateStatsForRow(br *blockResult, rowIdx int) int {
+	h := &sqp.h
+	stateSizeIncrease := 0
+
+	if sqp.sq.containsStar {
+		for _, c := range br.getColumns() {
+			f := c.getFloatValueAtRow(rowIdx)
+			if !math.IsNaN(f) {
+				stateSizeIncrease += h.update(f)
+			}
+		}
+	} else {
+		for _, field := range sqp.sq.fields {
+			c := br.getColumnByName(field)
+			f := c.getFloatValueAtRow(rowIdx)
+			if !math.IsNaN(f) {
+				stateSizeIncrease += h.update(f)
+			}
+		}
+	}
+
+	return stateSizeIncrease
+}
+
+func (sqp *statsQuantileProcessor) mergeState(sfp statsProcessor) {
+	src := sfp.(*statsQuantileProcessor)
+	sqp.h.mergeState(&src.h)
+}
+
+func (sqp *statsQuantileProcessor) finalizeStats() string {
+	q := sqp.h.quantile(sqp.sq.phi)
+	return strconv.FormatFloat(q, 'f', -1, 64)
+}
+
+func parseStatsQuantile(lex *lexer) (*statsQuantile, error) {
+	if !lex.isKeyword("quantile") {
+		return nil, fmt.Errorf("unexpected token: %q; want %q", lex.token, "quantile")
+	}
+	lex.nextToken()
+
+	fields, err := parseFieldNamesInParens(lex)
+	if err != nil {
+		return nil, fmt.Errorf("cannot parse 'quantile' args: %w", err)
+	}
+	if len(fields) < 2 {
+		return nil, fmt.Errorf("'quantile' must have at least two args: phi and field name")
+	}
+
+	// Parse phi
+	phi, ok := tryParseFloat64(fields[0])
+	if !ok {
+		return nil, fmt.Errorf("phi arg in 'quantile' must be floating point number; got %q", fields[0])
+	}
+	if phi < 0 || phi > 1 {
+		return nil, fmt.Errorf("phi arg in 'quantile' must be in the range [0..1]; got %q", fields[0])
+	}
+
+	// Parse fields
+	fields = fields[1:]
+	if slices.Contains(fields, "*") {
+		fields = []string{"*"}
+	}
+
+	sq := &statsQuantile{
+		fields:       fields,
+		containsStar: slices.Contains(fields, "*"),
+
+		phi: phi,
+	}
+	return sq, nil
+}
+
+type histogram struct {
+	a     []float64
+	min   float64
+	max   float64
+	count uint64
+
+	rng fastrand.RNG
+}
+
+func (h *histogram) update(f float64) int {
+	if h.count == 0 || f < h.min {
+		h.min = f
+	}
+	if h.count == 0 || f > h.max {
+		h.max = f
+	}
+
+	h.count++
+	if len(h.a) < maxHistogramSamples {
+		h.a = append(h.a, f)
+		return int(unsafe.Sizeof(f))
+	}
+
+	if n := h.rng.Uint32n(uint32(h.count)); n < uint32(len(h.a)) {
+		h.a[n] = f
+	}
+	return 0
+}
+
+const maxHistogramSamples = 100_000
+
+func (h *histogram) mergeState(src *histogram) {
+	if src.count == 0 {
+		// Nothing to merge
+		return
+	}
+	if h.count == 0 {
+		h.a = append(h.a, src.a...)
+		h.min = src.min
+		h.max = src.max
+		h.count = src.count
+		return
+	}
+
+	h.a = append(h.a, src.a...)
+	if src.min < h.min {
+		h.min = src.min
+	}
+	if src.max > h.max {
+		h.max = src.max
+	}
+	h.count += src.count
+}
+
+func (h *histogram) quantile(phi float64) float64 {
+	if len(h.a) == 0 {
+		return nan
+	}
+	if len(h.a) == 1 {
+		return h.a[0]
+	}
+	if phi <= 0 {
+		return h.min
+	}
+	if phi >= 1 {
+		return h.max
+	}
+
+	slices.Sort(h.a)
+	idx := int(phi * float64(len(h.a)))
+	if idx == len(h.a) {
+		return h.max
+	}
+	return h.a[idx]
+}

lib/logstorage/stats_quantile_test.go

@@ -0,0 +1,55 @@
+package logstorage
+
+import (
+	"math"
+	"testing"
+)
+
+func TestHistogramQuantile(t *testing.T) {
+	f := func(a []float64, phi, qExpected float64) {
+		t.Helper()
+
+		var h histogram
+		for _, f := range a {
+			h.update(f)
+		}
+		q := h.quantile(phi)
+
+		if math.IsNaN(qExpected) {
+			if !math.IsNaN(q) {
+				t.Fatalf("unexpected result for q=%v, phi=%v; got %v; want %v", a, phi, q, qExpected)
+			}
+		} else if q != qExpected {
+			t.Fatalf("unexpected result for q=%v, phi=%v; got %v; want %v", a, phi, q, qExpected)
+		}
+	}
+
+	f(nil, -1, nan)
+	f(nil, 0, nan)
+	f(nil, 0.5, nan)
+	f(nil, 1, nan)
+	f(nil, 10, nan)
+
+	f([]float64{123}, -1, 123)
+	f([]float64{123}, 0, 123)
+	f([]float64{123}, 0.5, 123)
+	f([]float64{123}, 1, 123)
+	f([]float64{123}, 10, 123)
+
+	f([]float64{5, 1}, -1, 1)
+	f([]float64{5, 1}, 0, 1)
+	f([]float64{5, 1}, 0.5-1e-5, 1)
+	f([]float64{5, 1}, 0.5, 5)
+	f([]float64{5, 1}, 1, 5)
+	f([]float64{5, 1}, 10, 5)
+
+	f([]float64{5, 1, 3}, -1, 1)
+	f([]float64{5, 1, 3}, 0, 1)
+	f([]float64{5, 1, 3}, 1.0/3-1e-5, 1)
+	f([]float64{5, 1, 3}, 1.0/3, 3)
+	f([]float64{5, 1, 3}, 2.0/3-1e-5, 3)
+	f([]float64{5, 1, 3}, 2.0/3, 5)
+	f([]float64{5, 1, 3}, 1-1e-5, 5)
+	f([]float64{5, 1, 3}, 1, 5)
+	f([]float64{5, 1, 3}, 10, 5)
+}

lib/logstorage/stats_sum_len.go

@@ -0,0 +1,89 @@
+package logstorage
+
+import (
+	"slices"
+	"strconv"
+	"unsafe"
+)
+
+type statsSumLen struct {
+	fields       []string
+	containsStar bool
+}
+
+func (ss *statsSumLen) String() string {
+	return "sum_len(" + fieldNamesString(ss.fields) + ")"
+}
+
+func (ss *statsSumLen) neededFields() []string {
+	return ss.fields
+}
+
+func (ss *statsSumLen) newStatsProcessor() (statsProcessor, int) {
+	ssp := &statsSumLenProcessor{
+		ss:     ss,
+		sumLen: 0,
+	}
+	return ssp, int(unsafe.Sizeof(*ssp))
+}
+
+type statsSumLenProcessor struct {
+	ss *statsSumLen
+
+	sumLen uint64
+}
+
+func (ssp *statsSumLenProcessor) updateStatsForAllRows(br *blockResult) int {
+	if ssp.ss.containsStar {
+		// Sum all the columns
+		for _, c := range br.getColumns() {
+			ssp.sumLen += c.sumLenValues(br)
+		}
+	} else {
+		// Sum the requested columns
+		for _, field := range ssp.ss.fields {
+			c := br.getColumnByName(field)
+			ssp.sumLen += c.sumLenValues(br)
+		}
+	}
+	return 0
+}
+
+func (ssp *statsSumLenProcessor) updateStatsForRow(br *blockResult, rowIdx int) int {
+	if ssp.ss.containsStar {
+		// Sum all the fields for the given row
+		for _, c := range br.getColumns() {
+			v := c.getValueAtRow(br, rowIdx)
+			ssp.sumLen += uint64(len(v))
+		}
+	} else {
+		// Sum only the given fields for the given row
+		for _, field := range ssp.ss.fields {
+			c := br.getColumnByName(field)
+			v := c.getValueAtRow(br, rowIdx)
+			ssp.sumLen += uint64(len(v))
+		}
+	}
+	return 0
+}
+
+func (ssp *statsSumLenProcessor) mergeState(sfp statsProcessor) {
+	src := sfp.(*statsSumLenProcessor)
+	ssp.sumLen += src.sumLen
+}
+
+func (ssp *statsSumLenProcessor) finalizeStats() string {
+	return strconv.FormatUint(ssp.sumLen, 10)
+}
+
+func parseStatsSumLen(lex *lexer) (*statsSumLen, error) {
+	fields, err := parseFieldNamesForStatsFunc(lex, "sum_len")
+	if err != nil {
+		return nil, err
+	}
+	ss := &statsSumLen{
+		fields:       fields,
+		containsStar: slices.Contains(fields, "*"),
+	}
+	return ss, nil
+}

lib/logstorage/storage_search.go

@@ -182,12 +182,10 @@ func (s *Storage) search(workersCount int, so *genericSearchOptions, stopCh <-ch
 				bsws := bswb.bsws
 				for i := range bsws {
 					bsw := &bsws[i]
-					select {
+					if needStop(stopCh) {
-					case <-stopCh:
 						// The search has been canceled. Just skip all the scheduled work in order to save CPU time.
 						bsw.reset()
 						continue
-					default:
 					}
 
 					bs.search(bsw)
@@ -266,11 +264,9 @@ var partitionSearchConcurrencyLimitCh = make(chan struct{}, cgroup.AvailableCPUs
 type partitionSearchFinalizer func()
 
 func (pt *partition) search(ft *filterTime, sf *StreamFilter, f filter, so *genericSearchOptions, workCh chan<- *blockSearchWorkBatch, stopCh <-chan struct{}) partitionSearchFinalizer {
-	select {
+	if needStop(stopCh) {
-	case <-stopCh:
 		// Do not spend CPU time on search, since it is already stopped.
 		return func() {}
-	default:
 	}
 
 	tenantIDs := so.tenantIDs
@@ -436,10 +432,8 @@ func (p *part) searchByTenantIDs(so *searchOptions, bhss *blockHeaders, workCh c
 	// it is assumed that ibhs are sorted
 	ibhs := p.indexBlockHeaders
 	for len(ibhs) > 0 && len(tenantIDs) > 0 {
-		select {
+		if needStop(stopCh) {
-		case <-stopCh:
 			return
-		default:
 		}
 
 		// locate tenantID equal or bigger than the tenantID in ibhs[0]
@@ -541,10 +535,8 @@ func (p *part) searchByStreamIDs(so *searchOptions, bhss *blockHeaders, workCh c
 	ibhs := p.indexBlockHeaders
 
 	for len(ibhs) > 0 && len(streamIDs) > 0 {
-		select {
+		if needStop(stopCh) {
-		case <-stopCh:
 			return
-		default:
 		}
 
 		// locate streamID equal or bigger than the streamID in ibhs[0]