pindex_bleve.go

//  Copyright 2014-Present Couchbase, Inc.
//
//  Use of this software is governed by the Business Source License included
//  in the file licenses/BSL-Couchbase.txt.  As of the Change Date specified
//  in that file, in accordance with the Business Source License, use of this
//  software will be governed by the Apache License, Version 2.0, included in
//  the file licenses/APL2.txt.

package cbft

import (
	"container/heap"
	"context"
	"encoding/binary"
	"encoding/json"
	"fmt"
	"io"
	"net/http"
	"net/url"
	"os"
	"reflect"
	"strconv"
	"strings"
	"sync"
	"sync/atomic"
	"time"

	"github.com/gorilla/mux"

	bleveMappingUI "github.com/blevesearch/bleve-mapping-ui"
	"github.com/blevesearch/bleve/v2"
	_ "github.com/blevesearch/bleve/v2/config"
	"github.com/blevesearch/bleve/v2/index/scorch"
	"github.com/blevesearch/bleve/v2/mapping"
	bleveRegistry "github.com/blevesearch/bleve/v2/registry"
	"github.com/blevesearch/bleve/v2/search"
	"github.com/blevesearch/bleve/v2/search/query"
	ftsHttp "github.com/couchbase/cbft/http"
	"github.com/couchbase/cbgt"
	"github.com/couchbase/cbgt/rest"
	log "github.com/couchbase/clog"
)

const (
	featureIndexType   = "indexType"
	FeatureScorchIndex = featureIndexType + ":" + scorch.Name

	FeatureCollections  = cbgt.SOURCE_GOCBCORE + ":collections"
	FeatureGeoSpatial   = "geoSpatial"
	featureVectorSearch = "vectors"
	FeatureXattrs       = "xattrs"

	// bleveLegacyZapVersion represents the default zap version.
	// This version is expected to remain a constant as all the
	// future indexes are going to have a default segment version.
	// Only pre CC indexes are expected to have an empty segment version
	// which would be treated like the default zap version.
	bleveLegacyZapVersion = int(11)

	// ZapVersion where Collections support was introduced.
	bleveCollectionsZapVersion = int(15)

	// ZapVersion where Vector support was introduced.
	bleveVectorZapVersion = int(16)

	// blevePreferredZapVersion is the recommended zap version for newer indexes.
	// This version needs to be bumped to reflect the latest recommended zap
	// version in any given release.
	blevePreferredZapVersion = int(16)

	xattrsMappingName          = "_$xattrs"
	DefaultBleveMaxClauseCount = 1024
)

var (
	FeatureBlevePreferredSegmentVersion = fmt.Sprintf("segmentVersion:%d", blevePreferredZapVersion)

	// Use sync/atomic to access these stats
	BatchBytesAdded      uint64
	BatchBytesRemoved    uint64
	NumBatchesIntroduced uint64

	TotBatchesFlushedOnMaxOps uint64
	TotBatchesFlushedOnTimer  uint64
	TotBatchesNew             uint64
	TotBatchesMerged          uint64

	TotRollbackPartial uint64
	TotRollbackFull    uint64

	BleveMaxOpsPerBatch = 200 // Unlimited when <= 0.

	BleveBatchFlushDuration = time.Duration(100 * time.Millisecond)

	// represents the number of async batch workers per pindex
	asyncBatchWorkerCount = 4 // need to make it configurable,

	TotBleveDestOpened uint64
	TotBleveDestClosed uint64

	defaultLimitingMinTime = 500
	defaultLimitingMaxTime = 120000
)

// BleveParams represents the bleve index params.  See also
// cbgt.IndexDef.Params.  A JSON'ified BleveParams looks like...
//
//	{
//	   "mapping": {
//	      // See bleve.mapping.IndexMapping.
//	   },
//	   "store": {
//	      // See BleveParamsStore.
//	   },
//	   "doc_config": {
//	      // See BleveDocumentConfig.
//	   }
//	}
type BleveParams struct {
	Mapping   mapping.IndexMapping   `json:"mapping"`
	Store     map[string]interface{} `json:"store"`
	DocConfig BleveDocumentConfig    `json:"doc_config"`
}

// BleveParamsStore represents some of the publically available
// options in the "store" part of a bleve index params.  See also the
// BleveParams.Store field.
type BleveParamsStore struct {
	// The indexType defaults to bleve.Config.DefaultIndexType.
	// Example: "scorch".  See bleve.index.scorch.Name and
	// bleve.registry.RegisterIndexType().
	IndexType string `json:"indexType"`

	// The kvStoreName defaults to bleve.Config.DefaultKVStore.
	// See also bleve.registry.RegisterKVStore().
	KvStoreName string `json:"kvStoreName"`
}

func NewBleveParams() *BleveParams {
	rv := &BleveParams{
		Mapping: bleve.NewIndexMapping(),
		Store: map[string]interface{}{
			"indexType":   bleve.Config.DefaultIndexType,
			"kvStoreName": bleve.Config.DefaultKVStore,
		},
		DocConfig: BleveDocumentConfig{
			Mode:      "type_field",
			TypeField: "type",
		},
	}

	return rv
}

func (sr *SearchRequest) decorateQuery(indexName string, q query.Query,
	cache *collMetaFieldCache) (query.Query, query.Query) {
	var docIDQuery *query.DocIDQuery
	var ok bool
	// bail out early if the query is not a docID one and there are
	// no target collections requested in search request.
	if docIDQuery, ok = q.(*query.DocIDQuery); !ok && len(sr.Collections) == 0 {
		return nil, q
	}
	if cache == nil {
		cache = metaFieldValCache
	}
	// bail out early if the index is a single collection index as there
	// won't be any docID decorations done during indexing as well as the
	// collection scoping during the querying also redundant.
	var sdm *sourceDetails
	if sdm, ok = cache.getSourceDetailsMap(indexName); !ok ||
		len(sdm.collUIDNameMap) <= 1 {
		return nil, q
	}

	// if this is a multi collection index and the query is for docID,
	// then decorate the target docIDs with cuid prefixes.
	if docIDQuery != nil {
		decoratedQuery := *docIDQuery

		hash := make(map[string]struct{})
		for _, cname := range sr.Collections {
			hash[cname] = struct{}{}
		}
		newIDs := make([]string, 0, len(decoratedQuery.IDs))
		for cuid, cname := range sdm.collUIDNameMap {
			if _, ok := hash[cname]; !ok && len(hash) > 0 {
				continue
			}
			cBytes := make([]byte, 4)
			binary.LittleEndian.PutUint32(cBytes, cuid)
			for _, docID := range decoratedQuery.IDs {
				newIDs = append(newIDs, string(append(cBytes,
					[]byte(docID)...)))
			}
		}
		decoratedQuery.IDs = newIDs
		return q, &decoratedQuery
	}

	// if the search is scoped to specific collections then add
	// collection specific conjunctions with multi collection indexes.
	cjnq := query.NewConjunctionQuery([]query.Query{q})
	djnq := query.NewDisjunctionQuery(nil)

	for _, col := range sr.Collections {
		queryStr := cache.getMetaFieldValue(indexName, col)
		mq := query.NewMatchQuery(queryStr)
		mq.Analyzer = "keyword"
		mq.SetField(CollMetaFieldName)
		djnq.AddQuery(mq)
	}
	djnq.SetMin(1)
	cjnq.AddQuery(djnq)
	return q, cjnq
}

type SearchRequest struct {
	Q                json.RawMessage         `json:"query"`
	Size             *int                    `json:"size"`
	From             *int                    `json:"from"`
	Highlight        *bleve.HighlightRequest `json:"highlight"`
	Fields           []string                `json:"fields"`
	Facets           bleve.FacetsRequest     `json:"facets"`
	Explain          bool                    `json:"explain"`
	Sort             []json.RawMessage       `json:"sort"`
	IncludeLocations bool                    `json:"includeLocations"`
	Score            string                  `json:"score,omitempty"`
	SearchAfter      []string                `json:"search_after,omitempty"`
	SearchBefore     []string                `json:"search_before,omitempty"`
	Limit            *int                    `json:"limit,omitempty"`
	Offset           *int                    `json:"offset,omitempty"`
	Collections      []string                `json:"collections,omitempty"`
	KNN              json.RawMessage         `json:"knn,omitempty"`
	KNNOperator      json.RawMessage         `json:"knn_operator,omitempty"`
	PreSearchData    json.RawMessage         `json:"pre_search_data,omitempty"`
}

func (sr *SearchRequest) ConvertToBleveSearchRequest() (*bleve.SearchRequest, error) {
	// size/from take precedence, but if not specified, overwrite with
	// limit/offset settings
	r := &bleve.SearchRequest{
		Highlight:        sr.Highlight,
		Fields:           sr.Fields,
		Facets:           sr.Facets,
		Explain:          sr.Explain,
		IncludeLocations: sr.IncludeLocations,
		Score:            sr.Score,
		SearchAfter:      sr.SearchAfter,
		SearchBefore:     sr.SearchBefore,
	}

	var err error
	r.Query, err = query.ParseQuery(sr.Q)
	if err != nil {
		return nil, err
	}
	if sr.PreSearchData != nil {
		r.PreSearchData, err = query.ParsePreSearchData(sr.PreSearchData)
		if err != nil {
			return nil, err
		}
	}

	if sr.Size == nil {
		if sr.Limit == nil || *sr.Limit < 0 {
			r.Size = 10
		} else {
			r.Size = *sr.Limit
		}
	} else if *sr.Size < 0 {
		r.Size = 10
	} else {
		r.Size = *sr.Size
	}

	if sr.From == nil {
		if sr.Offset == nil || *sr.Offset < 0 {
			r.From = 0
		} else {
			r.From = *sr.Offset
		}
	} else if *sr.From < 0 {
		r.From = 0
	} else {
		r.From = *sr.From
	}

	if sr.Sort == nil {
		r.Sort = search.SortOrder{&search.SortScore{Desc: true}}
	} else {
		r.Sort, err = search.ParseSortOrderJSON(sr.Sort)
		if err != nil {
			return nil, err
		}
	}

	if r, err = interpretKNNForRequest(sr.KNN, sr.KNNOperator, r); err != nil {
		return nil, err
	}

	return r, r.Validate()
}

type RollbackInfo struct {
	partition   string
	vBucketUUID uint64
	rollbackSeq uint64
}

type BleveDest struct {
	path       string
	sourceName string

	bleveDocConfig BleveDocumentConfig

	// Invoked when mgr should rollback this BleveDest
	rollback     func()
	rollbackInfo *RollbackInfo

	stats     *cbgt.PIndexStoreStats
	copyStats *CopyPartitionStats

	m                    sync.RWMutex // Protects the fields that follow.
	bindex               bleve.Index
	partitions           map[string]*BleveDestPartition
	rev                  uint64 // Incremented whenever bindex changes.
	batchReqChs          []chan *batchRequest
	stopCh               chan struct{}
	removeCh             chan struct{}
	isRollbackInProgress bool
}

// Used to track state for a single partition.
type BleveDestPartition struct {
	bdest           *BleveDest
	bindex          bleve.Index
	partition       string
	partitionBytes  []byte
	partitionOpaque []byte // Key used to implement OpaqueSet/OpaqueGet().

	seqMax      uint64       // Max seq # we've seen for this partition.
	seqMaxBatch uint64       // Max seq # that got through batch apply/commit.
	lastUUID    atomic.Value // type: string; Cache most recent partition UUID from lastOpaque.

	m                 sync.Mutex   // Protects the fields that follow.
	seqSnapEnd        uint64       // To track snapshot end seq # for this partition.
	osoSnapshot       bool         // Flag to track if current seq # is within an OSO Snapshot.
	osoSeqMax         uint64       // Max seq # received within an OSO Snapshot.
	batch             *bleve.Batch // Batch applied when we hit seqSnapEnd.
	lastOpaque        []byte       // Cache most recent value for OpaqueSet()/OpaqueGet().
	cwrQueue          cbgt.CwrQueue
	lastAsyncBatchErr error // for returning async batch err on next call
}

type batchRequest struct {
	bdp    *BleveDestPartition
	bindex bleve.Index
	batch  *bleve.Batch
	seqMax uint64 // a local copy of seqMax
}

func NewBleveDestEx(path string, bindex bleve.Index,
	rollback func(), bleveDocConfig BleveDocumentConfig, sourceName string) *BleveDest {

	bleveDest := &BleveDest{
		path:           path,
		bleveDocConfig: bleveDocConfig,
		rollback:       rollback,
		bindex:         bindex,
		partitions:     make(map[string]*BleveDestPartition),
		stats:          cbgt.NewPIndexStoreStats(),
		copyStats:      &CopyPartitionStats{},
		stopCh:         make(chan struct{}),
		removeCh:       make(chan struct{}),
		sourceName:     sourceName,
	}

	bleveDest.batchReqChs = make([]chan *batchRequest, asyncBatchWorkerCount)

	bleveDest.startBatchWorkers()

	atomic.AddUint64(&TotBleveDestOpened, 1)

	return bleveDest
}

func NewBleveDest(path string, bindex bleve.Index,
	rollback func(), bleveDocConfig BleveDocumentConfig) *BleveDest {

	bleveDest := &BleveDest{
		path:           path,
		bleveDocConfig: bleveDocConfig,
		rollback:       rollback,
		bindex:         bindex,
		partitions:     make(map[string]*BleveDestPartition),
		stats:          cbgt.NewPIndexStoreStats(),
		copyStats:      &CopyPartitionStats{},
		stopCh:         make(chan struct{}),
		removeCh:       make(chan struct{}),
	}

	bleveDest.batchReqChs = make([]chan *batchRequest, asyncBatchWorkerCount)

	bleveDest.startBatchWorkers()

	atomic.AddUint64(&TotBleveDestOpened, 1)

	return bleveDest
}

func (t *BleveDest) startBatchWorkers() {
	for i := 0; i < asyncBatchWorkerCount; i++ {
		t.batchReqChs[i] = make(chan *batchRequest, 1)
		go runBatchWorker(t.batchReqChs[i], t.stopCh, t.bindex, i)
	}
}

func (t *BleveDest) stopBatchWorkers() {
	t.m.Lock()
	ch := t.stopCh
	t.stopCh = make(chan struct{})
	t.m.Unlock()
	close(ch)
}

// ---------------------------------------------------------

var CurrentNodeDefsFetcher *NodeDefsFetcher

type NodeDefsFetcher struct {
	mgr *cbgt.Manager
}

func (ndf *NodeDefsFetcher) SetManager(mgr *cbgt.Manager) {
	ndf.mgr = mgr
}

func (ndf *NodeDefsFetcher) GetManager() *cbgt.Manager {
	return ndf.mgr
}

func (ndf *NodeDefsFetcher) Get() (*cbgt.NodeDefs, error) {
	if ndf.mgr != nil {
		return ndf.mgr.GetNodeDefs(cbgt.NODE_DEFS_WANTED, true)
	}
	return nil, fmt.Errorf("NodeDefsFetcher Get(): mgr is nil!")
}

// ---------------------------------------------------------

const bleveQueryHelp = `
<a href="https://docs.couchbase.com/server/7.6/fts/fts-supported-queries.html"
   target="_blank">
   query syntax help
</a>
`

func init() {
	cbgt.RegisterPIndexImplType("fulltext-index", &cbgt.PIndexImplType{
		Prepare:  PrepareIndexDef,
		Validate: ValidateBleve,
		OnDelete: OnDeleteIndex,

		New:       NewBlevePIndexImpl,
		NewEx:     NewBlevePIndexImplEx,
		Rollback:  RollbackBleve,
		Open:      OpenBlevePIndexImpl,
		OpenUsing: OpenBlevePIndexImplUsing,

		Count: CountBleve,
		Query: QueryBleve,

		Description: "general/fulltext-index " +
			" - a full text index powered by the bleve engine",
		StartSample:  NewBleveParams(),
		QuerySamples: BleveQuerySamples,
		QueryHelp:    bleveQueryHelp,
		InitRouter:   BleveInitRouter,
		DiagHandlers: []cbgt.DiagHandler{
			{Name: "/api/pindex-bleve", Handler: ftsHttp.NewListIndexesHandler(),
				HandlerFunc: nil},
		},
		MetaExtra: BleveMetaExtra,
		UI: map[string]string{
			"controllerInitName": "blevePIndexInitController",
			"controllerDoneName": "blevePIndexDoneController",
		},
		AnalyzeIndexDefUpdates: RestartOnIndexDefChanges,
		SubmitTaskRequest:      SubmitTaskRequest,
	})

}

func PrepareIndexDef(mgr *cbgt.Manager, indexDef *cbgt.IndexDef) (
	*cbgt.IndexDef, error) {
	if indexDef == nil {
		return nil, cbgt.NewBadRequestError("PrepareIndex, indexDef is nil")
	}

	if CurrentNodeDefsFetcher == nil {
		rv, err := LimitIndexDef(mgr, indexDef)
		if err != nil {
			return rv, cbgt.NewBadRequestError("%v", err)
		}
		return rv, nil
	}

	nodeDefs, err := CurrentNodeDefsFetcher.Get()
	if err != nil {
		return nil, cbgt.NewInternalServerError("PrepareIndex, nodeDefs unavailable: err: %v", err)
	}

	var collectionsSupported, s2SpatialSupported bool
	if isClusterCompatibleFor(FeatureGeoSpatialVersion) {
		s2SpatialSupported =
			cbgt.IsFeatureSupportedByCluster(FeatureGeoSpatial, nodeDefs)

		// check whether the spatial plugin usage is disabled for geo points.
		if v := mgr.GetOption("disableGeoPointSpatialPlugin"); v == "true" {
			s2SpatialSupported = false
		}

		// as its implicit since spatial is already on an advanced version.
		collectionsSupported = true
	} else if isClusterCompatibleFor(FeatureCollectionVersion) {
		collectionsSupported =
			cbgt.IsFeatureSupportedByCluster(FeatureCollections, nodeDefs)
	}

	if collectionsSupported {
		// Use "gocbcore" for DCP streaming if cluster is 7.0+
		indexDef.SourceType = cbgt.SOURCE_GOCBCORE
	}

	vectorSearchSupported := isClusterCompatibleFor(FeatureVectorSearchSupportVersion) &&
		cbgt.IsFeatureSupportedByCluster(featureVectorSearch, nodeDefs)
	var indexVectorPicture vectorPicture

	bp := NewBleveParams()
	if len(indexDef.Params) > 0 {
		b, err := bleveMappingUI.CleanseJSON([]byte(indexDef.Params))
		if err != nil {
			return nil, cbgt.NewBadRequestError("PrepareIndex, CleanseJSON err: %v", err)
		}

		err = json.Unmarshal(b, bp)
		if err != nil {
			if typeErr, ok := err.(*json.UnmarshalTypeError); ok {
				if typeErr.Type.String() == "map[string]json.RawMessage" {
					return nil, cbgt.NewBadRequestError("PrepareIndex,"+
						" JSON parse was expecting a string key/field-name"+
						" but instead saw a %s", typeErr.Value)
				}
			}
			return nil, cbgt.NewBadRequestError("bleve: Prepare, err: %v", err)
		}

		if indexType, ok := bp.Store["indexType"].(string); ok {
			if indexType == "scorch" {
				// If indexType were "scorch", the "kvStoreName" setting isn't
				// really applicable, so drop the setting.
				delete(bp.Store, "kvStoreName")

				// insert spatial config only if geopoint field exists in the mapping.
				if s2SpatialSupported && isGeoPointFieldInMapping(bp) {
					bp.Store["spatialPlugin"] = "s2"
				} else {
					delete(bp.Store, "spatialPlugin")
				}
			}
		}

		// figure out the scope/collection details from mappings
		// and perform the validation checks.
		if strings.HasPrefix(bp.DocConfig.Mode, ConfigModeCollPrefix) {
			if !collectionsSupported {
				return nil, cbgt.NewBadRequestError("PrepareIndex, collections not supported" +
					" across all nodes in the cluster")
			}

			if im, ok := bp.Mapping.(*mapping.IndexMappingImpl); ok {
				scope, err := validateScopeCollFromMappings(indexDef.SourceName,
					im, false)
				if err != nil {
					return nil, cbgt.NewBadRequestError("%v", err)
				}

				bucketName, scopeName := getKeyspaceFromScopedIndexName(indexDef.Name)
				if len(bucketName) > 0 && len(scopeName) > 0 {
					if !isClusterCompatibleFor(FeatureScopedIndexNamesVersion) {
						return nil, cbgt.NewBadRequestError("PrepareIndex, scoped indexes NOT" +
							" supported in mixed version cluster")
					}
					if bucketName != indexDef.SourceName || scopeName != scope.Name {
						return nil, cbgt.NewBadRequestError("PrepareIndex, validation of bucket" +
							" and/or scope names against index definition failed")
					}
				}
			}
		} else {
			bucketName, scopeName := getKeyspaceFromScopedIndexName(indexDef.Name)
			if len(bucketName) > 0 && len(scopeName) > 0 {
				if bucketName != indexDef.SourceName || scopeName != "_default" {
					return nil, cbgt.NewBadRequestError("PrepareIndex, changing a scoped index's" +
						" bucket/scope name is NOT allowed")
				}
			}
		}

		indexVectorPicture = vectorPictureFromIndexMapping(bp.Mapping)
		if indexVectorPicture.fields != noVectorFields && !vectorSearchSupported {
			// Vector indexing & search is NOT supported on this cluster
			// (lower version or mixed lower version)
			return nil, cbgt.NewBadRequestError("PrepareIndex, err: vector typed fields " +
				"not supported in this cluster")
		}

		if mappingContainsXAttrs(bp) {
			if !cbgt.IsFeatureSupportedByCluster(FeatureXattrs, nodeDefs) {
				// XAttrs is NOT supported on this cluster
				// (lower version or mixed lower version)
				return nil, cbgt.NewBadRequestError("PrepareIndex, err: xattr fields " +
					"and properties not supported in this cluster")
			}

			sourceParams := make(map[string]interface{})
			if len(indexDef.SourceParams) > 0 && indexDef.SourceParams != "null" {
				err = UnmarshalJSON([]byte(indexDef.SourceParams), &sourceParams)
				if err != nil {
					return nil, cbgt.NewBadRequestError("PrepareIndex, unable to unmarshal source params"+
						", err: %v", err)
				}
			}

			sourceParams["includeXAttrs"] = true

			updatedSourceParams, err := MarshalJSON(sourceParams)
			if err != nil {
				return nil, cbgt.NewBadRequestError("PrepareIndex, unable to marshal source params"+
					", err: %v", err)
			}

			indexDef.SourceParams = string(updatedSourceParams)
		}

		if indexVectorPicture.fields == vectorAndBase64Fields &&
			(len(featuresVectorBase64Dims4096) == 0 ||
				!cbgt.IsFeatureSupportedByCluster(featuresVectorBase64Dims4096, nodeDefs)) {
			return nil, cbgt.NewBadRequestError("PrepareIndex, err: vector_base64 typed fields " +
				"not supported in this cluster")
		}

		featureFlagToCheckForDims := featureFlagForDims(indexVectorPicture.maxDims)
		if len(featureFlagToCheckForDims) > 0 &&
			!cbgt.IsFeatureSupportedByCluster(featureFlagToCheckForDims, nodeDefs) {
			return nil, cbgt.NewBadRequestError(fmt.Sprintf("PrepareIndex, err: vector typed fields "+
				"with dims %v not supported in this cluster", indexVectorPicture.maxDims))
		}

		if indexVectorPicture.cosine &&
			(len(featureVectorCosineSimilarity) == 0 ||
				!cbgt.IsFeatureSupportedByCluster(featureVectorCosineSimilarity, nodeDefs)) {
			return nil, cbgt.NewBadRequestError("PrepareIndex, err: cosine similarity metric for vectors " +
				"not supported in this cluster")
		}
	}

	segmentVersionSupported := cbgt.IsFeatureSupportedByCluster(
		FeatureBlevePreferredSegmentVersion, nodeDefs)
	// if segment version is specified then perform the validations.
	if v, ok := bp.Store["segmentVersion"]; ok {
		if z, ok := v.(float64); ok {
			zv := int(z)
			if !segmentVersionSupported && zv >= blevePreferredZapVersion {
				// if the cluster isn't advanced enough then err out
				// on latest zap version request for new indexes.
				return nil, cbgt.NewBadRequestError("PrepareIndex, err: zap version %d isn't "+
					"supported in mixed version cluster", zv)
			}
			if zv > blevePreferredZapVersion || zv < bleveLegacyZapVersion {
				return nil, cbgt.NewBadRequestError("PrepareIndex, err: zap version %d isn't "+
					"supported", zv)
			}

			if indexVectorPicture.fields != noVectorFields && zv < bleveVectorZapVersion {
				// overrride segmentVersion to minimum version needed to support vector mappings
				versionToUse := bleveVectorZapVersion
				if segmentVersionSupported {
					versionToUse = blevePreferredZapVersion
				}
				bp.Store["segmentVersion"] = versionToUse
			}
		} else {
			return nil, cbgt.NewBadRequestError("PrepareIndex, err: segmentVersion %v "+
				"should be a numeric value", v)
		}
	} else {
		// if no zap version is specified then assume the preferred
		// zap version for newer indexes in a sufficiently advanced
		// cluster, else consider the default zap version.
		if segmentVersionSupported {
			bp.Store["segmentVersion"] = blevePreferredZapVersion
		} else if vectorSearchSupported {
			bp.Store["segmentVersion"] = bleveVectorZapVersion
		} else if collectionsSupported {
			bp.Store["segmentVersion"] = bleveCollectionsZapVersion
		} else {
			bp.Store["segmentVersion"] = bleveLegacyZapVersion
		}
	}

	updatedParams, err := json.Marshal(bp)
	if err != nil {
		return nil, cbgt.NewBadRequestError("PrepareIndex, Marshal err: %v", err)
	}
	indexDef.Params = string(updatedParams)

	if err := checkSourceCompatability(mgr, indexDef.SourceName); err != nil {
		return nil, cbgt.NewInternalServerError("PrepareIndex, err: %v", err)
	}

	rv, err := LimitIndexDef(mgr, indexDef)
	if err != nil {
		return rv, cbgt.NewInternalServerError("%v", err)
	}
	return rv, nil
}

func ValidateBleve(indexType, indexName, indexParams string) error {
	if len(indexParams) <= 0 {
		return nil
	}

	validateBleveIndexType := func(content interface{}) error {
		if CurrentNodeDefsFetcher == nil {
			return nil
		}

		nodeDefs, err := CurrentNodeDefsFetcher.Get()
		if err != nil {
			return cbgt.NewInternalServerError("ValidateIndex, nodeDefs unavailable: err: %v", err)
		}

		indexType := ""
		if entries, ok := content.(map[string]interface{}); ok {
			indexType = entries["indexType"].(string)
		}

		// Validate indexType
		if !cbgt.IsFeatureSupportedByCluster(featureIndexType+":"+indexType, nodeDefs) {
			return cbgt.NewBadRequestError("ValidateIndex, index validation failed:"+
				" indexType: %v not supported on all nodes in"+
				" cluster", indexType)
		}

		return nil
	}

	// Validate token filters in indexParams
	validateIndexParams := func() error {
		var iParams map[string]interface{}
		err := json.Unmarshal([]byte(indexParams), &iParams)
		if err != nil {
			// Ignore the JSON unmarshalling error, if in the case
			// indexParams isn't JSON.
			return nil
		}

		store, found := iParams["store"]
		if found {
			err = validateBleveIndexType(store)
			if err != nil {
				return err
			}
		}

		mapping, found := iParams["mapping"]
		if !found {
			// Entry for mapping not found
			return nil
		}

		analysis, found := mapping.(map[string]interface{})["analysis"]
		if !found {
			// No sub entry with the name analysis within mapping
			return nil
		}

		tokenfilters, found := analysis.(map[string]interface{})["token_filters"]
		if !found {
			// No entry for token_filters within mapping/analysis
			return nil
		}

		for _, val := range tokenfilters.(map[string]interface{}) {
			param := val.(map[string]interface{})
			switch param["type"] {
			case "edge_ngram", "length", "ngram", "shingle":
				if param["min"].(float64) > param["max"].(float64) {
					return cbgt.NewBadRequestError("bleve: token_filter validation failed"+
						" for %v => min(%v) > max(%v)", param["type"],
						param["min"], param["max"])
				}
			case "truncate_token":
				if param["length"].(float64) < 0 {
					return cbgt.NewBadRequestError("bleve: token_filter validation failed"+
						" for %v => length(%v) < 0", param["type"], param["length"])
				}
			default:
				break
			}
		}

		return nil
	}

	err := validateIndexParams()
	if err != nil {
		return cbgt.NewBadRequestError("ValidateIndex, err: %v", err)
	}

	b, err := bleveMappingUI.CleanseJSON([]byte(indexParams))
	if err != nil {
		return cbgt.NewBadRequestError("ValidateIndex, CleanseJSON err: %v", err)
	}

	bp := NewBleveParams()

	err = json.Unmarshal(b, bp)
	if err != nil {
		if typeErr, ok := err.(*json.UnmarshalTypeError); ok {
			if typeErr.Type.String() == "map[string]json.RawMessage" {
				return cbgt.NewBadRequestError("ValidateIndex, Params:"+
					" JSON parse was expecting a string key/field-name"+
					" but instead saw a %s", typeErr.Value)
			}
		}

		return cbgt.NewBadRequestError("ValidateIndex, Params err: %v", err)
	}

	// err out if there are no active type mapping.
	if im, ok := bp.Mapping.(*mapping.IndexMappingImpl); ok {
		var found bool
		for tp, dm := range im.TypeMapping {
			if !dm.Enabled || len(tp) == 0 {
				continue
			}
			found = true
			break
		}
		if !im.DefaultMapping.Enabled && !found {
			return cbgt.NewBadRequestError("ValidateIndex, Params: no valid type mappings found")
		}
	}

	err = bp.Mapping.Validate()
	if err != nil {
		return cbgt.NewBadRequestError("ValidateIndex, Mapping err: %v", err)
	}

	return nil
}

func OnDeleteIndex(indexDef *cbgt.IndexDef) {
	if indexDef == nil {
		return
	}

	// Reset focusStats of the index
	for k := range queryPaths {
		if indexQueryPathStats, exists := MapRESTPathStats[k]; exists {
			indexQueryPathStats.ResetFocusStats(indexDef.Name)
		}
	}

	// Reset gRPC focusStats of the index
	GrpcPathStats.ResetFocusStats(indexDef.Name)

	// Reset the metaFieldValCache
	metaFieldValCache.reset(indexDef.Name)

	// Reset QuerySupervisor's entry for last_access_time
	querySupervisor.deleteEntryForIndex(indexDef.Name)

	// Clear out source partition seqs cache for this index
	DropSourcePartitionSeqs(indexDef.SourceName, indexDef.SourceUUID)

	// Refresh the regulator stats. For eg, Deletion of a bucket leads
	// to index deletions, in which case the stale regulator stats
	// needs to be removed.
	RefreshRegulatorStats()

	// When deleting an index, remove the corresponding bucket from the manifest cache.
	// If other indexes still use the same bucket, the cache will be refreshed as needed.
	// Otherwise, unnecessary monitoring of the bucket can be avoided.
	removeBucketFromManifestCache(indexDef.SourceName)
}

func parseIndexParams(indexParams string) (
	bleveParams *BleveParams, kvConfig map[string]interface{},
	bleveIndexType string, kvStoreName string, err error) {
	var ip cbgt.IndexPrepParams
	err = json.Unmarshal([]byte(indexParams), &ip)
	if err != nil {
		return nil, nil, "", "",
			fmt.Errorf("bleve: new index, json marshal"+
				" err: %v", err)
	}

	indexParams = ip.Params
	bleveParams = NewBleveParams()
	if len(indexParams) > 0 {
		buf, err := bleveMappingUI.CleanseJSON([]byte(indexParams))
		if err != nil {
			return nil, nil, "", "",
				fmt.Errorf("bleve: cleanse params, err: %v", err)
		}

		err = json.Unmarshal(buf, bleveParams)
		if err != nil {
			return nil, nil, "", "",
				fmt.Errorf("bleve: parse params, err: %v", err)
		}
	}

	if strings.HasPrefix(bleveParams.DocConfig.Mode, ConfigModeCollPrefix) {
		if im, ok := bleveParams.Mapping.(*mapping.IndexMappingImpl); ok {
			scope, err := validateScopeCollFromMappings(ip.SourceName,
				im, false)
			if err != nil {
				return nil, nil, "", "", err
			}
			// if there are more than 1 collection then need to
			// insert $scope#$collection field into the mappings
			if multiCollection(scope.Collections) {
				err = enhanceMappingsWithCollMetaField(im.TypeMapping)
				if err != nil {
					return nil, nil, "", "", err
				}
				ipBytes, err := json.Marshal(bleveParams)
				if err != nil {
					return nil, nil, "", "",
						fmt.Errorf("bleve: new , json marshal,"+
							" err: %v", err)
				}
				indexParams = string(ipBytes)
			}

			bleveParams.DocConfig.CollPrefixLookup =
				initMetaFieldValCache(ip.IndexName, ip.SourceName, im, scope)
		}
	}

	kvConfig, bleveIndexType, kvStoreName = bleveRuntimeConfigMap(bleveParams)
	return bleveParams, kvConfig, bleveIndexType, kvStoreName, nil
}

func NewBlevePIndexImpl(indexType, indexParams, path string,
	rollback func()) (cbgt.PIndexImpl, cbgt.Dest, error) {
	bleveParams, kvConfig, bleveIndexType, kvStoreName, err :=
		parseIndexParams(indexParams)
	if err != nil {
		return nil, nil, err
	}

	if bleveIndexType == "upside_down" {
		return nil, nil, fmt.Errorf("bleve: new index, path: %s,"+
			" uses index type upside_down which is no longer supported", path)
	}

	bindex, err := bleve.NewUsing(path, bleveParams.Mapping,
		bleveIndexType, kvStoreName, kvConfig)
	if err != nil {
		return nil, nil, fmt.Errorf("bleve: new index, path: %s,"+
			" kvStoreName: %s, kvConfig: %#v, err: %s",
			path, kvStoreName, kvConfig, err)
	}

	pathMeta := path + string(os.PathSeparator) + "PINDEX_BLEVE_META"
	err = os.WriteFile(pathMeta, []byte(indexParams), 0600)
	if err != nil {
		return nil, nil, err
	}
	tmp := struct {
		SourceName string `json:"sourceName"`
	}{}

	err = json.Unmarshal([]byte(indexParams), &tmp)
	if err != nil {
		return nil, nil, fmt.Errorf("bleve: parse params: %v", err)
	}
	return bindex, &cbgt.DestForwarder{
		DestProvider: NewBleveDestEx(path, bindex, rollback, bleveParams.DocConfig,
			tmp.SourceName),
	}, nil
}

func RollbackBleve(indexType, indexParams, sourceParams, path string,
	mgr *cbgt.Manager, rollback func()) (cbgt.PIndexImpl, cbgt.Dest, error) {
	bleveParams, kvConfig, bleveIndexType, kvStoreName, err :=
		parseIndexParams(indexParams)
	if err != nil {
		return nil, nil, err
	}

	if bleveIndexType == "upside_down" {
		return nil, nil, fmt.Errorf("bleve: new index, path: %s,"+
			" uses index type upside_down which is no longer supported", path)
	}

	pindexName := cbgt.PIndexNameFromPath(path)

	pathMeta := path + string(os.PathSeparator) + "PINDEX_BLEVE_META"
	err = os.WriteFile(pathMeta, []byte(indexParams), 0600)
	if err != nil {
		return nil, nil, err
	}

	dest := newNoOpBleveDest(pindexName, path, bleveParams, rollback)
	destfwd := &cbgt.DestForwarder{DestProvider: dest}

	go func() {
		// Check if the pindex is closed before building it.
		if isClosed(dest.stopCh) {
			// Check if the pindex still exists.
			_, pindexes := mgr.CurrentMaps()
			pi, ok := pindexes[pindexName]
			if !ok {
				log.Printf("pindex_bleve: pindex: %s"+
					" no longer exists", pindexName)
				return
			}

			// Remove the pindex if it's still live.
			// No need to create a new bleve index if the underlying
			// pindex isn't live.
			_ = mgr.RemovePIndex(pi)
			return
		}

		bindex, err := bleve.NewUsing(path, bleveParams.Mapping,
			bleveIndexType, kvStoreName, kvConfig)
		if err != nil {
			log.Errorf("bleve: new index, path: %s,"+
				" kvStoreName: %s, kvConfig: %#v, err: %s",
				path, kvStoreName, kvConfig, err)

			// If there's an error creating the bleve index,
			// remove the directory and attempt to re-rollback.
			_ = os.RemoveAll(path)

			_, pindexes := mgr.CurrentMaps()
			pi, ok := pindexes[pindexName]
			if !ok {
				log.Printf("pindex_bleve: pindex: %s"+
					" no longer exists", pindexName)
				return
			}

			// Will be a full rollback since the path was removed.
			// As part of rollback, the pindex will be closed
			mgr.JanitorRollbackKick("rollback:"+pi.Name, pi)

			return
		}

		updateBleveIndex(pindexName, mgr, bindex, dest)
	}()

	return nil, destfwd, nil
}

// To be called ONLY when docConfig.Mode has "scope.collection" prefix
func initMetaFieldValCache(indexName, sourceName string,
	im *mapping.IndexMappingImpl, scope *Scope) map[uint32]*collMetaField {
	if im == nil {
		return nil
	}
	rv := make(map[uint32]*collMetaField, 1)
	// clean up any old data
	metaFieldValCache.reset(indexName)

	multiCollIndex := multiCollection(scope.Collections)
	for _, coll := range scope.Collections {
		cuid, err := strconv.ParseInt(coll.Uid, 16, 32)
		if err != nil {
			return nil
		}
		suid, err := strconv.ParseInt(scope.Uid, 16, 32)
		if err != nil {
			return nil
		}
		if cmf, ok := rv[uint32(cuid)]; !ok {
			rv[uint32(cuid)] = &collMetaField{
				scopeDotColl: scope.Name + "." + coll.Name,
				typeMappings: []string{coll.typeMapping},
				value:        encodeCollMetaFieldValue(suid, cuid),
			}
		} else {
			cmf.typeMappings = append(cmf.typeMappings, coll.typeMapping)
		}

		metaFieldValCache.setValue(indexName, scope.Name, suid,
			coll.Name, cuid, multiCollIndex)
	}
	return rv
}

func OpenBlevePIndexImpl(indexType, path string,
	rollback func()) (cbgt.PIndexImpl, cbgt.Dest, error) {
	return OpenBlevePIndexImplUsing(indexType, path, "", rollback)
}

func bleveRuntimeConfigMap(bleveParams *BleveParams) (map[string]interface{},
	string, string) {
	// check the indexType
	bleveIndexType, ok := bleveParams.Store["indexType"].(string)
	if !ok || bleveIndexType == "" {
		bleveIndexType = bleve.Config.DefaultIndexType
	}

	kvConfig := map[string]interface{}{
		"create_if_missing":        true,
		"error_if_exists":          true,
		"unsafe_batch":             true,
		"eventCallbackName":        "scorchEventCallbacks",
		"asyncErrorCallbackName":   "scorchAsyncErrorCallbacks",
		"numSnapshotsToKeep":       3,
		"rollbackSamplingInterval": "10m",
		"forceSegmentType":         "zap",
		"bolt_timeout":             "30s",
		// enable_concurrency will be added here.
	}
	for k, v := range bleveParams.Store {
		if k == "segmentVersion" {
			kvConfig["forceSegmentVersion"] = v
			continue
		}
		kvConfig[k] = v
	}

	if bleveIndexType != "scorch" {
		log.Warnf("unsupported index type: %s, overwriting to scorch", bleveIndexType)
		bleveIndexType = "scorch"
	}
	kvStoreName := "scorch"
	kvConfig["kvStoreName"] = "scorch"

	return kvConfig, bleveIndexType, kvStoreName
}

func OpenBlevePIndexImplUsing(indexType, path, indexParams string,
	rollback func()) (cbgt.PIndexImpl, cbgt.Dest, error) {
	buf := []byte(indexParams)
	var err error
	if len(buf) == 0 {
		buf, err = os.ReadFile(path +
			string(os.PathSeparator) + "PINDEX_BLEVE_META")
		if err != nil {
			return nil, nil, err
		}
	}

	bleveParams := NewBleveParams()
	if len(buf) > 0 {
		// It is possible that buf is empty when index params aren't set as
		// part of the index definition.
		buf, err = bleveMappingUI.CleanseJSON(buf)
		if err != nil {
			return nil, nil, fmt.Errorf("bleve: cleanse params, err: %v", err)
		}

		err = json.Unmarshal(buf, bleveParams)
		if err != nil {
			return nil, nil, fmt.Errorf("bleve: parse params: %v", err)
		}
	}

	if strings.HasPrefix(bleveParams.DocConfig.Mode, ConfigModeCollPrefix) {
		if am, ok := bleveParams.Mapping.(*mapping.IndexMappingImpl); ok {
			buf, err = os.ReadFile(path +
				string(os.PathSeparator) + "PINDEX_META")
			if err != nil {
				return nil, nil, err
			}
			tmp := struct {
				SourceName string `json:"sourceName"`
				IndexName  string `json:"indexName"`
			}{}

			err = json.Unmarshal(buf, &tmp)
			if err != nil {
				return nil, nil, fmt.Errorf("bleve: parse params: %v", err)
			}
			scope, err := validateScopeCollFromMappings(tmp.SourceName,
				am, false)
			if err != nil {
				return nil, nil, fmt.Errorf("bleve: validate scope/collection: %v", err)
			}
			// populate the collection meta field look up cache.
			bleveParams.DocConfig.CollPrefixLookup =
				initMetaFieldValCache(tmp.IndexName, tmp.SourceName, am, scope)
		}
	}

	kvConfig, _, _ := bleveRuntimeConfigMap(bleveParams)
	// TODO: boltdb sometimes locks on Open(), so need to investigate,
	// where perhaps there was a previous missing or race-y Close().
	startTime := time.Now()
	log.Printf("bleve: start open using: %s", path)
	bindex, err := bleve.OpenUsing(path, kvConfig)
	if err != nil {
		return nil, nil, err
	}
	log.Printf("bleve: finished open using: %s took: %s", path, time.Since(startTime).String())

	buf, err = os.ReadFile(path +
		string(os.PathSeparator) + "PINDEX_META")
	if err != nil {
		return nil, nil, err
	}
	tmp := struct {
		SourceName string `json:"sourceName"`
		IndexName  string `json:"indexName"`
	}{}

	err = json.Unmarshal(buf, &tmp)
	if err != nil {
		return nil, nil, fmt.Errorf("bleve: parse params: %v", err)
	}

	return bindex, &cbgt.DestForwarder{
		DestProvider: NewBleveDestEx(path, bindex, rollback, bleveParams.DocConfig,
			tmp.SourceName),
	}, nil
}

// ---------------------------------------------------------------

func CountBleve(mgr *cbgt.Manager, indexName, indexUUID string) (
	uint64, error) {
	alias, _, _, err := bleveIndexAlias(mgr, indexName, indexUUID, false, nil, nil,
		false, nil, "", addIndexClients)
	if err != nil {
		if _, ok := err.(*cbgt.ErrorLocalPIndexHealth); !ok {
			return 0, fmt.Errorf("bleve: CountBleve indexAlias error,"+
				" indexName: %s, indexUUID: %s, err: %v", indexName, indexUUID, err)
		}
	}
	return alias.DocCount()
}

func ValidateConsistencyParams(c *cbgt.ConsistencyParams) error {
	switch c.Level {
	case "":
		return nil
	case "at_plus":
		return nil
	}
	return fmt.Errorf("unsupported consistencyLevel: %s", c.Level)
}

// SubmitTaskRequest helps requesting for asynchronous tasks on
// indexes like force merge, cancel merge etc.
func SubmitTaskRequest(mgr *cbgt.Manager, indexName, indexUUID string,
	requestBody []byte) (*cbgt.TaskRequestStatus, error) {
	var reqMap map[string]interface{}
	err := UnmarshalJSON(requestBody, &reqMap)
	if err != nil {
		return nil, fmt.Errorf("bleve: SubmitTaskRequest"+
			" parsing request, err: %v", err)
	}
	var onlyPIndexes map[string]bool
	var targetPIndexes []string
	if vals, ok := reqMap["partitionNames"].([]interface{}); ok {
		for _, v := range vals {
			targetPIndexes = append(targetPIndexes, v.(string))
		}
		onlyPIndexes = cbgt.StringsToMap(targetPIndexes)
	}

	var op string
	var ok bool
	if op, ok = reqMap["op"].(string); !ok ||
		(op != "merge" && op != "cancel" && op != "get") {
		return nil, fmt.Errorf("bleve: SubmitTaskRequest"+
			" unsupported task type: %s", op)
	}

	var uuid string
	if uuid, ok = reqMap["uuid"].(string); !ok && op == "cancel" {
		return nil, fmt.Errorf("bleve: SubmitTaskRequest" +
			" missing task uuid from cancel request")
	}

	// uuid for tracking the task for cancellations or potential
	// progress monitoring etc
	if uuid == "" && op != "get" {
		uuid = cbgt.NewUUID()
		reqMap["uuid"] = uuid
		requestBody, err = json.Marshal(reqMap)
		if err != nil {
			return nil, err
		}
	}

	alias, _, _, err1 := bleveIndexAlias(mgr, indexName,
		indexUUID, true, nil, nil, true,
		onlyPIndexes, "", addIndexClients)
	if err1 != nil {
		if _, ok := err1.(*cbgt.ErrorLocalPIndexHealth); !ok {
			return nil, err1
		}
	}

	if bic, ok := alias.(BleveIndexCollector); ok {
		return cbgt.ScatterTaskRequest(requestBody, indexPartitions(bic))
	}

	return nil, fmt.Errorf("bleve: no BleveIndexCollector implementation found")
}

func indexPartitions(bic BleveIndexCollector) (handlers []cbgt.TaskRequestHandler) {
	gatherHandlers := func(i bleve.Index) {
		if h, ok := i.(cbgt.TaskRequestHandler); ok {
			handlers = append(handlers, h)
		}
	}
	bic.VisitIndexes(gatherHandlers)
	return handlers
}

// ---------------------------------------------------------
// Query Error Counters, use atomic/Sync* for access
var (
	// totQueryRejectOnNotEnoughQuota tracks the number of rejected
	// search requests on hitting the memory threshold for query
	totQueryRejectOnNotEnoughQuota uint64

	// bleve.IndexAlias's SearchInContext() failed
	totQuerySearchInContextErr uint64
	// requested query could not not be parsed, bad syntax
	totQueryBadRequestErr uint64
	// requested query failed validation against index
	totQueryValidationErr uint64
	// requested query could not meet consistency requirements
	totQueryConsistencyErr uint64
	// (Query.From + Query.Size) exceeded bleveMaxResultWindow
	totQueryMaxResultWindowExceededErr uint64
	// requested query could not be executed by some pindex(es)
	totQueryPartialResultsErr uint64
)

// number of search requests that carry a kNN request within them
// (only recorded at coordinator node - where req was received)
var numKNNSearchRequests uint64

// QueryPIndexes defines the part of the JSON query request that
// allows the client to specify which pindexes the server should
// consider during query processing.
type QueryPIndexes struct {
	// An empty or nil PIndexNames means the query should use all
	// the pindexes of the index.
	PIndexNames []string `json:"pindexNames,omitempty"`
}

func fireQueryEvent(depth int, kind QueryEventKind, dur time.Duration, size uint64) error {
	if RegistryQueryEventCallback != nil {
		return RegistryQueryEventCallback(depth, QueryEvent{Kind: kind, Duration: dur}, size)
	}
	return nil
}

// forcing the searcher creation to app_herder checks so that
// if there is high memory usage we don't do an expensive searcher creation
func bleveCtxSearcherStartCallback(size uint64) error {
	return fireQueryEvent(0, EventQueryStart, 0, size)
}

func bleveCtxSearcherEndCallback(size uint64) error {
	return fireQueryEvent(0, EventQueryEnd, 0, size)
}

func bleveCtxQueryStartCallback(size uint64) error {
	return fireQueryEvent(1, EventQueryStart, 0, size)
}

func bleveCtxQueryEndCallback(size uint64) error {
	return fireQueryEvent(1, EventQueryEnd, 0, size)
}

func QueryBleve(mgr *cbgt.Manager, indexName, indexUUID string,
	req []byte, res io.Writer) error {
	// Throttle if query has knn
	if GetKNNThrottleLimit() > 0 && QueryHasKNN(req) {
		err := fireQueryEvent(0, EventKNNQueryStart, 0, 0)
		if err != nil {
			return err
		}
		defer fireQueryEvent(0, EventKNNQueryEnd, 0, 0)
	}
	// phase 0 - parsing/validating query
	// could return err 400
	queryCtlParams := cbgt.QueryCtlParams{
		Ctl: cbgt.QueryCtl{
			Timeout: cbgt.QUERY_CTL_DEFAULT_TIMEOUT_MS,
		},
	}

	err := UnmarshalJSON(req, &queryCtlParams)
	if err != nil {
		atomic.AddUint64(&totQueryBadRequestErr, 1)
		return fmt.Errorf("bleve: QueryBleve"+
			" parsing queryCtlParams, err: %v", err)
	}

	queryPIndexes := QueryPIndexes{}
	err = UnmarshalJSON(req, &queryPIndexes)
	if err != nil {
		atomic.AddUint64(&totQueryBadRequestErr, 1)
		return fmt.Errorf("bleve: QueryBleve"+
			" parsing queryPIndexes, err: %v", err)
	}

	var sr *SearchRequest
	err = UnmarshalJSON(req, &sr)
	if err != nil {
		atomic.AddUint64(&totQueryBadRequestErr, 1)
		return fmt.Errorf("bleve: QueryBleve"+
			" parsing searchRequest, err: %v", err)
	}
	searchRequest, err := sr.ConvertToBleveSearchRequest()
	if err != nil {
		atomic.AddUint64(&totQueryBadRequestErr, 1)
		return fmt.Errorf("bleve: QueryBleve"+
			" parsing searchRequest, err: %v", err)
	}

	// Validate from queryCtlParams
	if queryCtlParams.Ctl.Validate {
		if err = validateSearchRequestAgainstIndex(mgr, indexName, indexUUID,
			sr.Collections, searchRequest); err != nil {
			atomic.AddUint64(&totQueryValidationErr, 1)
			return fmt.Errorf("bleve: QueryBleve"+
				" query validation failed against index, err: %v", err)
		}
	}

	if sr.KNN != nil {
		atomic.AddUint64(&numKNNSearchRequests, 1)
	}

	var undecoratedQuery query.Query
	// pre process the query with collections if applicable.
	if strings.Compare(cbgt.CfgAppVersion, "7.0.0") >= 0 {
		undecoratedQuery, searchRequest.Query = sr.decorateQuery(indexName,
			searchRequest.Query, nil)
	}

	if queryCtlParams.Ctl.Consistency != nil {
		err = ValidateConsistencyParams(queryCtlParams.Ctl.Consistency)
		if err != nil {
			atomic.AddUint64(&totQueryConsistencyErr, 1)
			return fmt.Errorf("bleve: QueryBleve"+
				" validating consistency, err: %v", err)
		}
	}

	if v := mgr.GetOption("bleveMaxResultWindow"); len(v) > 0 {
		var bleveMaxResultWindow int
		bleveMaxResultWindow, err = strconv.Atoi(v)
		if err != nil {
			return fmt.Errorf("bleve: QueryBleve atoi: %v, err: %v", v, err)
		}

		if (searchRequest.From+searchRequest.Size > bleveMaxResultWindow) ||
			(searchRequest.Size > bleveMaxResultWindow &&
				(searchRequest.SearchAfter != nil || searchRequest.SearchBefore != nil)) {
			atomic.AddUint64(&totQueryMaxResultWindowExceededErr, 1)
			return fmt.Errorf("bleve: bleveMaxResultWindow exceeded,"+
				" from: %d, size: %d, bleveMaxResultWindow: %d",
				searchRequest.From, searchRequest.Size, bleveMaxResultWindow)
		}
	}

	// phase 1 - set up timeouts, wait for local consistency reqiurements
	// to be satisfied, could return err 412

	// create a context with the appropriate timeout
	ctx, cancel, cancelCh := setupContextAndCancelCh(queryCtlParams, nil)
	// defer a call to cancel, this ensures that goroutine from
	// setupContextAndCancelCh always exits
	defer cancel()

	var onlyPIndexes map[string]bool
	if len(queryPIndexes.PIndexNames) > 0 {
		onlyPIndexes = cbgt.StringsToMap(queryPIndexes.PIndexNames)
	}

	alias, remoteClients, numPIndexes, err1 := bleveIndexAlias(mgr, indexName,
		indexUUID, true, queryCtlParams.Ctl.Consistency, cancelCh, true,
		onlyPIndexes, queryCtlParams.Ctl.PartitionSelection, getRemoteClients(mgr))
	if err1 != nil {
		if _, ok := err1.(*cbgt.ErrorLocalPIndexHealth); !ok {
			return err1
		}
	}

	// estimate memory needed for merging search results from all
	// the pindexes
	mergeEstimate := uint64(numPIndexes) * bleve.MemoryNeededForSearchResult(searchRequest)
	err = fireQueryEvent(0, EventQueryStart, 0, mergeEstimate)
	if err != nil {
		atomic.AddUint64(&totQueryRejectOnNotEnoughQuota, 1)
		return err
	}

	defer fireQueryEvent(0, EventQueryEnd, 0, mergeEstimate)

	// set query start/end callbacks
	ctx = context.WithValue(ctx, bleve.SearchQueryStartCallbackKey,
		bleve.SearchQueryStartCallbackFn(bleveCtxQueryStartCallback))
	ctx = context.WithValue(ctx, bleve.SearchQueryEndCallbackKey,
		bleve.SearchQueryEndCallbackFn(bleveCtxQueryEndCallback))

	// set the callbacks for searcher creation
	ctx = context.WithValue(ctx, search.SearcherStartCallbackKey,
		search.SearcherStartCallbackFn(bleveCtxSearcherStartCallback))
	ctx = context.WithValue(ctx, search.SearcherEndCallbackKey,
		search.SearcherEndCallbackFn(bleveCtxSearcherEndCallback))

	// register with the QuerySupervisor
	id := querySupervisor.AddEntry(&QuerySupervisorContext{
		Query:     searchRequest.Query,
		Cancel:    cancel,
		Size:      searchRequest.Size,
		From:      searchRequest.From,
		Timeout:   queryCtlParams.Ctl.Timeout,
		IndexName: indexName,
	})

	defer querySupervisor.DeleteEntry(id)

	searchResult, err := alias.SearchInContext(ctx, searchRequest)
	if err != nil {
		atomic.AddUint64(&totQuerySearchInContextErr, 1)
	}
	if searchResult != nil {
		// if the query decoration happens for collection targeted or docID
		// queries for multi collection indexes, then restore the original
		// user query in the search response, if the search request was echo'd
		// back in the search result.
		// Note: searchResult.Request will be non nil only when searchRequest.Explain is true
		// and its a bleve level setting
		if undecoratedQuery != nil && searchResult.Request != nil {
			searchResult.Request.Query = undecoratedQuery
		}
		err = processSearchResult(&queryCtlParams, indexName, searchResult,
			remoteClients, err, err1)

		if searchResult.Status != nil &&
			len(searchResult.Status.Errors) > 0 &&
			queryCtlParams.Ctl.Consistency != nil &&
			queryCtlParams.Ctl.Consistency.Results == "complete" {
			atomic.AddUint64(&totQueryPartialResultsErr, 1)

			// complete results expected, do not propagate partial results
			return fmt.Errorf("bleve: results weren't retrieved from some"+
				" index partitions: %d", len(searchResult.Status.Errors))
		}

		if searchResult.Status != nil &&
			searchResult.Status.Failed == searchResult.Status.Total {

			// query rejected by all pindexes
			isRejected := true
			for _, e := range searchResult.Status.Errors {
				if !strings.Contains(e.Error(), rest.ErrorQueryReqRejected.Error()) {
					isRejected = false
					break
				}
			}

			if isRejected {
				return rest.ErrorQueryReqRejected
			}
		}

		mustEncode(res, searchResult)

		// update return error status to indicate any errors within the
		// search result that was already propagated as response.
		if searchResult.Status != nil && len(searchResult.Status.Errors) > 0 {
			err = rest.ErrorAlreadyPropagated
		}
	}

	return err
}

func processSearchResult(queryCtlParams *cbgt.QueryCtlParams, indexName string,
	searchResult *bleve.SearchResult, remoteClients []RemoteClient,
	searchErr, aliasErr error) error {
	if searchResult != nil {
		if len(searchResult.Hits) > 0 {
			// if this is a multi collection index, then strip the collection UID
			// from the hit ID and fill the details of source collection
			if sdm, multiCollIndex :=
				metaFieldValCache.getSourceDetailsMap(indexName); multiCollIndex {
				for _, hit := range searchResult.Hits {
					if _, exists := hit.Fields["_$c"]; exists {
						// collection name has already been retrieved for this hit;
						// on the non-coordinating node (scatter-gather)
						continue
					}
					idBytes := []byte(hit.ID)
					cuid := binary.LittleEndian.Uint32(idBytes[:4])
					if collName, ok := sdm.collUIDNameMap[cuid]; ok {
						hit.ID = string(idBytes[4:])
						if hit.Fields == nil {
							hit.Fields = make(map[string]interface{})
						}
						hit.Fields["_$c"] = collName
					}
				}
			}
		}

		// check to see if any of the remote searches returned anything
		// other than 0, 200, 412, 429, these are returned to the user as
		// error status 400, and appear as phase 0 errors detected late.
		// 0 means we never heard anything back, and that is dealt with
		// in the following section
		for _, remoteClient := range remoteClients {
			lastStatus, lastErrBody := remoteClient.GetLast()
			if lastStatus == http.StatusTooManyRequests {
				log.Printf("bleve: remoteClient: %s query reject, statusCode: %d,"+
					" err: %v", remoteClient.GetHostPort(), lastStatus, searchErr)
				continue
			}
			if lastStatus != http.StatusOK &&
				lastStatus != http.StatusPreconditionFailed &&
				lastStatus != 0 {
				return fmt.Errorf("bleve: QueryBleve remote client"+
					" returned status: %d body: %s", lastStatus, lastErrBody)
			}
		}
		// now see if any of the remote searches returned 412; these should be
		// collated into a single 412 response at this level and will
		// be presented as phase 1 errors detected late
		remoteConsistencyWaitError := cbgt.ErrorConsistencyWait{
			Status:       "remote consistency error",
			StartEndSeqs: make(map[string][]uint64),
		}
		numRemoteSilent := 0
		for _, remoteClient := range remoteClients {
			lastStatus, lastErrBody := remoteClient.GetLast()
			if lastStatus == 0 {
				numRemoteSilent++
			}
			if lastStatus == http.StatusPreconditionFailed {
				var remoteConsistencyErr = struct {
					StartEndSeqs map[string][]uint64 `json:"startEndSeqs"`
				}{}
				err := UnmarshalJSON(lastErrBody, &remoteConsistencyErr)
				if err == nil {
					for k, v := range remoteConsistencyErr.StartEndSeqs {
						remoteConsistencyWaitError.StartEndSeqs[k] = v
					}
				}
			}
		}
		// if we had any explicitly returned consistency errors, return those
		if len(remoteConsistencyWaitError.StartEndSeqs) > 0 {
			atomic.AddUint64(&totQueryConsistencyErr, 1)
			return &remoteConsistencyWaitError
		}

		// we had *some* consistency requirements, but we never heard back
		// from some of the remote pindexes; just punt for now and return
		// a mostly empty 412 indicating we aren't sure
		if queryCtlParams.Ctl.Consistency != nil &&
			len(queryCtlParams.Ctl.Consistency.Vectors) > 0 &&
			numRemoteSilent > 0 {
			atomic.AddUint64(&totQueryConsistencyErr, 1)
			return &remoteConsistencyWaitError
		}

		if aliasErr != nil {
			if err2, ok := aliasErr.(*cbgt.ErrorLocalPIndexHealth); ok && len(err2.IndexErrMap) > 0 {
				// populate the searchResuls with the details of
				// pindexes not searched/covered in this query.
				if searchResult.Status.Errors == nil {
					searchResult.Status.Errors = make(map[string]error)
				}
				for pi, e := range err2.IndexErrMap {
					searchResult.Status.Errors[pi] = e
					searchResult.Status.Failed++
					searchResult.Status.Total++
				}
			}
		}
	}
	return nil
}

// ---------------------------------------------------------

func (t *BleveDest) resetBIndex(bindex bleve.Index) {
	t.m.Lock()
	t.bindex = bindex
	t.m.Unlock()
}

func (t *BleveDest) Dest(partition string) (cbgt.Dest, error) {
	t.m.Lock()
	d, err := t.getPartitionLOCKED(partition)
	t.m.Unlock()
	return d, err
}

func (t *BleveDest) getPartitionLOCKED(partition string) (
	*BleveDestPartition, error) {
	if t.bindex == nil {
		return nil, fmt.Errorf("bleve: BleveDest already closed")
	}

	bdp, exists := t.partitions[partition]
	if !exists || bdp == nil {
		bdp = &BleveDestPartition{
			bdest:           t,
			bindex:          t.bindex,
			partition:       partition,
			partitionBytes:  []byte(partition),
			partitionOpaque: []byte("o:" + partition),
			batch:           t.bindex.NewBatch(),
			cwrQueue:        cbgt.CwrQueue{},
		}
		heap.Init(&bdp.cwrQueue)

		t.partitions[partition] = bdp
	}

	return bdp, nil
}

// ---------------------------------------------------------

func (t *BleveDest) Close(remove bool) error {
	t.m.Lock()
	err := t.closeLOCKED(remove)
	t.m.Unlock()
	return err
}

func (t *BleveDest) removeLOCKED() {
	if _, ok := t.bindex.(*noopBleveIndex); ok {
		if !isClosed(t.removeCh) {
			close(t.removeCh)
		}
	} else {
		os.RemoveAll(t.path)
	}
}

func (t *BleveDest) closeLOCKED(remove bool) error {
	if t.bindex == nil {
		return nil // Already closed.
	}

	atomic.AddUint64(&TotBleveDestClosed, 1)

	close(t.stopCh)

	partitions := t.partitions
	t.partitions = make(map[string]*BleveDestPartition)

	t.bindex.Close()
	if remove {
		t.removeLOCKED()
	}
	t.bindex = nil

	go func() {
		// Cancel/error any consistency wait requests.
		err := fmt.Errorf("bleve: closeLOCKED")

		for _, bdp := range partitions {
			bdp.m.Lock()
			for _, cwr := range bdp.cwrQueue {
				cwr.DoneCh <- err
				close(cwr.DoneCh)
			}
			bdp.m.Unlock()
		}
	}()

	return nil
}

// ---------------------------------------------------------

func (t *BleveDest) ConsistencyWait(partition, partitionUUID string,
	consistencyLevel string,
	consistencySeq uint64,
	cancelCh <-chan bool) error {
	if consistencyLevel == "" {
		return nil
	}
	if consistencyLevel != "at_plus" {
		return fmt.Errorf("bleve: unsupported consistencyLevel: %s",
			consistencyLevel)
	}

	cwr := &cbgt.ConsistencyWaitReq{
		PartitionUUID:    partitionUUID,
		ConsistencyLevel: consistencyLevel,
		ConsistencySeq:   consistencySeq,
		CancelCh:         cancelCh,
		DoneCh:           make(chan error, 1),
	}

	t.m.Lock()

	bdp, err := t.getPartitionLOCKED(partition)
	if err != nil {
		t.m.Unlock()
		return err
	}

	uuid, _ := bdp.lastUUID.Load().(string)
	seq := atomic.LoadUint64(&bdp.seqMaxBatch)

	bdp.m.Lock()
	if cwr.PartitionUUID != "" && cwr.PartitionUUID != uuid {
		cwr.DoneCh <- fmt.Errorf("bleve: pindex_consistency"+
			" mismatched partition, uuid: %s, cwr: %#v", uuid, cwr)
		close(cwr.DoneCh)
	} else if cwr.ConsistencySeq > seq {
		heap.Push(&bdp.cwrQueue, cwr)
	} else {
		close(cwr.DoneCh)
	}
	bdp.m.Unlock()

	t.m.Unlock()

	return cbgt.ConsistencyWaitDone(partition, cancelCh, cwr.DoneCh,
		func() uint64 {
			return atomic.LoadUint64(&bdp.seqMaxBatch)
		})
}

// ---------------------------------------------------------

func (t *BleveDest) Count(pindex *cbgt.PIndex, cancelCh <-chan bool) (
	uint64, error) {
	t.m.RLock()
	defer t.m.RUnlock()

	if t.bindex == nil {
		return 0, fmt.Errorf("bleve: Count, bindex already closed")
	}

	return t.bindex.DocCount()
}

// ---------------------------------------------------------

func (t *BleveDest) Query(pindex *cbgt.PIndex, req []byte, res io.Writer,
	parentCancelCh <-chan bool) error {
	// phase 0 - parsing/validating query
	// could return err 400
	queryCtlParams := cbgt.QueryCtlParams{
		Ctl: cbgt.QueryCtl{
			Timeout: cbgt.QUERY_CTL_DEFAULT_TIMEOUT_MS,
		},
	}
	err := UnmarshalJSON(req, &queryCtlParams)
	if err != nil {
		return fmt.Errorf("bleve: BleveDest.Query"+
			" parsing queryCtlParams, err: %v", err)
	}

	var sr *SearchRequest
	err = UnmarshalJSON(req, &sr)
	if err != nil {
		return fmt.Errorf("bleve: BleveDest.Query"+
			" parsing searchRequest, err: %v", err)
	}
	searchRequest, err := sr.ConvertToBleveSearchRequest()
	if err != nil {
		return fmt.Errorf("bleve: BleveDest.Query"+
			" parsing searchRequest, err: %v", err)
	}

	// phase 1 - set up timeouts, wait to satisfy consistency requirements
	// could return err 412

	// create a context with the appropriate timeout
	ctx, cancel, cancelCh := setupContextAndCancelCh(queryCtlParams, parentCancelCh)
	// defer a call to cancel, this ensures that goroutine from
	// setupContextAndCancelCh always exits
	defer cancel()

	err = cbgt.ConsistencyWaitPIndex(pindex, t,
		queryCtlParams.Ctl.Consistency, cancelCh)
	if err != nil {
		if _, ok := err.(*cbgt.ErrorConsistencyWait); !ok {
			// not a consistency wait error
			// check to see if context error
			if ctx.Err() != nil {
				// return this as search response error
				sendSearchResultErr(searchRequest, res, []string{pindex.Name}, ctx.Err())
				return nil
			}
		}
		// some other error occurred return this as 400
		return err
	}

	// phase 2 - execute query
	// always 200, possibly with errors inside status
	t.m.RLock()
	bindex := t.bindex
	t.m.RUnlock()

	if bindex == nil {
		err = fmt.Errorf("bleve: Query, bindex already closed")
		sendSearchResultErr(searchRequest, res, []string{pindex.Name}, err)
		return nil
	}

	// register with the QuerySupervisor
	id := querySupervisor.AddEntry(&QuerySupervisorContext{
		Query:   searchRequest.Query,
		Cancel:  cancel,
		Size:    searchRequest.Size,
		From:    searchRequest.From,
		Timeout: queryCtlParams.Ctl.Timeout,
	})

	defer querySupervisor.DeleteEntry(id)

	searchResponse, err := bindex.SearchInContext(ctx, searchRequest)
	if err != nil {
		sendSearchResultErr(searchRequest, res, []string{pindex.Name}, err)
		return nil
	}

	rest.MustEncode(res, searchResponse)
	return nil
}

// ---------------------------------------------------------

func sendSearchResultErr(req *bleve.SearchRequest, res io.Writer,
	pindexNames []string, err error) {
	rest.MustEncode(res, makeSearchResultErr(req, pindexNames, err))
}

func makeSearchResultErr(req *bleve.SearchRequest,
	pindexNames []string, err error) *bleve.SearchResult {
	rv := &bleve.SearchResult{
		Request: req,
		Status: &bleve.SearchStatus{
			Total:      len(pindexNames),
			Failed:     len(pindexNames),
			Successful: 0,
			Errors:     make(map[string]error),
		},
	}
	for _, pindexName := range pindexNames {
		rv.Status.Errors[pindexName] = err
	}
	return rv
}

// ---------------------------------------------------------

func setupContextAndCancelCh(queryCtlParams cbgt.QueryCtlParams,
	parentCancelCh <-chan bool) (ctx context.Context, cancel context.CancelFunc,
	cancelChRv <-chan bool) {
	if queryCtlParams.Ctl.Timeout > 0 {
		ctx, cancel = context.WithTimeout(context.Background(),
			time.Duration(queryCtlParams.Ctl.Timeout)*time.Millisecond)
	} else {
		ctx, cancel = context.WithCancel(context.Background())
	}
	// now create cbgt compatible cancel channel
	cancelCh := make(chan bool, 1)
	cancelChRv = cancelCh
	// spawn a goroutine to close the cancelCh when either:
	//   - the context is Done()
	// or
	//   - the parentCancelCh is closed
	go func() {
		select {
		case <-parentCancelCh:
			close(cancelCh)
		case <-ctx.Done():
			close(cancelCh)
		}
	}()
	return
}

// ---------------------------------------------------------

func (t *BleveDest) AddError(op, partition string,
	key []byte, seq uint64, val []byte, err error) {
	// avoid log flooding from non-json inputs in bucket
	if !strings.HasPrefix(op, "json") {
		log.Printf("bleve: %s, partition: %s, key: %q, seq: %d,"+
			" err: %v", op, partition, log.Tag(log.UserData, key), seq, err)
	}

	e := struct {
		Time      string
		Op        string
		Partition string
		Key       string
		Seq       uint64
		Err       string
	}{
		Time:      time.Now().Format(time.RFC3339Nano),
		Op:        op,
		Partition: partition,
		Key:       string(key),
		Seq:       seq,
		Err:       fmt.Sprintf("%v", err),
	}

	buf, err := json.Marshal(&e)
	if err == nil {
		t.stats.AddError(string(buf))
	}
}

// ---------------------------------------------------------

type JSONStatsWriter interface {
	WriteJSON(w io.Writer) error
}

var prefixPIndexStoreStats = []byte(`{"pindexStoreStats":`)

var prefixCopyPartitionStats = []byte(`,"copyPartitionStats":`)

var prefixHerderStats = []byte(`,"herderStats":`)

func writeHerderStatsJSON(w io.Writer) {
	t := atomic.LoadUint64(&TotHerderWaitingIn)
	stats := fmt.Sprintf(`{"TotWaitingIn":%d`, t)
	t = atomic.LoadUint64(&TotHerderWaitingOut)
	stats += fmt.Sprintf(`,"TotWaitingOut":%d`, t)

	w.Write([]byte(stats))
	w.Write(cbgt.JsonCloseBrace)
}

func (t *BleveDest) Stats(w io.Writer) (err error) {
	var vbstats, verbose bool
	var indexDef *cbgt.IndexDef
	var sourcePartitionSeqs map[string]cbgt.UUIDSeq
	if w, ok := w.(rest.PartitionStatsWriter); ok {
		vbstats = w.VbStats()
		verbose = w.Verbose()
		indexDef = w.IndexDef()
		sourcePartitionSeqs = w.SourcePartitionSeqs()
	}

	// exit early if all details are disabled.
	if !verbose && !vbstats {
		_, err = w.Write([]byte(`{}`))
		return err
	}

	// if verbose stats is requested then send most of the index stats.
	if verbose {
		_, err = w.Write(prefixPIndexStoreStats)
		if err != nil {
			return
		}

		t.stats.WriteJSON(w)

		var statsMap map[string]interface{}
		var docCount uint64
		t.m.RLock()
		if t.bindex != nil {
			statsMap = t.bindex.StatsMap()
			docCount, err = t.bindex.DocCount()
		}
		t.m.RUnlock()

		if err != nil {
			return
		}

		if statsMap != nil {
			_, err = w.Write([]byte(`,"bleveIndexStats":`))
			if err != nil {
				return
			}
			var idxStatsJSON []byte
			idxStatsJSON, err = MarshalJSON(statsMap)
			if err != nil {
				log.Errorf("json failed to marshal was: %#v", statsMap)
				return
			}
			_, err = w.Write(idxStatsJSON)
			if err != nil {
				return
			}
		}

		_, err = w.Write([]byte(`,"basic":{"DocCount":` + strconv.FormatUint(docCount, 10)))
		if err != nil {
			return
		}
		_, err = w.Write(cbgt.JsonCloseBrace)
		if err != nil {
			return
		}

		// skip the vbucket stats if vbstats is not requested.
		if !vbstats {
			_, _ = w.Write(cbgt.JsonCloseBrace)
			return
		}

		_, err = w.Write([]byte(`,`))
		if err != nil {
			return
		}
	} else {
		// in case only vbstats are requested.
		_, err = w.Write(cbgt.JsonOpenBrace)
		if err != nil {
			return
		}
	}

	_, err = w.Write([]byte(`"partitions":{`))
	if err != nil {
		return
	}

	// obtain scope, collection names
	scope, collections, _ := GetScopeCollectionsFromIndexDef(indexDef)

	t.m.RLock()
	partitionSeqs := make([][]byte, len(t.partitions))
	i := 0
	for partition, bdp := range t.partitions {
		bdpSeqMax := atomic.LoadUint64(&bdp.seqMax)
		bdpSeqMaxBatch := atomic.LoadUint64(&bdp.seqMaxBatch)
		bdpLastUUID, _ := bdp.lastUUID.Load().(string)

		partitionSeq := partition +
			`":{"seq":` + strconv.FormatUint(bdpSeqMaxBatch, 10) +
			`,"seqReceived":` + strconv.FormatUint(bdpSeqMax, 10)

		if len(scope) > 0 && len(collections) > 0 && sourcePartitionSeqs != nil {
			var highSeq uint64
			// determine which collection holds the highest sequence number
			for _, coll := range collections {
				if uuidSeq, exists :=
					sourcePartitionSeqs[partition+":"+scope+":"+coll]; exists {
					if highSeq < uuidSeq.Seq {
						highSeq = uuidSeq.Seq
					}
				}
			}
			partitionSeq += `,"sourceSeq":` + strconv.FormatUint(highSeq, 10)
		}

		partitionSeq += `,"uuid":"` + bdpLastUUID + `"}`

		partitionSeqs[i] = []byte(partitionSeq)

		i++
	}
	t.m.RUnlock()

	for i, partitionSeq := range partitionSeqs {
		if i == 0 {
			_, err = w.Write([]byte(`"`))
			if err != nil {
				return
			}
		} else {
			_, err = w.Write([]byte(`,"`))
			if err != nil {
				return
			}
		}

		_, err = w.Write(partitionSeq)
		if err != nil {
			return
		}
	}
	_, err = w.Write(cbgt.JsonCloseBrace)
	if err != nil {
		return
	}

	_, err = w.Write(prefixCopyPartitionStats)
	if err != nil {
		return err
	}

	t.copyStats.WriteJSON(w)

	_, err = w.Write(prefixHerderStats)
	if err != nil {
		return err
	}

	// Since herder and its stats are unique to cbft, writing these stats here
	// instead of in cbgt/rest.
	writeHerderStatsJSON(w)

	_, err = w.Write(cbgt.JsonCloseBrace)
	if err != nil {
		return
	}

	return nil
}

func CustomSeqTimeoutCheck(pindex string, data []byte) bool {
	m := struct {
		PIndexes map[string]struct {
			HerderStats struct {
				TotWaitingIn  uint64 `json:"TotWaitingIn,omitempty"`
				TotWaitingOut uint64 `json:"TotWaitingOut,omitempty"`
			} `json:"herderStats,omitempty"`
		} `json:"pindexes"`
	}{}

	err := json.Unmarshal(data, &m)
	if err != nil {
		return false
	}

	// true since batches blocked by herder for this node's cbft process.
	if m.PIndexes[pindex].HerderStats.TotWaitingIn > 0 &&
		m.PIndexes[pindex].HerderStats.TotWaitingIn >
			m.PIndexes[pindex].HerderStats.TotWaitingOut {
		return true
	}

	// Return false since none of the batches/index are blocked.
	return false
}

func (t *BleveDest) StatsMap() (rv map[string]interface{}, err error) {
	rv = make(map[string]interface{})

	t.m.RLock()
	defer t.m.RUnlock()

	if t.bindex != nil {
		rv["bleveIndexStats"] = t.bindex.StatsMap()
		var c uint64
		c, err = t.bindex.DocCount()
		if err != nil {
			return
		}

		rv["DocCount"] = c
	}

	return
}

// ---------------------------------------------------------

// Implements the PartitionSeqProvider interface.
func (t *BleveDest) PartitionSeqs() (map[string]cbgt.UUIDSeq, error) {
	rv := map[string]cbgt.UUIDSeq{}
	t.m.RLock()
	for partition, bdp := range t.partitions {
		bdpSeqMaxBatch := atomic.LoadUint64(&bdp.seqMaxBatch)
		bdpLastUUID, _ := bdp.lastUUID.Load().(string)

		rv[partition] = cbgt.UUIDSeq{
			UUID: bdpLastUUID,
			Seq:  bdpSeqMaxBatch,
		}
	}
	t.m.RUnlock()

	return rv, nil
}

// ---------------------------------------------------------

func (t *BleveDestPartition) Close(remove bool) error {
	return t.bdest.Close(remove)
}

func (t *BleveDestPartition) PrepareFeedParams(partition string,
	params *cbgt.DCPFeedParams) error {
	// nothing to be done for bucket based indexes.
	if !strings.HasPrefix(t.bdest.bleveDocConfig.Mode, ConfigModeCollPrefix) {
		return nil
	}
	// if already set, then return early.
	if params != nil && params.Scope != "" && len(params.Collections) > 0 {
		return nil
	}

	buf, err := os.ReadFile(t.bdest.path +
		string(os.PathSeparator) + "PINDEX_META")
	if err != nil {
		return err
	}
	if len(buf) == 0 {
		return fmt.Errorf("bleve: empty PINDEX_META contents")
	}

	in := struct {
		IndexParams string `json:"indexParams"`
		SourceName  string `json:"sourceName"`
	}{}
	err = json.Unmarshal(buf, &in)
	if err != nil {
		return fmt.Errorf("bleve: parse params: %v", err)
	}

	tmp := struct {
		Mapping mapping.IndexMapping `json:"mapping"`
	}{Mapping: bleve.NewIndexMapping()}

	err = json.Unmarshal([]byte(in.IndexParams), &tmp)
	if err != nil {
		return fmt.Errorf("bleve: parse params: %v", err)
	}

	if im, ok := tmp.Mapping.(*mapping.IndexMappingImpl); ok {
		scope, err := validateScopeCollFromMappings(in.SourceName,
			im, true)
		if err != nil {
			return err
		}

		if scope != nil && len(scope.Collections) > 0 {
			params.Scope = scope.Name
			uniqueCollections := map[string]struct{}{}
			for _, coll := range scope.Collections {
				if _, exists := uniqueCollections[coll.Name]; !exists {
					uniqueCollections[coll.Name] = struct{}{}
					params.Collections = append(params.Collections, coll.Name)
				}
			}
		}
	}

	return nil
}

func (t *BleveDestPartition) dataUpdate(partition string,
	key []byte, seq uint64, val []byte, cas uint64,
	extrasType cbgt.DestExtrasType, req interface{}, extras []byte) error {

	atomic.AddUint64(&aggregateBDPStats.TotDataUpdateBeg, 1)

	t.m.Lock()
	if t.batch == nil {
		t.m.Unlock()
		atomic.AddUint64(&aggregateBDPStats.TotDataUpdateEnd, 1)
		return fmt.Errorf("bleve: DataUpdate nil batch")
	}

	defaultType := "_default"
	if imi, ok := t.bindex.Mapping().(*mapping.IndexMappingImpl); ok {
		defaultType = imi.DefaultType
	}

	cbftDoc, key, errv := t.bdest.bleveDocConfig.BuildDocumentEx(key, val,
		defaultType, extrasType, req, extras)

	erri := t.batch.Index(string(key), cbftDoc)

	revNeedsUpdate, err := t.updateSeqLOCKED(seq)

	t.m.Unlock()

	if err == nil && revNeedsUpdate {
		t.incRev()
	}
	if errv != nil {
		t.bdest.AddError("json.Unmarshal", partition, key, seq, val, errv)
	}
	if erri != nil {
		t.bdest.AddError("batch.Index", partition, key, seq, val, erri)
	}

	atomic.AddUint64(&aggregateBDPStats.TotDataUpdateEnd, 1)
	return err
}

func (t *BleveDestPartition) DataUpdate(partition string,
	key []byte, seq uint64, val []byte, cas uint64,
	extrasType cbgt.DestExtrasType, extras []byte) error {

	return t.dataUpdate(partition, key, seq, val, cas, extrasType, nil, extras)
}

func (t *BleveDestPartition) DataUpdateEx(partition string,
	key []byte, seq uint64, val []byte, cas uint64,
	extrasType cbgt.DestExtrasType, req interface{}) error {

	return t.dataUpdate(partition, key, seq, val, cas, extrasType, req, nil)
}

func (t *BleveDestPartition) DataDelete(partition string,
	key []byte, seq uint64,
	cas uint64,
	extrasType cbgt.DestExtrasType, extras []byte) error {
	atomic.AddUint64(&aggregateBDPStats.TotDataDeleteBeg, 1)

	t.m.Lock()

	if t.batch == nil {
		t.m.Unlock()
		atomic.AddUint64(&aggregateBDPStats.TotDataDeleteEnd, 1)
		return fmt.Errorf("bleve: DataDelete nil batch")
	}

	// need to apply the key decoration with multicollection indexes.
	if t.bdest.bleveDocConfig.multiCollection() {
		key = append(extras[4:8], key...)
	}

	t.batch.Delete(string(key)) // TODO: string(key) makes garbage?

	revNeedsUpdate, err := t.updateSeqLOCKED(seq)

	t.m.Unlock()

	if err == nil && revNeedsUpdate {
		t.incRev()
	}

	atomic.AddUint64(&aggregateBDPStats.TotDataDeleteEnd, 1)
	return err
}

func (t *BleveDestPartition) DataDeleteEx(partition string, key []byte, seq uint64,
	cas uint64,
	extrasType cbgt.DestExtrasType, req interface{}) error {

	dcpExtras, ok := req.(cbgt.GocbcoreDCPExtras)
	if !ok {
		return fmt.Errorf("bleve: DataDeleteEx unable to typecast GocbcoreDCPExtras")
	}

	extras := make([]byte, 8)
	binary.LittleEndian.PutUint32(extras[0:], dcpExtras.ScopeId)
	binary.LittleEndian.PutUint32(extras[4:], dcpExtras.CollectionId)

	return t.DataDelete(partition, key, seq, cas, extrasType, extras)
}

// ---------------------------------------------------------

const osoSnapshotStart = uint32(1)
const osoSnapshotEnd = uint32(2)

func (t *BleveDestPartition) OSOSnapshot(partition string,
	snapshotType uint32) error {
	if snapshotType == osoSnapshotStart {
		t.m.Lock()
		t.osoSnapshot = true
		t.osoSeqMax = atomic.LoadUint64(&t.seqMax)
		revNeedsUpdate, err := t.submitAsyncBatchRequestLOCKED()
		t.m.Unlock()
		if err == nil && revNeedsUpdate {
			t.incRev()
		}

		return err
	} else if snapshotType == osoSnapshotEnd {
		t.m.Lock()
		t.osoSnapshot = false
		// When the OSO snapshot end message is received, update seqMax with
		// the max seq received while in the OSO snapshot and flush the batch.
		if atomic.LoadUint64(&t.seqMax) < t.osoSeqMax {
			atomic.StoreUint64(&t.seqMax, t.osoSeqMax)
		}
		revNeedsUpdate, err := t.submitAsyncBatchRequestLOCKED()
		t.m.Unlock()
		if err == nil && revNeedsUpdate {
			t.incRev()
		}

		return err
	}

	return fmt.Errorf("bleve: OSOSnapshot unknown snapshotType: %v",
		snapshotType)
}

func (t *BleveDestPartition) SeqNoAdvanced(partition string, seq uint64) error {
	// Can be received when the feed is subscribed to collections;
	// This message is to be viewed as a snapshot END message, indicating that
	// no other sequence numbers will be received in that snapshot;
	// Also, this seqno is to be used in the event of re-connection.
	t.m.Lock()
	revNeedsUpdate, err := t.updateSeqLOCKED(seq)
	t.m.Unlock()
	if err == nil && revNeedsUpdate {
		t.incRev()
	}

	return err
}

func (t *BleveDestPartition) CreateCollection(partition string,
	manifestUid uint64, scopeId, collecitonId uint32, seq uint64) error {
	t.m.Lock()
	revNeedsUpdate, err := t.updateSeqLOCKED(seq)
	t.m.Unlock()
	if err == nil && revNeedsUpdate {
		t.incRev()
	}

	return err
}

func (t *BleveDestPartition) DeleteCollection(partition string,
	manifestUid uint64, scopeId, collecitonId uint32, seq uint64) error {
	// NO-OP
	return nil
}

func (t *BleveDestPartition) FlushCollection(partition string,
	manifestUid uint64, scopeId, collecitonId uint32, seq uint64) error {
	// NO-OP
	return nil
}

func (t *BleveDestPartition) ModifyCollection(partition string,
	manifestUid uint64, scopeId, collecitonId uint32, seq uint64) error {
	// NO-OP
	return nil
}

// ---------------------------------------------------------

type bleveDestPartitionStats struct {
	TotDataUpdateBeg uint64
	TotDataUpdateEnd uint64

	TotDataDeleteBeg uint64
	TotDataDeleteEnd uint64

	TotExecuteBatchBeg uint64
	TotExecuteBatchEnd uint64
}

var aggregateBDPStats bleveDestPartitionStats

func AggregateBleveDestPartitionStats() map[string]interface{} {
	return map[string]interface{}{
		"TotDataUpdateBeg": atomic.LoadUint64(&aggregateBDPStats.TotDataUpdateBeg),
		"TotDataUpdateEnd": atomic.LoadUint64(&aggregateBDPStats.TotDataUpdateEnd),

		"TotDataDeleteBeg": atomic.LoadUint64(&aggregateBDPStats.TotDataDeleteBeg),
		"TotDataDeleteEnd": atomic.LoadUint64(&aggregateBDPStats.TotDataDeleteEnd),

		"TotExecuteBatchBeg": atomic.LoadUint64(&aggregateBDPStats.TotExecuteBatchBeg),
		"TotExecuteBatchEnd": atomic.LoadUint64(&aggregateBDPStats.TotExecuteBatchEnd),
	}
}

// ---------------------------------------------------------

func (t *BleveDestPartition) SnapshotStart(partition string,
	snapStart, snapEnd uint64) error {
	t.m.Lock()
	t.osoSnapshot = false
	revNeedsUpdate, err := t.submitAsyncBatchRequestLOCKED()
	if err != nil {
		t.m.Unlock()
		return err
	}

	t.seqSnapEnd = snapEnd

	t.m.Unlock()

	if revNeedsUpdate {
		t.incRev()
	}

	return nil
}

func (t *BleveDestPartition) OpaqueGet(partition string) ([]byte, uint64, error) {
	t.m.Lock()

	if t.lastOpaque == nil {
		// TODO: Need way to control memory alloc during GetInternal(),
		// perhaps with optional memory allocator func() parameter?
		value, err := t.bindex.GetInternal(t.partitionOpaque)
		if err != nil {
			t.m.Unlock()
			return nil, 0, err
		}
		t.lastOpaque = append([]byte(nil), value...) // Note: copies value.
		t.lastUUID.Store(cbgt.ParseOpaqueToUUID(value))
	}

	lastOpaque := append([]byte(nil), t.lastOpaque...) // Another copy of value.

	if atomic.LoadUint64(&t.seqMax) <= 0 {
		// TODO: Need way to control memory alloc during GetInternal(),
		// perhaps with optional memory allocator func() parameter?
		buf, err := t.bindex.GetInternal(t.partitionBytes)
		if err != nil {
			t.m.Unlock()
			return nil, 0, err
		}
		if len(buf) <= 0 {
			t.m.Unlock()
			return lastOpaque, 0, nil // No seqMax buf is a valid case.
		}
		if len(buf) != 8 {
			t.m.Unlock()
			return nil, 0, fmt.Errorf("bleve: unexpected size for seqMax bytes")
		}

		atomic.StoreUint64(&t.seqMax, binary.BigEndian.Uint64(buf[0:8]))
		atomic.CompareAndSwapUint64(&t.seqMaxBatch, 0, atomic.LoadUint64(&t.seqMax))
	}

	lastSeq := atomic.LoadUint64(&t.seqMax)
	lastCommittedSeq := atomic.LoadUint64(&t.seqMaxBatch)
	if lastSeq > lastCommittedSeq {
		lastSeq = lastCommittedSeq
	}

	t.m.Unlock()

	return lastOpaque, lastSeq, nil
}

func (t *BleveDestPartition) OpaqueSet(partition string, value []byte) error {
	t.m.Lock()

	if t.batch == nil {
		t.m.Unlock()
		return fmt.Errorf("bleve: OpaqueSet nil batch")
	}

	t.lastOpaque = append(t.lastOpaque[0:0], value...)
	t.lastUUID.Store(cbgt.ParseOpaqueToUUID(value))

	anotherCopy := append([]byte(nil), value...)
	t.batch.SetInternal(t.partitionOpaque, anotherCopy)

	t.m.Unlock()
	return nil
}

func (t *BleveDestPartition) ConsistencyWait(
	partition, partitionUUID string,
	consistencyLevel string,
	consistencySeq uint64,
	cancelCh <-chan bool) error {
	return t.bdest.ConsistencyWait(partition, partitionUUID,
		consistencyLevel, consistencySeq, cancelCh)
}

func (t *BleveDestPartition) Count(pindex *cbgt.PIndex,
	cancelCh <-chan bool) (
	uint64, error) {
	return t.bdest.Count(pindex, cancelCh)
}

func (t *BleveDestPartition) Query(pindex *cbgt.PIndex,
	req []byte, res io.Writer,
	cancelCh <-chan bool) error {
	return t.bdest.Query(pindex, req, res, cancelCh)
}

func (t *BleveDestPartition) Stats(w io.Writer) error {
	return t.bdest.Stats(w)
}

// ---------------------------------------------------------

func (t *BleveDestPartition) updateSeqLOCKED(seq uint64) (bool, error) {
	if t.osoSnapshot {
		if t.osoSeqMax < seq {
			t.osoSeqMax = seq
		}
	} else if atomic.LoadUint64(&t.seqMax) < seq {
		atomic.StoreUint64(&t.seqMax, seq)
	}

	if (t.osoSnapshot || seq < t.seqSnapEnd) &&
		(BleveMaxOpsPerBatch <= 0 || BleveMaxOpsPerBatch > t.batch.Size()) {
		return false, t.lastAsyncBatchErr
	}

	return t.submitAsyncBatchRequestLOCKED()
}

func (t *BleveDestPartition) submitAsyncBatchRequestLOCKED() (bool, error) {
	// fetch the needed parameters and remain unlocked until requestCh
	// is ready to accommodate this request
	bindex := t.bindex
	seqMax := atomic.LoadUint64(&t.seqMax)
	seqMaxBuf := make([]byte, 8)
	binary.BigEndian.PutUint64(seqMaxBuf, seqMax)
	t.batch.SetInternal(t.partitionBytes, seqMaxBuf)
	batch := t.batch
	t.batch = t.bindex.NewBatch()
	p := t.partition
	batchReqChs := t.bdest.batchReqChs
	stopCh := t.bdest.stopCh
	t.m.Unlock()

	// ensure that batch requests from a given partition always goes
	// to the same worker queue so that the order of seq numbers are maintained
	partition, err := strconv.Atoi(p)
	if err != nil {
		log.Errorf("pindex_bleve: submitAsyncBatchRequestLOCKED over `%v`, err: %v",
			bindex.Name(), err)
		t.m.Lock()
		return false, err
	}

	reqChIndex := partition % asyncBatchWorkerCount
	br := &batchRequest{bdp: t, bindex: bindex,
		batch: batch, seqMax: seqMax,
	}
	select {
	case <-stopCh:
		log.Printf("pindex_bleve: submitAsyncBatchRequestLOCKED stopped for `%v`",
			bindex.Name())
		t.m.Lock()
		return false, t.lastAsyncBatchErr

	case batchReqChs[reqChIndex] <- br:
	}

	// acquire lock
	t.m.Lock()
	return false, t.lastAsyncBatchErr
}

func (t *BleveDestPartition) setLastAsyncBatchErr(err error) {
	t.m.Lock()
	t.lastAsyncBatchErr = err
	t.m.Unlock()
}

func runBatchWorker(requestCh chan *batchRequest, stopCh chan struct{},
	bindex bleve.Index, workerID int) {
	var targetBatch *bleve.Batch
	bdp := make([]*BleveDestPartition, 0, 50)
	bdpMaxSeqNums := make([]uint64, 0, 50)
	var ticker *time.Ticker
	batchFlushDuration := BleveBatchFlushDuration

	if batchFlushDuration > 0 {
		ticker = time.NewTicker(batchFlushDuration)
		defer ticker.Stop()
	}
	var tickerCh <-chan time.Time

	log.Printf("pindex_bleve: started runBatchWorker: %d for pindex: %s", workerID, bindex.Name())

	for {
		// trigger batch execution if we have enough items in batch
		if targetBatch != nil && targetBatch.Size() >= BleveMaxOpsPerBatch {
			executeBatch(bdp, bdpMaxSeqNums, bindex, targetBatch)
			targetBatch = nil
			atomic.AddUint64(&TotBatchesFlushedOnMaxOps, 1)
		}

		// wait for more mutations for a bigger target batch
		if targetBatch != nil && ticker != nil {
			tickerCh = ticker.C
		}

		select {
		case batchReq := <-requestCh:
			if batchReq == nil {
				log.Printf("pindex_bleve: batchWorker stopped, batchReq: nil")
				return
			}
			if batchReq.bdp == nil || batchReq.bindex == nil {
				break
			}

			// if batch merging is disabled then execute the batch
			if batchFlushDuration == 0 {
				bdp = bdp[:0]
				bdpMaxSeqNums = bdpMaxSeqNums[:0]
				batchReq.bdp.m.Lock()
				bdp = append(bdp, batchReq.bdp)
				bdpMaxSeqNums = append(bdpMaxSeqNums, batchReq.seqMax)
				batchReq.bdp.m.Unlock()
				executeBatch(bdp, bdpMaxSeqNums, batchReq.bindex, batchReq.batch)
				break
			}

			if targetBatch == nil {
				bdp = bdp[:0]
				bdpMaxSeqNums = bdpMaxSeqNums[:0]
				batchReq.bdp.m.Lock()
				bdp = append(bdp, batchReq.bdp)
				bdpMaxSeqNums = append(bdpMaxSeqNums, batchReq.seqMax)
				batchReq.bdp.m.Unlock()
				bindex = batchReq.bindex
				targetBatch = batchReq.batch
				atomic.AddUint64(&TotBatchesNew, 1)
				break
			}

			targetBatch.Merge(batchReq.batch)
			atomic.AddUint64(&TotBatchesMerged, 1)
			batchReq.bdp.m.Lock()
			bdp = append(bdp, batchReq.bdp)
			bdpMaxSeqNums = append(bdpMaxSeqNums, batchReq.seqMax)
			batchReq.bdp.m.Unlock()

		case <-tickerCh:
			if targetBatch != nil {
				executeBatch(bdp, bdpMaxSeqNums, bindex, targetBatch)
				targetBatch = nil
				atomic.AddUint64(&TotBatchesFlushedOnTimer, 1)
			}
			tickerCh = nil

		case <-stopCh:
			log.Printf("pindex_bleve: batchWorker %d stopped for `%v`", workerID, bindex.Name())
			return
		}

	}
}

func executeBatch(bdp []*BleveDestPartition, bdpMaxSeqNums []uint64,
	index bleve.Index, batch *bleve.Batch) {

	// perform a regulator check and then the execute() in case of
	// serverless mode. If that is successful, perform the corresponding
	// metering.
	regulateAndExecute(bdp, bdpMaxSeqNums, index, batch)
}

func regulateAndExecute(bdp []*BleveDestPartition, bdpMaxSeqNums []uint64,
	index bleve.Index, batch *bleve.Batch) {

	// The metering and checkQuota calls must happen on
	// indexes with valid sources, i.e. their dests should contain
	// valid sourceName. So, in scenarios like rebalance and hibernation
	// when there is a no-op bleve dest involved, those operations
	//  shouldn't happen.
	if !ServerlessMode || bdp[0].bdest.sourceName == "" {
		_, err := execute(bdp, bdpMaxSeqNums, index, batch)
		if err != nil {
			bdp[0].setLastAsyncBatchErr(err)
		}
		return
	}

	result, _, err := CheckQuotaWrite(bdp[0].bdest.stopCh,
		bdp[0].bdest.sourceName, "", true, index)

	// NOTE: At this point, it's guaranteed that the action is either
	// CheckResultError (which is the case when the bdp's bleve index is closed)
	// or CheckResultNormal, in which case we continue the execution
	if result == CheckResultError {
		log.Error(err)
		return
	}

	_, err = execute(bdp, bdpMaxSeqNums, index, batch)
	if err != nil {
		bdp[0].setLastAsyncBatchErr(err)
		return
	}

	// NOTE: each partition is going to fetch its bleve index's
	// bytes written to disk, and meter this to its local node
	// in the cluster.
	MeterWrites(bdp[0].bdest.stopCh, bdp[0].bdest.sourceName, index)
}

func execute(bdp []*BleveDestPartition, bdpMaxSeqNums []uint64,
	bindex bleve.Index, batch *bleve.Batch) (bool, error) {
	if batch == nil {
		return false, fmt.Errorf("pindex_bleve: executeBatch batch nil")
	}

	if bindex == nil {
		return false, fmt.Errorf("pindex_bleve: executeBatch bindex already closed")
	}

	batchTotalDocsSize := batch.TotalDocsSize()
	atomic.AddUint64(&BatchBytesAdded, batchTotalDocsSize)

	err := cbgt.Timer(func() error {
		atomic.AddUint64(&aggregateBDPStats.TotExecuteBatchBeg, 1)
		err := bindex.Batch(batch)
		atomic.AddUint64(&aggregateBDPStats.TotExecuteBatchEnd, 1)
		if err != nil && err != bleve.ErrorIndexClosed {
			log.Errorf("pindex_bleve: executeBatch over `%v`, err: %+v ",
				bindex.Name(), err)
		}
		return err
	}, bdp[0].bdest.stats.TimerBatchStore)

	if err != nil {
		return false, err
	}

	atomic.AddUint64(&NumBatchesIntroduced, 1)
	atomic.AddUint64(&BatchBytesRemoved, batchTotalDocsSize)

	for i, t := range bdp {
		t.m.Lock()
		if bdpMaxSeqNums[i] > atomic.LoadUint64(&t.seqMaxBatch) {
			atomic.StoreUint64(&t.seqMaxBatch, bdpMaxSeqNums[i])
		}
		for t.cwrQueue.Len() > 0 &&
			t.cwrQueue[0].ConsistencySeq <= atomic.LoadUint64(&t.seqMaxBatch) {
			cwr := heap.Pop(&t.cwrQueue).(*cbgt.ConsistencyWaitReq)
			if cwr != nil && cwr.DoneCh != nil {
				close(cwr.DoneCh)
			}
		}
		t.m.Unlock()
	}

	return true, nil
}

// ---------------------------------------------------------

func (t *BleveDestPartition) incRev() {
	t.bdest.m.Lock()
	t.bdest.rev++
	t.bdest.m.Unlock()
}

// ---------------------------------------------------------

// Atomic counters that keep track of the number of times http and http-ssl
// were used for scatter gather over remote pindexes.
var totRemoteHttp uint64
var totRemoteHttpSsl uint64

// ---------------------------------------------------------

// Returns a bleve.IndexAlias that represents all the PIndexes for the
// index, including perhaps bleve remote client PIndexes.
//
// TODO: Perhaps need a tighter check around indexUUID, as the current
// implementation might have a race where old pindexes with a matching
// (but invalid) indexUUID might be hit.
//
// TODO: If this returns an error, perhaps the caller somewhere up the
// chain should close the cancelCh to help stop any other inflight
// activities.
func bleveIndexAlias(mgr *cbgt.Manager, indexName, indexUUID string,
	ensureCanRead bool, consistencyParams *cbgt.ConsistencyParams,
	cancelCh <-chan bool, groupByNode bool, onlyPIndexes map[string]bool,
	partitionSelection string, rcAdder addRemoteClients) (
	bleve.IndexAlias, []RemoteClient, int, error) {

	alias := bleve.NewIndexAlias()

	remoteClients, numPIndexes, err := bleveIndexTargets(mgr, indexName, indexUUID,
		ensureCanRead, consistencyParams, cancelCh,
		groupByNode, onlyPIndexes, alias, partitionSelection, rcAdder)
	if err != nil {
		if _, ok := err.(*cbgt.ErrorLocalPIndexHealth); ok {
			return alias, remoteClients, numPIndexes, err
		}
		return nil, nil, 0, err
	}

	return alias, remoteClients, numPIndexes, nil
}

// BleveIndexCollector interface is a subset of the bleve.IndexAlias
// interface, with just the Add() method, allowing alternative
// implementations that need to collect "backend" bleve indexes based
// on a user defined index.
type BleveIndexCollector interface {
	Add(i ...bleve.Index)
	VisitIndexes(func(bleve.Index))
}

func addIndexClients(mgr *cbgt.Manager, indexName, indexUUID string,
	remotePlanPIndexes []*cbgt.RemotePlanPIndex, consistencyParams *cbgt.ConsistencyParams,
	onlyPIndexes map[string]bool, collector BleveIndexCollector,
	groupByNode bool) ([]RemoteClient, error) {
	prefix := mgr.GetOption("urlPrefix")
	remoteClients := make([]*IndexClient, 0, len(remotePlanPIndexes))
	rv := make([]RemoteClient, 0, len(remotePlanPIndexes))
	ss := cbgt.GetSecuritySetting()
	for _, remotePlanPIndex := range remotePlanPIndexes {
		if (onlyPIndexes != nil && !onlyPIndexes[remotePlanPIndex.PlanPIndex.Name]) ||
			remotePlanPIndex.NodeDef == nil {
			continue
		}

		hostPortUrl := "http://" + remotePlanPIndex.NodeDef.HostPort
		if ss.EncryptionEnabled {
			if u, err := remotePlanPIndex.NodeDef.HttpsURL(); err == nil {
				hostPortUrl = u
				atomic.AddUint64(&totRemoteHttpSsl, 1)
			} else {
				atomic.AddUint64(&totRemoteHttp, 1)
			}
		} else {
			atomic.AddUint64(&totRemoteHttp, 1)
		}

		baseURL := hostPortUrl + prefix +
			"/api/pindex/" + remotePlanPIndex.PlanPIndex.Name

		indexClient := &IndexClient{
			mgr:         mgr,
			name:        fmt.Sprintf("IndexClient - %s", baseURL),
			HostPort:    remotePlanPIndex.NodeDef.HostPort,
			IndexName:   indexName,
			IndexUUID:   indexUUID,
			PIndexNames: []string{remotePlanPIndex.PlanPIndex.Name},
			QueryURL:    baseURL + "/query",
			CountURL:    baseURL + "/count",
			Consistency: consistencyParams,
			httpClient:  cbgt.HttpClient(),
		}

		remoteClients = append(remoteClients, indexClient)
	}

	if groupByNode {
		remoteClients, _ = GroupIndexClientsByHostPort(remoteClients)
	}

	for _, remoteClient := range remoteClients {
		collector.Add(remoteClient)
		rv = append(rv, remoteClient)
	}

	return rv, nil
}

// PartitionSelectionStrategy lets clients specify any selection
// preferences for the query serving index partitions spread across
// the cluster.
// PartitionSelectionStrategy recognized options are,
//   - ""              : primary partitions are selected
//   - local           : local partitions are favored, pseudorandom selection from remote
//   - random          : pseudorandom selection from available local and remote
//   - random_balanced : pseudorandom selection from available local and remote nodes by
//     equally distributing the query load across all nodes.
type PartitionSelectionStrategy string

var FetchBleveTargets = func(mgr *cbgt.Manager, indexName, indexUUID string,
	planPIndexFilterName string, partitionSelection PartitionSelectionStrategy) (
	[]*cbgt.PIndex, []*cbgt.RemotePlanPIndex, []string, error) {
	if mgr == nil {
		return nil, nil, nil, fmt.Errorf("manager not defined")
	}

	return mgr.CoveringPIndexesEx(cbgt.CoveringPIndexesSpec{
		IndexName:            indexName,
		IndexUUID:            indexUUID,
		PlanPIndexFilterName: planPIndexFilterName,
	}, nil, false)
}

func bleveIndexTargets(mgr *cbgt.Manager, indexName, indexUUID string,
	ensureCanRead bool, consistencyParams *cbgt.ConsistencyParams,
	cancelCh <-chan bool, groupByNode bool, onlyPIndexes map[string]bool,
	collector BleveIndexCollector, partitionSelection string,
	rcAdder addRemoteClients) (
	[]RemoteClient, int, error) {
	planPIndexFilterName := "ok"
	if ensureCanRead {
		planPIndexFilterName = "canRead"
	}

	if onlyPIndexes != nil && partitionSelection != "" {
		// select all local pindexes
		partitionSelection = "local"
	}

	localPIndexesAll, remotePlanPIndexes, missingPIndexNames, err :=
		FetchBleveTargets(mgr, indexName, indexUUID,
			planPIndexFilterName, PartitionSelectionStrategy(partitionSelection))
	if err != nil {
		return nil, 0, fmt.Errorf("bleve: bleveIndexTargets, err: %v", err)
	}
	if consistencyParams != nil &&
		consistencyParams.Results == "complete" &&
		len(missingPIndexNames) > 0 {
		return nil, 0, fmt.Errorf("bleve: some index partitions aren't reachable,"+
			" missing: %v", len(missingPIndexNames))
	}

	numPIndexes := len(localPIndexesAll) + len(remotePlanPIndexes)

	localPIndexes := localPIndexesAll
	if onlyPIndexes != nil {
		localPIndexes = make([]*cbgt.PIndex, 0, len(localPIndexesAll))
		for _, localPIndex := range localPIndexesAll {
			if onlyPIndexes[localPIndex.Name] {
				localPIndexes = append(localPIndexes, localPIndex)
			}
		}
		if len(localPIndexes) == 0 {
			var wantedPIndexNames []string
			for pIndexName := range onlyPIndexes {
				wantedPIndexNames = append(wantedPIndexNames, pIndexName)
			}
			var hadPIndexNames []string
			for _, pIndex := range localPIndexesAll {
				hadPIndexNames = append(hadPIndexNames, pIndex.Name)
			}
			log.Warnf("pindex_bleve: bleveIndexTargets 0 localPIndexes, wanted: %v, had: %v for index: %s",
				wantedPIndexNames, hadPIndexNames, indexName)
		}
	}

	for _, missingPIndexName := range missingPIndexNames {
		if onlyPIndexes == nil || onlyPIndexes[missingPIndexName] {
			collector.Add(&MissingPIndex{
				name: missingPIndexName,
			})
		}
	}

	remoteClients, err := rcAdder(mgr, indexName, indexUUID,
		remotePlanPIndexes, consistencyParams, onlyPIndexes,
		collector, groupByNode)
	if err != nil {
		return nil, numPIndexes, err
	}

	// TODO: Should kickoff remote queries concurrently before we wait.

	return remoteClients, numPIndexes, cbgt.ConsistencyWaitGroup(indexName, consistencyParams,
		cancelCh, localPIndexes,
		func(localPIndex *cbgt.PIndex) error {
			bindex, _, rev, err := bleveIndex(localPIndex)
			if err != nil {
				return err
			}

			collector.Add(&cacheBleveIndex{
				pindex: localPIndex,
				bindex: bindex,
				rev:    rev,
				name:   bindex.Name(),
			})

			return nil
		})
}

func bleveIndex(localPIndex *cbgt.PIndex) (bleve.Index, *BleveDest, uint64, error) {
	if !strings.HasPrefix(localPIndex.IndexType, "fulltext-index") {
		return nil, nil, 0, fmt.Errorf("bleve: bleveIndexTargets, wrong type,"+
			" localPIndex: %s, type: %s", localPIndex.Name, localPIndex.IndexType)
	}

	destFwd, ok := localPIndex.Dest.(*cbgt.DestForwarder)
	if !ok || destFwd == nil {
		return nil, nil, 0, fmt.Errorf("bleve: bleveIndexTargets, wrong destFwd type,"+
			" localPIndex: %s, destFwd type: %T", localPIndex.Name, localPIndex.Dest)
	}

	bdest, ok := destFwd.DestProvider.(*BleveDest)
	if !ok || bdest == nil {
		return nil, nil, 0, fmt.Errorf("bleve: bleveIndexTargets, wrong provider type,"+
			" localPIndex: %s, provider type: %T", localPIndex.Name, destFwd.DestProvider)
	}

	bdest.m.RLock()
	bindex := bdest.bindex
	rev := bdest.rev
	bdest.m.RUnlock()

	if bindex == nil {
		return nil, nil, 0, fmt.Errorf("bleve: bleveIndexTargets, nil bindex,"+
			" localPIndex: %s", localPIndex.Name)
	}

	return bindex, bdest, rev, nil
}

// ---------------------------------------------------------

var BleveRouteMethods map[string]string

func init() {
	BleveRouteMethods = make(map[string]string)
}

func BleveInitRouter(r *mux.Router, phase string,
	mgr *cbgt.Manager) {
	prefix := ""
	if mgr != nil {
		prefix = mgr.GetOption("urlPrefix")
	}

	if phase == "static.before" {
		staticBleveMapping := http.FileServer(bleveMappingUI.AssetFS())

		staticBleveMappingRoutes := AssetNames()

		for _, route := range staticBleveMappingRoutes {
			if strings.Contains(route, "static-bleve-mapping") {
				route = strings.TrimPrefix(route, "ns_server_static/fts")
				r.Handle(prefix+route, http.StripPrefix(prefix+"/static-bleve-mapping/",
					staticBleveMapping))
			}
		}
	}

	if phase == "manager.before" {
		r.Handle(prefix+"/api/index",
			NewFilteredListIndexHandler(mgr)).
			Methods("GET").Name(prefix + "/api/index")
		BleveRouteMethods[prefix+"/api/index"] = "GET"
		r.Handle(prefix+"/api/bucket/{bucketName}/scope/{scopeName}/index",
			NewFilteredListIndexHandler(mgr)).
			Methods("GET").Name(prefix +
			"/api/bucket/{bucketName}/scope/{scopeName}/index")
		BleveRouteMethods[prefix+
			"/api/bucket/{bucketName}/scope/{scopeName}/index"] = "GET"
	}

	if phase == "manager.after" {
		bleveMappingUI.RegisterHandlers(r, prefix+"/api")

		listIndexesHandler := ftsHttp.NewListIndexesHandler()
		r.Handle(prefix+"/api/pindex-bleve",
			listIndexesHandler).Methods("GET")
		BleveRouteMethods[prefix+"/api/pindex-bleve"] = "GET"

		getIndexHandler := ftsHttp.NewGetIndexHandler()
		getIndexHandler.IndexNameLookup = rest.PIndexNameLookup
		r.Handle(prefix+"/api/pindex-bleve/{pindexName}",
			getIndexHandler).Methods("GET")
		BleveRouteMethods[prefix+"/api/pindex-bleve/{pindexName}"] = "GET"

		docCountHandler := ftsHttp.NewDocCountHandler("")
		docCountHandler.IndexNameLookup = rest.PIndexNameLookup
		r.Handle(prefix+"/api/pindex-bleve/{pindexName}/count",
			docCountHandler).Methods("GET")
		BleveRouteMethods[prefix+"/api/pindex-bleve/{pindexName}/count"] = "GET"

		searchHandler := ftsHttp.NewSearchHandler("")
		searchHandler.IndexNameLookup = rest.PIndexNameLookup
		r.Handle(prefix+"/api/pindex-bleve/{pindexName}/query",
			searchHandler).Methods("POST")
		BleveRouteMethods[prefix+"/api/pindex-bleve/{pindexName}/query"] = "POST"

		docGetHandler := ftsHttp.NewDocGetHandler("")
		docGetHandler.IndexNameLookup = rest.PIndexNameLookup
		docGetHandler.DocIDLookup = rest.DocIDLookup
		r.Handle(prefix+"/api/pindex-bleve/{pindexName}/doc/{docID}",
			docGetHandler).Methods("GET")
		BleveRouteMethods[prefix+"/api/pindex-bleve/{pindexName}/doc/{docID}"] = "GET"

		listFieldsHandler := ftsHttp.NewListFieldsHandler("")
		listFieldsHandler.IndexNameLookup = rest.PIndexNameLookup
		r.Handle(prefix+"/api/pindex-bleve/{pindexName}/fields",
			listFieldsHandler).Methods("GET")
		BleveRouteMethods[prefix+"/api/pindex-bleve/{pindexName}/fields"] = "GET"
	}
}

func BleveMetaExtra(m map[string]interface{}) {
	br := make(map[string]map[string][]string)

	t, i := bleveRegistry.AnalyzerTypesAndInstances()
	br["Analyzer"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.CharFilterTypesAndInstances()
	br["CharFilter"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.DateTimeParserTypesAndInstances()
	br["DateTimeParser"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.FragmentFormatterTypesAndInstances()
	br["FragmentFormatter"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.FragmenterTypesAndInstances()
	br["Fragmenter"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.HighlighterTypesAndInstances()
	br["Highlighter"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.KVStoreTypesAndInstances()
	br["KVStore"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.TokenFilterTypesAndInstances()
	br["TokenFilter"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.TokenMapTypesAndInstances()
	br["TokenMap"] = map[string][]string{"types": t, "instances": i}
	t, i = bleveRegistry.TokenizerTypesAndInstances()
	br["Tokenizer"] = map[string][]string{"types": t, "instances": i}

	m["regBleve"] = br
}

// ---------------------------------------------------------

func BleveQuerySamples() []cbgt.Documentation {
	return []cbgt.Documentation{
		{
			Text: "A simple bleve query POST body:",
			JSON: &struct {
				*cbgt.QueryCtlParams
				*bleve.SearchRequest
			}{
				nil,
				&bleve.SearchRequest{
					From:  0,
					Size:  10,
					Query: bleve.NewQueryStringQuery("a sample query"),
				},
			},
		},
		{
			Text: `An example POST body using from/size for results paging,
using ctl for a timeout and for "at_plus" consistency level.
On consistency, the index must have incorporated at least mutation
sequence-number 123 for partition (vbucket) 0 and mutation
sequence-number 234 for partition (vbucket) 1 (where vbucket 1
should have a vbucketUUID of a0b1c2):`,
			JSON: &struct {
				*cbgt.QueryCtlParams
				*bleve.SearchRequest
			}{
				&cbgt.QueryCtlParams{
					Ctl: cbgt.QueryCtl{
						Timeout: cbgt.QUERY_CTL_DEFAULT_TIMEOUT_MS,
						Consistency: &cbgt.ConsistencyParams{
							Level: "at_plus",
							Vectors: map[string]cbgt.ConsistencyVector{
								"customerIndex": {
									"0":        123,
									"1/a0b1c2": 234,
								},
							},
						},
					},
				},
				&bleve.SearchRequest{
					From:      20,
					Size:      10,
					Fields:    []string{"*"},
					Query:     bleve.NewQueryStringQuery("alice smith"),
					Highlight: bleve.NewHighlight(),
					Explain:   true,
				},
			},
		},
	}
}

func reloadableIndexDefParamChange(paramPrev, paramCur string) bool {
	bpPrev := NewBleveParams()
	if len(paramPrev) == 0 {
		// make it a json unmarshal-able string
		paramPrev = "{}"
	}
	err := json.Unmarshal([]byte(paramPrev), bpPrev)
	if err != nil {
		return false
	}

	bpCur := NewBleveParams()
	if len(paramCur) == 0 {
		// make it a json unmarshal-able string
		paramCur = "{}"
	}
	err = json.Unmarshal([]byte(paramCur), bpCur)
	if err != nil {
		return false
	}

	// check for any changes in the spatial engines.
	if bpPrev.Store["spatialPlugin"] != bpCur.Store["spatialPlugin"] {
		return false
	}

	// check for non store parameter differences
	if !reflect.DeepEqual(bpCur.Mapping, bpPrev.Mapping) ||
		!reflect.DeepEqual(bpCur.DocConfig, bpPrev.DocConfig) {
		return false
	}
	// check for indexType updates
	prevType := bpPrev.Store["indexType"]
	curType := bpCur.Store["indexType"]
	if prevType != curType {
		return false
	}
	// always reboot partitions on scorch option changes
	if curType == "scorch" {
		log.Printf("bleve: reloadable scorch option change "+
			" detected, before: %s, after: %s", paramPrev, paramCur)
		return true
	}

	log.Warnf("bleve: unsupported index type: %s", curType)
	return true
}

func reloadableSourceParamsChange(paramPrev, paramCur string) bool {
	if paramPrev == paramCur {
		return true
	}

	if len(paramPrev) == 0 {
		// make it a json unmarshal-able string
		paramPrev = "{}"
	}

	var prevMap map[string]interface{}
	err := json.Unmarshal([]byte(paramPrev), &prevMap)
	if err != nil {
		log.Printf("pindex_bleve: reloadableSourceParamsChange"+
			" json parse paramPrev: %s, err: %v",
			paramPrev, err)
		return false
	}

	if len(paramCur) == 0 {
		// make it a json unmarshal-able string
		paramCur = "{}"
	}

	var curMap map[string]interface{}
	err = json.Unmarshal([]byte(paramCur), &curMap)
	if err != nil {
		log.Printf("pindex_bleve: reloadableSourceParamsChange"+
			" json parse paramCur: %s, err: %v",
			paramCur, err)
		return false
	}

	// any parsing err doesn't matter here.
	po, _ := cbgt.ParseFeedAllotmentOption(paramPrev)
	co, _ := cbgt.ParseFeedAllotmentOption(paramCur)
	if po != co {
		prevMap["feedAllotment"] = ""
		curMap["feedAllotment"] = ""
	}

	return reflect.DeepEqual(prevMap, curMap)
}

// RestartOnIndexDefChanges checks whether the changes in the indexDefns are
// quickly adoptable over a reboot of the pindex implementations.
// eg: kvstore configs updates like compaction percentage.
func RestartOnIndexDefChanges(
	configRequest *cbgt.ConfigAnalyzeRequest) cbgt.ResultCode {
	if configRequest == nil || configRequest.IndexDefnCur == nil ||
		configRequest.IndexDefnPrev == nil {
		return ""
	}
	if configRequest.IndexDefnPrev.Name != configRequest.IndexDefnCur.Name ||
		configRequest.IndexDefnPrev.SourceName !=
			configRequest.IndexDefnCur.SourceName ||
		configRequest.IndexDefnPrev.SourceUUID !=
			configRequest.IndexDefnCur.SourceUUID ||
		!reloadableSourceParamsChange(configRequest.IndexDefnPrev.SourceParams,
			configRequest.IndexDefnCur.SourceParams) ||
		configRequest.IndexDefnPrev.Type !=
			configRequest.IndexDefnCur.Type ||
		!reflect.DeepEqual(configRequest.SourcePartitionsCur,
			configRequest.SourcePartitionsPrev) ||
		!reloadableIndexDefParamChange(configRequest.IndexDefnPrev.Params,
			configRequest.IndexDefnCur.Params) {
		return ""
	}
	return cbgt.PINDEXES_RESTART
}

func mustEncode(w io.Writer, i interface{}) {
	if JSONImpl != nil && JSONImpl.GetManagerOption("jsonImpl") != "std" {
		MustEncodeWithParser(w, i)
	} else {
		rest.MustEncode(w, i)
	}
}

// MustEncodeWithParser encode with the registered parserType,
func MustEncodeWithParser(w io.Writer, i interface{}) {
	rw, rwOk := w.(http.ResponseWriter)
	if rwOk {
		h := rw.Header()
		if h != nil {
			h.Set("Cache-Control", "no-cache")
			if h.Get("Content-type") == "" {
				h.Set("Content-type", "application/json")
			}
		}
	}
	var err error
	if JSONImpl != nil {
		err = JSONImpl.Encode(w, i)
	} else {
		err = fmt.Errorf("bleve: MustEncodeWithParser fails as no custom parser found")
	}
	if err != nil {
		if rwOk {
			crw, ok := rw.(*rest.CountResponseWriter)
			if ok && crw.Wrote {
				return
			}
			rest.PropagateError(rw, nil,
				fmt.Sprintf("rest: custom JSON: %s encode, err: %v",
					JSONImpl.GetParserType(), err), http.StatusInternalServerError)
		}
	}
}

// -----------------------------------------------------------------------------

// This API determines if the sourceName provided within the index
// definition is compatible for FTS indexes by obtaining bucket
// information from the /pools/default/bucket/{bucketName} endpoint.
func checkSourceCompatability(mgr *cbgt.Manager, sourceName string) error {
	if mgr == nil {
		return nil
	}

	if len(sourceName) == 0 {
		return fmt.Errorf("source name not provided")
	}

	url := mgr.Server() + "/pools/default/b/" + url.QueryEscape(sourceName)
	u, err := cbgt.CBAuthURL(url)
	if err != nil {
		return err
	}

	resp, err := HttpGet(cbgt.HttpClient(), u)
	if err != nil {
		return err
	}
	defer resp.Body.Close()

	respBuf, err := io.ReadAll(resp.Body)
	if err != nil {
		return err
	}

	rv := struct {
		BucketType     string `json:"bucketType"`
		StorageBackend string `json:"storageBackend"`
	}{}

	err = UnmarshalJSON(respBuf, &rv)
	if err != nil {
		return err
	}

	if rv.BucketType != "membase" && rv.BucketType != "ephemeral" {
		return fmt.Errorf("unsupported bucket type")
	}

	if rv.BucketType == "membase" {
		switch rv.StorageBackend {
		case "couchstore":
		case "magma":
		default:
			return fmt.Errorf("unsupported storage/backend for bucket: %v",
				rv.StorageBackend)
		}
	}

	return nil
}

func findGeoPoint(dm *mapping.DocumentMapping) bool {
	if dm != nil && dm.Enabled {
		for _, fm := range dm.Fields {
			if fm.Index && fm.Type == "geopoint" {
				return true
			}
		}
		for _, dmapping := range dm.Properties {
			found := findGeoPoint(dmapping)
			if found {
				return true
			}
		}
	}
	return false
}

func isGeoPointFieldInMapping(bp *BleveParams) bool {
	if bp != nil && bp.Mapping != nil {
		if im, ok := bp.Mapping.(*mapping.IndexMappingImpl); ok {
			// look for in the default mapping.
			res := findGeoPoint(im.DefaultMapping)
			if res {
				return true
			}

			// look for among the custom mappings.
			for _, dm := range im.TypeMapping {
				res := findGeoPoint(dm)
				if res {
					return true
				}
			}
		}
	}

	return false
}

func mappingContainsXAttrs(bp *BleveParams) bool {
	if bp == nil {
		return false
	}

	if im, ok := bp.Mapping.(*mapping.IndexMappingImpl); ok {
		if im.DefaultMapping.Enabled {
			if _, ok := im.DefaultMapping.Properties[xattrsMappingName]; ok {
				return true
			}
		}

		for _, tm := range im.TypeMapping {
			if tm.Enabled {
				if _, ok := tm.Properties[xattrsMappingName]; ok {
					return true
				}
			}
		}
	}

	return false
}

const (
	noVectorFields        int = iota
	vectorFields              // only vector fields (7.6.0+)
	vectorAndBase64Fields     // vector + vector_base64 fields (7.6.2+)
)

type vectorPicture struct {
	fields  int
	maxDims int
	cosine  bool
}

// Utility function check if a "vector" typed field is present within
// the index mapping
func vectorPictureFromIndexMapping(m mapping.IndexMapping) vectorPicture {
	im, ok := m.(*mapping.IndexMappingImpl)
	if !ok {
		// cannot interpret index mapping
		return vectorPicture{}
	}

	var vectorPictureFromDocMapping func(*mapping.DocumentMapping) vectorPicture
	vectorPictureFromDocMapping = func(d *mapping.DocumentMapping) vectorPicture {
		rv := vectorPicture{}
		if d != nil && d.Enabled {
			for _, v := range d.Properties {
				val := vectorPictureFromDocMapping(v)
				if val.fields > rv.fields {
					rv.fields = val.fields
				}
				if val.maxDims > rv.maxDims {
					rv.maxDims = val.maxDims
				}
				rv.cosine = rv.cosine || val.cosine
			}

			for _, field := range d.Fields {
				if field.Type == "vector" && vectorFields > rv.fields {
					rv.fields = vectorFields
				}
				if field.Type == "vector_base64" && vectorAndBase64Fields > rv.fields {
					rv.fields = vectorAndBase64Fields
				}
				if field.Dims > rv.maxDims {
					rv.maxDims = field.Dims
				}
				if field.Similarity == "cosine" {
					rv.cosine = true
				}
			}
		}

		return rv
	}

	rv := vectorPicture{}

	// Check DefaultMapping
	val := vectorPictureFromDocMapping(im.DefaultMapping)
	if val.fields > rv.fields {
		rv.fields = val.fields
	}
	if val.maxDims > rv.maxDims {
		rv.maxDims = val.maxDims
	}
	rv.cosine = rv.cosine || val.cosine

	// Iterate over TypeMapping(s)
	for _, d := range im.TypeMapping {
		val := vectorPictureFromDocMapping(d)
		if val.fields > rv.fields {
			rv.fields = val.fields
		}
		if val.maxDims > rv.maxDims {
			rv.maxDims = val.maxDims
		}
		rv.cosine = rv.cosine || val.cosine
	}

	return rv
}