refactor: use sqlbuider for queries

This is to avoid the contract change in existing v3

docs/api/openapi.yml

@@ -18948,6 +18948,77 @@ paths:
       summary: Get waterfall view for a trace
       tags:
       - tracedetail
+  /api/v4/traces/{traceID}/waterfall:
+    post:
+      deprecated: false
+      description: 'Two-step fetch: minimal fields for all spans to build the tree,
+        full fields only for the visible window. Aggregations are not included in
+        the response.'
+      operationId: GetWaterfallV4
+      parameters:
+      - in: path
+        name: traceID
+        required: true
+        schema:
+          type: string
+      requestBody:
+        content:
+          application/json:
+            schema:
+              $ref: '#/components/schemas/SpantypesPostableWaterfall'
+      responses:
+        "200":
+          content:
+            application/json:
+              schema:
+                properties:
+                  data:
+                    $ref: '#/components/schemas/SpantypesGettableWaterfallTrace'
+                  status:
+                    type: string
+                required:
+                - status
+                - data
+                type: object
+          description: OK
+        "400":
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/RenderErrorResponse'
+          description: Bad Request
+        "401":
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/RenderErrorResponse'
+          description: Unauthorized
+        "403":
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/RenderErrorResponse'
+          description: Forbidden
+        "404":
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/RenderErrorResponse'
+          description: Not Found
+        "500":
+          content:
+            application/json:
+              schema:
+                $ref: '#/components/schemas/RenderErrorResponse'
+          description: Internal Server Error
+      security:
+      - api_key:
+        - VIEWER
+      - tokenizer:
+        - VIEWER
+      summary: Get waterfall view for a trace (OOM-safe)
+      tags:
+      - tracedetail
   /api/v5/query_range:
     post:
       deprecated: false

frontend/src/api/generated/services/sigNoz.schemas.ts

@@ -9232,6 +9232,17 @@ export type GetWaterfall200 = {
 	status: string;
 };
 
+export type GetWaterfallV4PathParameters = {
+	traceID: string;
+};
+export type GetWaterfallV4200 = {
+	data: SpantypesGettableWaterfallTraceDTO;
+	/**
+	 * @type string
+	 */
+	status: string;
+};
+
 export type QueryRangeV5200 = {
 	data: Querybuildertypesv5QueryRangeResponseDTO;
 	/**

frontend/src/api/generated/services/tracedetail/index.ts

@@ -14,6 +14,8 @@ import type {
 import type {
 	GetWaterfall200,
 	GetWaterfallPathParameters,
+	GetWaterfallV4200,
+	GetWaterfallV4PathParameters,
 	RenderErrorResponseDTO,
 	SpantypesPostableWaterfallDTO,
 } from '../sigNoz.schemas';
@@ -120,3 +122,102 @@ export const useGetWaterfall = <
 > => {
 	return useMutation(getGetWaterfallMutationOptions(options));
 };
+/**
+ * Two-step fetch: minimal fields for all spans to build the tree, full fields only for the visible window. Aggregations are not included in the response.
+ * @summary Get waterfall view for a trace (OOM-safe)
+ */
+export const getWaterfallV4 = (
+	{ traceID }: GetWaterfallV4PathParameters,
+	spantypesPostableWaterfallDTO?: BodyType<SpantypesPostableWaterfallDTO>,
+	signal?: AbortSignal,
+) => {
+	return GeneratedAPIInstance<GetWaterfallV4200>({
+		url: `/api/v4/traces/${traceID}/waterfall`,
+		method: 'POST',
+		headers: { 'Content-Type': 'application/json' },
+		data: spantypesPostableWaterfallDTO,
+		signal,
+	});
+};
+
+export const getGetWaterfallV4MutationOptions = <
+	TError = ErrorType<RenderErrorResponseDTO>,
+	TContext = unknown,
+>(options?: {
+	mutation?: UseMutationOptions<
+		Awaited<ReturnType<typeof getWaterfallV4>>,
+		TError,
+		{
+			pathParams: GetWaterfallV4PathParameters;
+			data?: BodyType<SpantypesPostableWaterfallDTO>;
+		},
+		TContext
+	>;
+}): UseMutationOptions<
+	Awaited<ReturnType<typeof getWaterfallV4>>,
+	TError,
+	{
+		pathParams: GetWaterfallV4PathParameters;
+		data?: BodyType<SpantypesPostableWaterfallDTO>;
+	},
+	TContext
+> => {
+	const mutationKey = ['getWaterfallV4'];
+	const { mutation: mutationOptions } = options
+		? options.mutation &&
+			'mutationKey' in options.mutation &&
+			options.mutation.mutationKey
+			? options
+			: { ...options, mutation: { ...options.mutation, mutationKey } }
+		: { mutation: { mutationKey } };
+
+	const mutationFn: MutationFunction<
+		Awaited<ReturnType<typeof getWaterfallV4>>,
+		{
+			pathParams: GetWaterfallV4PathParameters;
+			data?: BodyType<SpantypesPostableWaterfallDTO>;
+		}
+	> = (props) => {
+		const { pathParams, data } = props ?? {};
+
+		return getWaterfallV4(pathParams, data);
+	};
+
+	return { mutationFn, ...mutationOptions };
+};
+
+export type GetWaterfallV4MutationResult = NonNullable<
+	Awaited<ReturnType<typeof getWaterfallV4>>
+>;
+export type GetWaterfallV4MutationBody =
+	| BodyType<SpantypesPostableWaterfallDTO>
+	| undefined;
+export type GetWaterfallV4MutationError = ErrorType<RenderErrorResponseDTO>;
+
+/**
+ * @summary Get waterfall view for a trace (OOM-safe)
+ */
+export const useGetWaterfallV4 = <
+	TError = ErrorType<RenderErrorResponseDTO>,
+	TContext = unknown,
+>(options?: {
+	mutation?: UseMutationOptions<
+		Awaited<ReturnType<typeof getWaterfallV4>>,
+		TError,
+		{
+			pathParams: GetWaterfallV4PathParameters;
+			data?: BodyType<SpantypesPostableWaterfallDTO>;
+		},
+		TContext
+	>;
+}): UseMutationResult<
+	Awaited<ReturnType<typeof getWaterfallV4>>,
+	TError,
+	{
+		pathParams: GetWaterfallV4PathParameters;
+		data?: BodyType<SpantypesPostableWaterfallDTO>;
+	},
+	TContext
+> => {
+	return useMutation(getGetWaterfallV4MutationOptions(options));
+};

pkg/apiserver/signozapiserver/tracedetail.go

@@ -29,5 +29,24 @@ func (provider *provider) addTraceDetailRoutes(router *mux.Router) error {
 		return err
 	}
 
+	if err := router.Handle("/api/v4/traces/{traceID}/waterfall", handler.New(
+		provider.authzMiddleware.ViewAccess(provider.traceDetailHandler.GetWaterfallV4),
+		handler.OpenAPIDef{
+			ID:                  "GetWaterfallV4",
+			Tags:                []string{"tracedetail"},
+			Summary:             "Get waterfall view for a trace (OOM-safe)",
+			Description:         "Two-step fetch: minimal fields for all spans to build the tree, full fields only for the visible window. Aggregations are not included in the response.",
+			Request:             new(spantypes.PostableWaterfall),
+			RequestContentType:  "application/json",
+			Response:            new(spantypes.GettableWaterfallTrace),
+			ResponseContentType: "application/json",
+			SuccessStatusCode:   http.StatusOK,
+			ErrorStatusCodes:    []int{http.StatusBadRequest, http.StatusNotFound},
+			SecuritySchemes:     newSecuritySchemes(types.RoleViewer),
+		},
+	)).Methods(http.MethodPost).GetError(); err != nil {
+		return err
+	}
+
 	return nil
 }

pkg/modules/tracedetail/impltracedetail/handler.go

@@ -38,3 +38,24 @@ func (h *handler) GetWaterfall(rw http.ResponseWriter, r *http.Request) {
 
 	render.Success(rw, http.StatusOK, result)
 }
+
+func (h *handler) GetWaterfallV4(rw http.ResponseWriter, r *http.Request) {
+	req := new(spantypes.PostableWaterfall)
+	if err := binding.JSON.BindBody(r.Body, req); err != nil {
+		render.Error(rw, err)
+		return
+	}
+
+	if err := req.Validate(); err != nil {
+		render.Error(rw, err)
+		return
+	}
+
+	result, err := h.module.GetWaterfallV4(r.Context(), mux.Vars(r)["traceID"], req)
+	if err != nil {
+		render.Error(rw, err)
+		return
+	}
+
+	render.Success(rw, http.StatusOK, result)
+}

pkg/modules/tracedetail/impltracedetail/module.go

@@ -45,7 +45,7 @@ func (m *module) GetWaterfall(ctx context.Context, traceID string, req *spantype
 	return spantypes.NewGettableWaterfallTrace(waterfallTrace, selectedSpans, uncollapsedSpans, selectedAllSpans, aggregationResults), nil
 }
 
-// getTraceData returns the waterfall cache for the given traceID with fallback on DB.
+// getTraceData fetches all spans for a trace and builds the WaterfallTrace.
 func (m *module) getTraceData(ctx context.Context, traceID string) (*spantypes.WaterfallTrace, error) {
 	summary, err := m.store.GetTraceSummary(ctx, traceID)
 	if err != nil {
@@ -61,6 +61,86 @@ func (m *module) getTraceData(ctx context.Context, traceID string) (*spantypes.W
 		return nil, spantypes.ErrTraceNotFound
 	}
 
-	traceData := spantypes.NewWaterfallTraceFromSpans(spanItems)
-	return traceData, nil
+	nodes := make([]*spantypes.WaterfallSpan, len(spanItems))
+	for i := range spanItems {
+		nodes[i] = spanItems[i].ToWaterfallSpan()
+	}
+	return spantypes.NewWaterfallTraceFromSpans(nodes), nil
+}
+
+// GetWaterfallV4 is the OOM-safe V4 waterfall.
+// For large traces (NumSpans > effectiveLimit) it uses a two-step fetch:
+// minimal fields for all spans to build the tree, then full fields for the
+// visible window only. Aggregations are not returned.
+func (m *module) GetWaterfallV4(ctx context.Context, traceID string, req *spantypes.PostableWaterfall) (*spantypes.GettableWaterfallTrace, error) {
+	summary, err := m.store.GetTraceSummary(ctx, traceID)
+	if err != nil {
+		return nil, err
+	}
+	effectiveLimit := min(req.Limit, m.config.Waterfall.MaxLimitToSelectAllSpans)
+	if summary.NumSpans > uint64(effectiveLimit) {
+		return m.getWindowedWaterfall(ctx, traceID, req, summary, effectiveLimit)
+	}
+	return m.getFullWaterfall(ctx, traceID, summary)
+}
+
+func (m *module) getFullWaterfall(ctx context.Context, traceID string, summary *spantypes.TraceSummary) (*spantypes.GettableWaterfallTrace, error) {
+	spanItems, err := m.store.GetTraceSpans(ctx, traceID, summary)
+	if err != nil {
+		return nil, err
+	}
+
+	if len(spanItems) == 0 {
+		return nil, spantypes.ErrTraceNotFound
+	}
+
+	nodes := make([]*spantypes.WaterfallSpan, len(spanItems))
+	for i := range spanItems {
+		nodes[i] = spanItems[i].ToWaterfallSpan()
+	}
+	waterfallTrace := spantypes.NewWaterfallTraceFromSpans(nodes)
+	selectedSpans := waterfallTrace.GetAllSpans()
+
+	return spantypes.NewGettableWaterfallTrace(waterfallTrace, selectedSpans, nil, true, nil), nil
+}
+
+// getWindowedWaterfall builds the waterfall tree with minimal data and then returns only a window of full spans.
+func (m *module) getWindowedWaterfall(ctx context.Context, traceID string, req *spantypes.PostableWaterfall, summary *spantypes.TraceSummary, effectiveLimit uint) (*spantypes.GettableWaterfallTrace, error) {
+	// Step 1: minimal fetch → build full tree → select visible window
+	minimalSpans, err := m.store.GetMinimalSpans(ctx, traceID, summary)
+	if err != nil {
+		return nil, err
+	}
+	if len(minimalSpans) == 0 {
+		return nil, spantypes.ErrTraceNotFound
+	}
+
+	nodes := make([]*spantypes.WaterfallSpan, len(minimalSpans))
+	for i := range minimalSpans {
+		nodes[i] = minimalSpans[i].ToWaterfallSpan()
+	}
+	waterfallTrace := spantypes.NewWaterfallTraceFromSpans(nodes)
+
+	selectedSpans, uncollapsedSpans := waterfallTrace.GetSelectedSpans(
+		req.UncollapsedSpans,
+		req.SelectedSpanID,
+		m.config.Waterfall.SpanPageSize,
+		m.config.Waterfall.MaxDepthToAutoExpand,
+	)
+
+	// Step 2: full fetch for the selected window only
+	spanIDs := make([]string, len(selectedSpans))
+	for i, s := range selectedSpans {
+		spanIDs[i] = s.SpanID
+	}
+	fullSpans, err := m.store.GetTraceSpansByIDs(ctx, traceID, summary, spanIDs)
+	if err != nil {
+		return nil, err
+	}
+
+	spantypes.EnrichSelectedSpans(selectedSpans, fullSpans)
+
+	return spantypes.NewGettableWaterfallTrace(
+		waterfallTrace, selectedSpans, uncollapsedSpans, false, nil,
+	), nil
 }

pkg/modules/tracedetail/impltracedetail/store.go

@@ -12,6 +12,9 @@ import (
 	"github.com/SigNoz/signoz/pkg/types/spantypes"
 )
 
+// The $$$$ becomes $$ since go-sqlbuilder escapes $ sign
+const serviceNameCol = "resource_string_service$$$$name"
+
 type traceStore struct {
 	telemetryStore telemetrystore.TelemetryStore
 }
@@ -69,3 +72,64 @@ func (s *traceStore) GetTraceSpans(ctx context.Context, traceID string, summary
 	}
 	return spanItems, nil
 }
+
+func (s *traceStore) GetMinimalSpans(ctx context.Context, traceID string, summary *spantypes.TraceSummary) ([]spantypes.MinimalSpan, error) {
+	sb := sqlbuilder.NewSelectBuilder()
+	sb.Select(
+		"DISTINCT ON (span_id) span_id",
+		"parent_span_id", "timestamp", "duration_nano", "has_error",
+		serviceNameCol,
+	)
+	sb.From(fmt.Sprintf("%s.%s", spantypes.TraceDB, spantypes.TraceTable))
+	sb.Where(
+		sb.E("trace_id", traceID),
+		sb.GE("ts_bucket_start", summary.Start.Unix()-1800),
+		sb.LE("ts_bucket_start", summary.End.Unix()),
+	)
+	sb.OrderByAsc("timestamp")
+	sb.OrderByAsc("name")
+	query, args := sb.BuildWithFlavor(sqlbuilder.ClickHouse)
+
+	var spans []spantypes.MinimalSpan
+	if err := s.telemetryStore.ClickhouseDB().Select(ctx, &spans, query, args...); err != nil {
+		return nil, errors.WrapInternalf(err, errors.CodeInternal, "error querying minimal spans")
+	}
+	return spans, nil
+}
+
+func (s *traceStore) GetTraceSpansByIDs(ctx context.Context, traceID string, summary *spantypes.TraceSummary, spanIDs []string) ([]spantypes.StorableSpan, error) {
+	if len(spanIDs) == 0 {
+		return []spantypes.StorableSpan{}, nil
+	}
+	sb := sqlbuilder.NewSelectBuilder()
+	sb.Select(
+		"DISTINCT ON (span_id) timestamp",
+		"duration_nano", "span_id", "trace_id", "has_error", "kind",
+		serviceNameCol, "name", "links as references",
+		"attributes_string", "attributes_number", "attributes_bool", "resources_string",
+		"events", "status_message", "status_code_string", "kind_string", "parent_span_id",
+		"flags", "is_remote", "trace_state", "status_code",
+		"db_name", "db_operation", "http_method", "http_url", "http_host",
+		"external_http_method", "external_http_url", "response_status_code",
+	)
+	sb.From(fmt.Sprintf("%s.%s", spantypes.TraceDB, spantypes.TraceTable))
+	ids := make([]any, len(spanIDs))
+	for i, id := range spanIDs {
+		ids[i] = id
+	}
+	sb.Where(
+		sb.E("trace_id", traceID),
+		sb.In("span_id", ids...),
+		sb.GE("ts_bucket_start", summary.Start.Unix()-1800),
+		sb.LE("ts_bucket_start", summary.End.Unix()),
+	)
+	sb.OrderByAsc("timestamp")
+	sb.OrderByAsc("name")
+	query, args := sb.BuildWithFlavor(sqlbuilder.ClickHouse)
+
+	var spans []spantypes.StorableSpan
+	if err := s.telemetryStore.ClickhouseDB().Select(ctx, &spans, query, args...); err != nil {
+		return nil, errors.WrapInternalf(err, errors.CodeInternal, "error querying trace spans by IDs")
+	}
+	return spans, nil
+}

pkg/modules/tracedetail/tracedetail.go

@@ -10,9 +10,11 @@ import (
 // Handler exposes HTTP handlers for trace detail APIs.
 type Handler interface {
 	GetWaterfall(http.ResponseWriter, *http.Request)
+	GetWaterfallV4(http.ResponseWriter, *http.Request)
 }
 
 // Module defines the business logic for trace detail operations.
 type Module interface {
 	GetWaterfall(ctx context.Context, traceID string, req *spantypes.PostableWaterfall) (*spantypes.GettableWaterfallTrace, error)
+	GetWaterfallV4(ctx context.Context, traceID string, req *spantypes.PostableWaterfall) (*spantypes.GettableWaterfallTrace, error)
 }

pkg/types/spantypes/store.go

@@ -26,4 +26,6 @@ type SpanMapperStore interface {
 type TraceStore interface {
 	GetTraceSummary(ctx context.Context, traceID string) (*TraceSummary, error)
 	GetTraceSpans(ctx context.Context, traceID string, summary *TraceSummary) ([]StorableSpan, error)
+	GetMinimalSpans(ctx context.Context, traceID string, summary *TraceSummary) ([]MinimalSpan, error)
+	GetTraceSpansByIDs(ctx context.Context, traceID string, summary *TraceSummary, spanIDs []string) ([]StorableSpan, error)
 }

pkg/types/spantypes/waterfall_span.go

@@ -132,6 +132,31 @@ type StorableSpan struct {
 	ResponseStatusCode string             `ch:"response_status_code"`
 }
 
+// MinimalSpan with only the fields needed to build the parent-child tree.
+type MinimalSpan struct {
+	SpanID       string    `ch:"span_id"`
+	ParentSpanID string    `ch:"parent_span_id"`
+	StartTime    time.Time `ch:"timestamp"`
+	DurationNano uint64    `ch:"duration_nano"`
+	HasError     bool      `ch:"has_error"`
+	ServiceName  string    `ch:"resource_string_service$$name"`
+}
+
+func (item *MinimalSpan) ToWaterfallSpan() *WaterfallSpan {
+	return &WaterfallSpan{
+		SpanID:       item.SpanID,
+		ParentSpanID: item.ParentSpanID,
+		TimeUnix:     uint64(item.StartTime.UnixNano()),
+		DurationNano: item.DurationNano,
+		HasError:     item.HasError,
+		ServiceName:  item.ServiceName,
+		Resource:     map[string]string{"service.name": item.ServiceName},
+		Children:     make([]*WaterfallSpan, 0),
+		Attributes:   make(map[string]any),
+		Events:       make([]Event, 0),
+	}
+}
+
 // NewMissingWaterfallSpan creates a synthetic placeholder span for a parent that has no recorded data.
 func NewMissingWaterfallSpan(spanID, traceID string, timeUnixNano, durationNano uint64) *WaterfallSpan {
 	return &WaterfallSpan{
@@ -297,6 +322,24 @@ func (item *StorableSpan) ToWaterfallSpan() *WaterfallSpan {
 	}
 }
 
+func EnrichSelectedSpans(window []*WaterfallSpan, fullSpans []StorableSpan) {
+	fullByID := make(map[string]*StorableSpan, len(fullSpans))
+	for i := range fullSpans {
+		fullByID[fullSpans[i].SpanID] = &fullSpans[i]
+	}
+	for i, ws := range window {
+		full, ok := fullByID[ws.SpanID]
+		if !ok {
+			continue // synthesized MissingSpan — keep empty shell
+		}
+		newWS := full.ToWaterfallSpan()
+		newWS.Level = ws.Level
+		newWS.HasChildren = ws.HasChildren
+		newWS.SubTreeNodeCount = ws.SubTreeNodeCount
+		window[i] = newWS
+	}
+}
+
 // getSpanIndex returns the index of matched span and -1 for no match.
 func getSpanIndex(spans []*WaterfallSpan, targetSpanID string) int {
 	for i, s := range spans {

pkg/types/spantypes/waterfall_trace.go

@@ -62,26 +62,24 @@ func NewWaterfallTrace(
 	}
 }
 
-func NewWaterfallTraceFromSpans(spans []StorableSpan) *WaterfallTrace {
+// NewWaterfallTraceFromSpans requires WaterfallSpan nodes with only below fields:
+// SpanID, ParentSpanID, TimeUnix, DurationNano, HasError, and ServiceName.
+func NewWaterfallTraceFromSpans(nodes []*WaterfallSpan) *WaterfallTrace {
 	var (
 		startTime, endTime, totalErrorSpans uint64
-		spanIDToSpanNodeMap                 = make(map[string]*WaterfallSpan, len(spans))
+		spanIDToSpanNodeMap                 = make(map[string]*WaterfallSpan, len(nodes))
 		traceRoots                          []*WaterfallSpan
 		hasMissingSpans                     bool
 	)
 
-	for _, item := range spans {
-		span := item.ToWaterfallSpan()
-		startTimeUnixNano := uint64(item.StartTime.UnixNano())
-		if startTime == 0 || startTimeUnixNano < startTime {
-			startTime = startTimeUnixNano
+	for _, span := range nodes {
+		if startTime == 0 || span.TimeUnix < startTime {
+			startTime = span.TimeUnix
 		}
-		endTime = max(endTime, startTimeUnixNano+span.DurationNano)
-
+		endTime = max(endTime, span.TimeUnix+span.DurationNano)
 		if span.HasError {
 			totalErrorSpans++
 		}
-
 		spanIDToSpanNodeMap[span.SpanID] = span
 	}
 
@@ -116,7 +114,7 @@ func NewWaterfallTraceFromSpans(spans []StorableSpan) *WaterfallTrace {
 	return NewWaterfallTrace(
 		startTime,
 		endTime,
-		uint64(len(spans)),
+		uint64(len(nodes)),
 		totalErrorSpans,
 		spanIDToSpanNodeMap,
 		traceRoots,