herald.go/server/search.go

975 lines
30 KiB
Go
Raw Normal View History

2021-04-19 21:25:34 +02:00
package server
import (
"context"
"encoding/json"
"fmt"
2021-06-18 06:01:47 +02:00
"log"
"math"
"reflect"
"strings"
2021-07-06 02:20:38 +02:00
"time"
2021-06-18 06:01:47 +02:00
"github.com/lbryio/hub/internal"
"github.com/lbryio/hub/internal/metrics"
2021-05-31 20:53:08 +02:00
pb "github.com/lbryio/hub/protobuf/go"
2021-04-19 21:25:34 +02:00
"github.com/olivere/elastic/v7"
"github.com/prometheus/client_golang/prometheus"
"golang.org/x/text/cases"
2021-06-01 04:19:10 +02:00
"golang.org/x/text/language"
2021-10-01 19:54:03 +02:00
"google.golang.org/protobuf/encoding/protojson"
"gopkg.in/karalabe/cookiejar.v1/collections/deque"
2021-04-19 21:25:34 +02:00
)
// DefaultSearchSize is the default max number of items an
// es search will return.
2021-06-18 06:01:47 +02:00
const DefaultSearchSize = 1000
// record is a struct for the response from es.
2021-04-19 21:25:34 +02:00
type record struct {
2021-07-06 02:20:38 +02:00
Txid string `json:"tx_id"`
Nout uint32 `json:"tx_nout"`
Height uint32 `json:"height"`
ClaimId string `json:"claim_id"`
ChannelId string `json:"channel_id"`
RepostedClaimId string `json:"reposted_claim_id"`
CensorType uint32 `json:"censor_type"`
CensoringChannelId string `json:"censoring_channel_id"`
ShortUrl string `json:"short_url"`
CanonicalUrl string `json:"canonical_url"`
IsControlling bool `json:"is_controlling"`
TakeOverHeight uint32 `json:"last_take_over_height"`
CreationHeight uint32 `json:"creation_height"`
ActivationHeight uint32 `json:"activation_height"`
ExpirationHeight uint32 `json:"expiration_height"`
ClaimsInChannel uint32 `json:"claims_in_channel"`
RepostCount uint32 `json:"repost_count"`
EffectiveAmount uint64 `json:"effective_amount"`
SupportAmount uint64 `json:"support_amount"`
TrendingScore float64 `json:"trending_score"`
ClaimName string `json:"claim_name"`
2021-04-19 21:25:34 +02:00
}
// orderField is struct for specifying ordering of es search results.
2021-05-18 12:02:55 +02:00
type orderField struct {
Field string
2021-06-06 06:57:37 +02:00
IsAsc bool
2021-05-18 12:02:55 +02:00
}
// StrArrToInterface takes an array of strings and returns them as an array of
// interfaces.
2021-05-16 05:13:14 +02:00
func StrArrToInterface(arr []string) []interface{} {
searchVals := make([]interface{}, len(arr))
for i := 0; i < len(arr); i++ {
searchVals[i] = arr[i]
}
return searchVals
}
// Int32ArrToInterface takes an array of int32 and returns them as an array of
// interfaces.
func Int32ArrToInterface(arr []int32) []interface{} {
searchVals := make([]interface{}, len(arr))
for i := 0; i < len(arr); i++ {
searchVals[i] = arr[i]
}
return searchVals
}
// AddTermsField takes an es bool query, array of string values and a term
// name and adds a TermsQuery for that name matching those values to the
// bool query.
2021-06-18 06:01:47 +02:00
func AddTermsField(q *elastic.BoolQuery, arr []string, name string) *elastic.BoolQuery {
if len(arr) == 0 {
return q
2021-05-16 05:13:14 +02:00
}
2021-06-18 06:01:47 +02:00
searchVals := StrArrToInterface(arr)
return q.Must(elastic.NewTermsQuery(name, searchVals...))
2021-05-16 05:13:14 +02:00
}
// AddTermsFieldInt32 takes an es bool query, array of int32 values and a term
// name and adds a TermsQuery for that name matching those values to the
// bool query.
func AddTermsFieldInt32(q *elastic.BoolQuery, arr []int32, name string) *elastic.BoolQuery {
if len(arr) == 0 {
return q
}
searchVals := Int32ArrToInterface(arr)
return q.Must(elastic.NewTermsQuery(name, searchVals...))
}
// AddTermField takes an es bool query, a string value and a term name
// and adds a TermQuery for that name matching that value to the bool
// query.
2021-08-12 09:19:25 +02:00
func AddTermField(q *elastic.BoolQuery, value string, name string) *elastic.BoolQuery {
if value != "" {
return q.Must(elastic.NewTermQuery(name, value))
}
return q
}
// AddIndividualTermFields takes a bool query, an array of string values
// a term name, and a bool to invert the query, and adds multiple individual
// TermQuerys for that name matching each of the values.
2021-06-18 06:01:47 +02:00
func AddIndividualTermFields(q *elastic.BoolQuery, arr []string, name string, invert bool) *elastic.BoolQuery {
for _, x := range arr {
if invert {
q = q.MustNot(elastic.NewTermQuery(name, x))
} else {
q = q.Must(elastic.NewTermQuery(name, x))
2021-06-01 04:19:10 +02:00
}
}
return q
}
// AddRangeField takes a bool query, a range field struct and a term name
// and adds a term query for that name matching that range field.
func AddRangeField(q *elastic.BoolQuery, rqs []*pb.RangeField, name string) *elastic.BoolQuery {
if len(rqs) == 0 {
2021-05-16 05:13:14 +02:00
return q
}
for _, rq := range rqs {
if len(rq.Value) > 1 {
if rq.Op != pb.RangeField_EQ {
continue
}
q = AddTermsFieldInt32(q, rq.Value, name)
continue
}
if rq.Op == pb.RangeField_EQ {
q = q.Must(elastic.NewTermQuery(name, rq.Value[0]))
} else if rq.Op == pb.RangeField_LT {
q = q.Must(elastic.NewRangeQuery(name).Lt(rq.Value[0]))
} else if rq.Op == pb.RangeField_LTE {
q = q.Must(elastic.NewRangeQuery(name).Lte(rq.Value[0]))
} else if rq.Op == pb.RangeField_GT {
q = q.Must(elastic.NewRangeQuery(name).Gt(rq.Value[0]))
} else { // pb.RangeField_GTE
q = q.Must(elastic.NewRangeQuery(name).Gte(rq.Value[0]))
2021-05-16 05:13:14 +02:00
}
}
return q
2021-05-16 05:13:14 +02:00
}
// AddInvertibleField takes a bool query, an invertible field and a term name
// and adds a term query for that name matching that invertible field.
2021-06-18 06:01:47 +02:00
func AddInvertibleField(q *elastic.BoolQuery, field *pb.InvertibleField, name string) *elastic.BoolQuery {
2021-05-18 12:02:55 +02:00
if field == nil {
return q
}
searchVals := StrArrToInterface(field.Value)
if field.Invert {
2021-06-01 04:19:10 +02:00
q = q.MustNot(elastic.NewTermsQuery(name, searchVals...))
if name == "channel_id.keyword" {
q = q.MustNot(elastic.NewTermsQuery("_id", searchVals...))
}
return q
2021-05-18 12:02:55 +02:00
} else {
return q.Must(elastic.NewTermsQuery(name, searchVals...))
}
}
2021-07-06 02:20:38 +02:00
// recordErrorAndDie is for fatal errors. It takes an error, increments the
// fatal error metric in prometheus and prints a fatal error message.
2021-07-06 02:20:38 +02:00
func (s *Server) recordErrorAndDie(err error) {
metrics.ErrorsCounter.With(prometheus.Labels{"error_type": "fatal"}).Inc()
2021-07-06 02:20:38 +02:00
log.Fatalln(err)
}
2021-05-18 12:02:55 +02:00
// RoundUpReleaseTime take a bool query, a range query and a term name
// and adds a term query for that name (this is for the release time
// field) with the value rounded up.
func RoundUpReleaseTime(q *elastic.BoolQuery, rqs []*pb.RangeField, name string) *elastic.BoolQuery {
if len(rqs) == 0 {
2021-09-15 19:25:19 +02:00
return q
}
for _, rq := range rqs {
releaseTimeInt := rq.Value[0]
if releaseTimeInt < 0 {
releaseTimeInt *= -1
}
releaseTime := ((releaseTimeInt / 360) + 1) * 360
if rq.Op == pb.RangeField_EQ {
q = q.Must(elastic.NewTermQuery(name, releaseTime))
} else if rq.Op == pb.RangeField_LT {
q = q.Must(elastic.NewRangeQuery(name).Lt(releaseTime))
} else if rq.Op == pb.RangeField_LTE {
q = q.Must(elastic.NewRangeQuery(name).Lte(releaseTime))
} else if rq.Op == pb.RangeField_GT {
q = q.Must(elastic.NewRangeQuery(name).Gt(releaseTime))
} else { // pb.RangeField_GTE
q = q.Must(elastic.NewRangeQuery(name).Gte(releaseTime))
}
}
return q
}
// Search /*
// Search logic is as follows:
// 1) Setup query with params given
// 2) Do query with limit of 1000
// 3) remove blocked content (these are returned separately)
// 4) remove duplicates (these are not returned)
// 5) limit claims per channel logic
// 6) get claims referenced by reposts
// 7) get channels references by claims and repost claims
// 8) return streams referenced by repost and all channel referenced in extra_txos
//*/
2021-06-04 07:56:50 +02:00
func (s *Server) Search(ctx context.Context, in *pb.SearchRequest) (*pb.Outputs, error) {
if s.Args.DisableEs {
log.Println("ElasticSearch disable, return nil to search")
return nil, nil
}
metrics.RequestsCount.With(prometheus.Labels{"method": "search"}).Inc()
defer func(t time.Time) {
delta := time.Since(t).Seconds()
metrics.
QueryTime.
With(prometheus.Labels{"method": "search"}).
Observe(delta)
}(time.Now())
2021-06-04 07:56:50 +02:00
2021-06-18 06:01:47 +02:00
var from = 0
var pageSize = 10
var orderBy []orderField
2021-08-24 10:45:30 +02:00
var searchIndices []string
2021-09-30 21:03:00 +02:00
var searchResult *elastic.SearchResult = nil
client := s.EsClient
2021-08-11 05:39:37 +02:00
searchIndices = make([]string, 0, 1)
2021-10-03 05:13:47 +02:00
searchIndices = append(searchIndices, s.Args.EsIndex)
2021-10-01 19:54:03 +02:00
2021-10-03 04:49:49 +02:00
//Code for debugging locally
2021-10-03 05:13:47 +02:00
//indices, _ := client.IndexNames()
//for _, index := range indices {
// if index != "claims" {
// log.Println(index)
// searchIndices = append(searchIndices, index)
// }
//}
2021-10-01 19:54:03 +02:00
2021-10-03 04:49:49 +02:00
// If it's been more than RefreshDelta time since we last checked if the
// es index has been refreshed, we check (this is 2 seconds in prod,
// 0 seconds in debug / unit testing). If the index has been refreshed
// a different number of times since we last checked, we purge the cache
if time.Now().After(s.LastRefreshCheck.Add(s.RefreshDelta)) {
res, err := client.IndexStats(searchIndices[0]).Do(ctx)
if err != nil {
log.Printf("Error on ES index stats\n%v\n", err)
return &pb.Outputs{}, nil
}
2021-10-03 04:49:49 +02:00
numRefreshes := res.Indices[searchIndices[0]].Primaries.Refresh.Total
if numRefreshes != s.NumESRefreshes {
_ = s.QueryCache.Purge()
s.NumESRefreshes = numRefreshes
}
}
2021-10-01 19:54:03 +02:00
var records []*record
cacheKey := s.serializeSearchRequest(in)
2021-06-18 06:01:47 +02:00
2021-10-03 04:49:49 +02:00
setPageVars(in, &pageSize, &from)
2021-09-30 21:03:00 +02:00
/*
2021-11-25 00:24:06 +01:00
cache based on search request params
include from value and number of results.
When another search request comes in with same search params
and same or increased offset (which we currently don't even use?)
that will be a cache hit.
FIXME: For now the cache is turned off when in debugging mode
(for unit tests) because it breaks on some of them.
FIXME: Currently the cache just skips the initial search,
the mgets and post processing are still done. There's probably
a more efficient way to store the final result.
2021-09-30 21:03:00 +02:00
*/
2021-06-18 06:01:47 +02:00
2021-10-03 04:49:49 +02:00
if val, err := s.QueryCache.Get(cacheKey); err != nil {
2021-10-01 19:54:03 +02:00
2021-09-30 21:03:00 +02:00
q := elastic.NewBoolQuery()
2021-06-18 06:01:47 +02:00
2021-09-30 21:03:00 +02:00
err := s.checkQuery(in)
if err != nil {
return nil, err
}
2021-10-03 04:49:49 +02:00
q = s.setupEsQuery(q, in, &orderBy)
2021-06-18 06:01:47 +02:00
2021-09-30 21:03:00 +02:00
fsc := elastic.NewFetchSourceContext(true).Exclude("description", "title")
search := client.Search().
Index(searchIndices...).
FetchSourceContext(fsc).
Query(q). // specify the query
From(0).Size(DefaultSearchSize)
2021-06-18 06:01:47 +02:00
2021-09-30 21:03:00 +02:00
for _, x := range orderBy {
search = search.Sort(x.Field, x.IsAsc)
}
2021-08-13 19:08:25 +02:00
2021-09-30 21:03:00 +02:00
searchResult, err = search.Do(ctx) // execute
if err != nil && elastic.IsNotFound(err) {
log.Println("Index returned 404! Check writer. Index: ", searchIndices)
return &pb.Outputs{}, nil
2021-06-18 06:01:47 +02:00
2021-09-30 21:03:00 +02:00
} else if err != nil {
metrics.ErrorsCounter.With(prometheus.Labels{"error_type": "search"}).Inc()
2021-09-30 21:03:00 +02:00
log.Println("Error executing query: ", err)
return nil, err
}
2021-06-18 06:01:47 +02:00
2021-09-30 21:03:00 +02:00
log.Printf("%s: found %d results in %dms\n", in.Text, len(searchResult.Hits.Hits), searchResult.TookInMillis)
2021-10-01 19:54:03 +02:00
records = s.searchResultToRecords(searchResult)
err = s.QueryCache.Set(cacheKey, records)
if err != nil {
//FIXME: Should this be fatal?
log.Println("Error storing records in cache: ", err)
}
2021-09-30 21:03:00 +02:00
} else {
2021-10-01 19:54:03 +02:00
records = val.([]*record)
2021-09-30 21:03:00 +02:00
}
2021-10-01 19:54:03 +02:00
txos, extraTxos, blocked := s.postProcessResults(ctx, client, records, in, pageSize, from, searchIndices)
2021-06-18 06:01:47 +02:00
2021-09-18 19:21:32 +02:00
if in.NoTotals {
2021-06-18 06:01:47 +02:00
return &pb.Outputs{
Txos: txos,
2021-06-18 06:01:47 +02:00
ExtraTxos: extraTxos,
Offset: uint32(int64(from) + searchResult.TotalHits()),
Blocked: blocked,
2021-06-18 06:01:47 +02:00
}, nil
}
var blockedTotal uint32 = 0
for _, b := range blocked {
blockedTotal += b.Count
}
return &pb.Outputs{
Txos: txos,
ExtraTxos: extraTxos,
Total: uint32(searchResult.TotalHits()),
Offset: uint32(int64(from) + searchResult.TotalHits()),
Blocked: blocked,
2021-06-18 06:01:47 +02:00
BlockedTotal: blockedTotal,
}, nil
}
// normalizeTag takes a string and normalizes it for search in es.
2021-06-18 06:01:47 +02:00
func (s *Server) normalizeTag(tag string) string {
c := cases.Lower(language.English)
res := s.MultiSpaceRe.ReplaceAll(
s.WeirdCharsRe.ReplaceAll(
[]byte(strings.TrimSpace(strings.Replace(c.String(tag), "'", "", -1))),
[]byte(" ")),
[]byte(" "))
return string(res)
}
// cleanTags takes an array of tags and normalizes them.
2021-06-18 06:01:47 +02:00
func (s *Server) cleanTags(tags []string) []string {
cleanedTags := make([]string, len(tags))
for i, tag := range tags {
cleanedTags[i] = s.normalizeTag(tag)
}
return cleanedTags
}
// searchResultToRecords takes an elastic.SearchResult object and converts
// them to internal record structures.
2021-10-01 19:54:03 +02:00
func (s *Server) searchResultToRecords(
searchResult *elastic.SearchResult) []*record {
records := make([]*record, 0, searchResult.TotalHits())
var r record
for _, item := range searchResult.Each(reflect.TypeOf(r)) {
if t, ok := item.(record); ok {
records = append(records, &t)
}
}
return records
}
// postProcessResults takes es search result records and runs our
// post processing on them.
// TODO: more in depth description.
2021-06-18 06:01:47 +02:00
func (s *Server) postProcessResults(
ctx context.Context,
client *elastic.Client,
2021-10-01 19:54:03 +02:00
records []*record,
2021-06-18 06:01:47 +02:00
in *pb.SearchRequest,
pageSize int,
from int,
2021-10-01 19:54:03 +02:00
searchIndices []string) ([]*pb.Output, []*pb.Output, []*pb.Blocked) {
2021-06-18 06:01:47 +02:00
var txos []*pb.Output
var blockedRecords []*record
var blocked []*pb.Blocked
var blockedMap map[string]*pb.Blocked
//printJsonFullResults(searchResult)
records, blockedRecords, blockedMap = removeBlocked(records)
2021-08-12 09:19:25 +02:00
if in.RemoveDuplicates {
2021-06-18 06:01:47 +02:00
records = removeDuplicates(records)
}
2021-08-12 09:19:25 +02:00
if in.LimitClaimsPerChannel > 0 {
records = searchAhead(records, pageSize, int(in.LimitClaimsPerChannel))
2021-06-18 06:01:47 +02:00
}
finalLength := int(math.Min(float64(len(records)), float64(pageSize)))
txos = make([]*pb.Output, 0, finalLength)
var j = 0
for i := from; i < from+finalLength && i < len(records) && j < finalLength; i++ {
2021-06-18 06:01:47 +02:00
t := records[i]
res := t.recordToOutput()
txos = append(txos, res)
j += 1
}
//Get claims for reposts
2021-07-06 02:20:38 +02:00
repostClaims, repostRecords, repostedMap := s.getClaimsForReposts(ctx, client, records, searchIndices)
2021-06-18 06:01:47 +02:00
//get all unique channels
2021-07-06 02:20:38 +02:00
channels, channelMap := s.getUniqueChannels(append(append(records, repostRecords...), blockedRecords...), client, ctx, searchIndices)
2021-06-18 06:01:47 +02:00
//add these to extra txos
extraTxos := append(repostClaims, channels...)
//Fill in channel / repost data for txos and blocked
for i, txo := range txos {
channel, cOk := channelMap[records[i].ChannelId]
repostClaim, rOk := repostedMap[records[i].RepostedClaimId]
if cOk {
txo.GetClaim().Channel = channel
}
if rOk {
txo.GetClaim().Repost = repostClaim
}
}
blocked = make([]*pb.Blocked, 0, len(blockedMap))
for k, v := range blockedMap {
if channel, ok := channelMap[k]; ok {
v.Channel = channel
}
blocked = append(blocked, v)
}
return txos, extraTxos, blocked
}
// checkQuery takes a search request and does a sanity check on it for
// validity.
func (s *Server) checkQuery(in *pb.SearchRequest) error {
limit := 2048
checks := map[string]bool{
"claim_ids": in.ClaimId != nil && !in.ClaimId.Invert && len(in.ClaimId.Value) > limit,
"not_claim_ids": in.ClaimId != nil && in.ClaimId.Invert && len(in.ClaimId.Value) > limit,
"channel_ids": in.ChannelId != nil && !in.ChannelId.Invert && len(in.ChannelId.Value) > limit,
"not_channel_ids": in.ChannelId != nil && in.ChannelId.Invert && len(in.ChannelId.Value) > limit,
"not_tags": len(in.NotTags) > limit,
"all_tags": len(in.AllTags) > limit,
"any_tags": len(in.AnyTags) > limit,
"any_languages": len(in.AnyLanguages) > limit,
}
for name, failed := range checks {
if failed {
time.Sleep(time.Second * 2) // throttle
2021-10-03 04:49:49 +02:00
return fmt.Errorf("%s cant have more than %d items.", name, limit)
}
}
return nil
}
// setPageVars takes a search request and pointers to the local pageSize
// and from variables and sets them from the struct.
2021-10-03 04:49:49 +02:00
func setPageVars(in *pb.SearchRequest, pageSize *int, from *int) {
if in.Limit > 0 {
log.Printf("############ limit: %d\n", in.Limit)
*pageSize = int(in.Limit)
}
if in.Offset > 0 {
*from = int(in.Offset)
}
}
// setupEsQuery takes an elastic.BoolQuery, pb.SearchRequest and orderField
// and adds the search request terms to the bool query.
2021-06-18 06:01:47 +02:00
func (s *Server) setupEsQuery(
q *elastic.BoolQuery,
in *pb.SearchRequest,
orderBy *[]orderField) *elastic.BoolQuery {
claimTypes := map[string]int{
"stream": 1,
"channel": 2,
"repost": 3,
"collection": 4,
}
2021-05-18 12:02:55 +02:00
streamTypes := map[string]int{
"video": 1,
"audio": 2,
"image": 3,
2021-05-18 12:02:55 +02:00
"document": 4,
"binary": 5,
"model": 6,
2021-05-18 12:02:55 +02:00
}
replacements := map[string]string{
2021-11-25 00:24:06 +01:00
"name": "normalized_name",
"normalized": "normalized_name",
"claim_name": "normalized_name",
"txid": "tx_id",
"nout": "tx_nout",
"reposted": "repost_count",
"valid_channel_signature": "is_signature_valid",
2021-11-25 00:24:06 +01:00
"claim_id": "_id",
"signature_digest": "signature",
2021-05-18 12:02:55 +02:00
}
textFields := map[string]bool{
"author": true,
"canonical_url": true,
"channel_id": true,
"claim_name": true,
"description": true,
"claim_id": true,
"media_type": true,
"normalized_name": true,
"public_key_bytes": true,
"public_key_id": true,
"short_url": true,
"signature": true,
"stream_type": true,
"title": true,
"tx_id": true,
"fee_currency": true,
2021-05-18 12:02:55 +02:00
"reposted_claim_id": true,
"tags": true,
2021-05-18 12:02:55 +02:00
}
2021-08-12 09:19:25 +02:00
if in.IsControlling {
q = q.Must(elastic.NewTermQuery("is_controlling", in.IsControlling))
2021-05-31 20:53:08 +02:00
}
2021-08-12 09:19:25 +02:00
if len(in.ClaimName) > 0 {
in.NormalizedName = internal.NormalizeName(in.ClaimName)
}
2021-05-18 12:02:55 +02:00
if len(in.OrderBy) > 0 {
for _, x := range in.OrderBy {
var toAppend string
2021-06-06 06:57:37 +02:00
var isAsc = false
2021-05-18 12:02:55 +02:00
if x[0] == '^' {
2021-06-06 06:57:37 +02:00
isAsc = true
2021-05-18 12:02:55 +02:00
x = x[1:]
}
if _, ok := replacements[x]; ok {
toAppend = replacements[x]
2021-05-31 20:53:08 +02:00
} else {
toAppend = x
2021-05-18 12:02:55 +02:00
}
2021-05-31 20:53:08 +02:00
if _, ok := textFields[toAppend]; ok {
toAppend = toAppend + ".keyword"
2021-05-18 12:02:55 +02:00
}
2021-06-18 06:01:47 +02:00
*orderBy = append(*orderBy, orderField{toAppend, isAsc})
2021-06-06 06:57:37 +02:00
}
2021-05-18 12:02:55 +02:00
}
if len(in.ClaimType) > 0 {
searchVals := make([]interface{}, len(in.ClaimType))
for i := 0; i < len(in.ClaimType); i++ {
searchVals[i] = claimTypes[in.ClaimType[i]]
}
q = q.Must(elastic.NewTermsQuery("claim_type", searchVals...))
}
2021-05-18 12:02:55 +02:00
if len(in.StreamType) > 0 {
searchVals := make([]interface{}, len(in.StreamType))
for i := 0; i < len(in.StreamType); i++ {
searchVals[i] = streamTypes[in.StreamType[i]]
}
2021-06-01 04:19:10 +02:00
q = q.Must(elastic.NewTermsQuery("stream_type", searchVals...))
2021-05-18 12:02:55 +02:00
}
if in.SdHash != "" {
q.Must(elastic.NewPrefixQuery("sd_hash.keyword", in.SdHash))
}
2021-05-18 12:02:55 +02:00
if in.ClaimId != nil {
searchVals := StrArrToInterface(in.ClaimId.Value)
if len(in.ClaimId.Value) == 1 && len(in.ClaimId.Value[0]) < 20 {
if in.ClaimId.Invert {
q = q.MustNot(elastic.NewPrefixQuery("claim_id.keyword", in.ClaimId.Value[0]))
} else {
q = q.Must(elastic.NewPrefixQuery("claim_id.keyword", in.ClaimId.Value[0]))
}
2021-05-16 05:13:14 +02:00
} else {
2021-05-18 12:02:55 +02:00
if in.ClaimId.Invert {
q = q.MustNot(elastic.NewTermsQuery("claim_id.keyword", searchVals...))
} else {
q = q.Must(elastic.NewTermsQuery("claim_id.keyword", searchVals...))
}
2021-05-16 05:13:14 +02:00
}
}
if in.PublicKeyId != "" {
2021-10-04 18:44:55 +02:00
q = q.Must(elastic.NewTermQuery("public_key_id.keyword", in.PublicKeyId))
2021-05-16 05:13:14 +02:00
}
2021-08-12 09:19:25 +02:00
if in.HasChannelSignature {
q = q.Must(elastic.NewExistsQuery("signature"))
2021-07-06 02:20:38 +02:00
if in.IsSignatureValid != nil {
q = q.Must(elastic.NewTermQuery("is_signature_valid", in.IsSignatureValid.Value))
2021-05-18 12:02:55 +02:00
}
2021-07-06 02:20:38 +02:00
} else if in.IsSignatureValid != nil {
2021-05-18 12:02:55 +02:00
q = q.MinimumNumberShouldMatch(1)
q = q.Should(elastic.NewBoolQuery().MustNot(elastic.NewExistsQuery("signature")))
2021-07-06 02:20:38 +02:00
q = q.Should(elastic.NewTermQuery("is_signature_valid", in.IsSignatureValid.Value))
2021-05-18 12:02:55 +02:00
}
2021-05-31 20:53:08 +02:00
if in.HasSource != nil {
2021-05-18 12:02:55 +02:00
q = q.MinimumNumberShouldMatch(1)
2021-05-31 20:53:08 +02:00
isStreamOrRepost := elastic.NewTermsQuery("claim_type", claimTypes["stream"], claimTypes["repost"])
q = q.Should(elastic.NewBoolQuery().Must(isStreamOrRepost, elastic.NewMatchQuery("has_source", in.HasSource.Value)))
q = q.Should(elastic.NewBoolQuery().MustNot(isStreamOrRepost))
2021-05-18 12:02:55 +02:00
q = q.Should(elastic.NewBoolQuery().Must(elastic.NewTermQuery("reposted_claim_type", claimTypes["channel"])))
}
2021-05-31 20:53:08 +02:00
if in.TxNout != nil {
q = q.Must(elastic.NewTermQuery("tx_nout", in.TxNout.Value))
}
2021-05-18 12:02:55 +02:00
2021-08-12 09:19:25 +02:00
q = AddTermField(q, in.Author, "author.keyword")
q = AddTermField(q, in.Title, "title.keyword")
q = AddTermField(q, in.CanonicalUrl, "canonical_url.keyword")
q = AddTermField(q, in.ClaimName, "claim_name.keyword")
q = AddTermField(q, in.Description, "description.keyword")
2021-06-18 06:01:47 +02:00
q = AddTermsField(q, in.MediaType, "media_type.keyword")
2021-08-12 09:19:25 +02:00
q = AddTermField(q, in.NormalizedName, "normalized_name.keyword")
q = AddTermField(q, in.ShortUrl, "short_url.keyword")
q = AddTermField(q, in.Signature, "signature.keyword")
q = AddTermField(q, in.TxId, "tx_id.keyword")
2021-10-04 18:44:55 +02:00
q = AddTermField(q, strings.ToUpper(in.FeeCurrency), "fee_currency.keyword")
2021-08-12 09:19:25 +02:00
q = AddTermField(q, in.RepostedClaimId, "reposted_claim_id.keyword")
2021-06-18 06:01:47 +02:00
q = AddTermsField(q, s.cleanTags(in.AnyTags), "tags.keyword")
q = AddIndividualTermFields(q, s.cleanTags(in.AllTags), "tags.keyword", false)
q = AddIndividualTermFields(q, s.cleanTags(in.NotTags), "tags.keyword", true)
q = AddTermsField(q, in.AnyLanguages, "languages")
q = AddIndividualTermFields(q, in.AllLanguages, "languages", false)
q = AddInvertibleField(q, in.ChannelId, "channel_id.keyword")
q = AddRangeField(q, in.TxPosition, "tx_position")
q = AddRangeField(q, in.Amount, "amount")
q = AddRangeField(q, in.Timestamp, "timestamp")
q = AddRangeField(q, in.CreationTimestamp, "creation_timestamp")
q = AddRangeField(q, in.Height, "height")
q = AddRangeField(q, in.CreationHeight, "creation_height")
q = AddRangeField(q, in.ActivationHeight, "activation_height")
q = AddRangeField(q, in.ExpirationHeight, "expiration_height")
q = RoundUpReleaseTime(q, in.ReleaseTime, "release_time")
2021-07-06 02:20:38 +02:00
q = AddRangeField(q, in.RepostCount, "repost_count")
2021-06-18 06:01:47 +02:00
q = AddRangeField(q, in.FeeAmount, "fee_amount")
q = AddRangeField(q, in.Duration, "duration")
q = AddRangeField(q, in.CensorType, "censor_type")
q = AddRangeField(q, in.EffectiveAmount, "effective_amount")
q = AddRangeField(q, in.SupportAmount, "support_amount")
q = AddRangeField(q, in.TrendingScore, "trending_score")
2021-05-16 05:13:14 +02:00
2021-05-31 20:53:08 +02:00
if in.Text != "" {
textQuery := elastic.NewSimpleQueryStringQuery(in.Text).
FieldWithBoost("claim_name", 4).
FieldWithBoost("channel_name", 8).
FieldWithBoost("title", 1).
FieldWithBoost("description", 0.5).
FieldWithBoost("author", 1).
FieldWithBoost("tags", 0.5)
q = q.Must(textQuery)
}
2021-06-18 06:01:47 +02:00
return q
2021-04-19 21:25:34 +02:00
}
// getUniqueChannels takes the record results from the es search and returns
// the unique channels from those records as a list and a map.
2021-07-06 02:20:38 +02:00
func (s *Server) getUniqueChannels(records []*record, client *elastic.Client, ctx context.Context, searchIndices []string) ([]*pb.Output, map[string]*pb.Output) {
2021-06-12 00:18:17 +02:00
channels := make(map[string]*pb.Output)
channelsSet := make(map[string]bool)
var mget = client.Mget()
var totalChannels = 0
for _, r := range records {
2021-06-17 18:15:05 +02:00
for _, searchIndex := range searchIndices {
if r.ChannelId != "" && !channelsSet[r.ChannelId] {
channelsSet[r.ChannelId] = true
nmget := elastic.NewMultiGetItem().Id(r.ChannelId).Index(searchIndex)
mget = mget.Add(nmget)
totalChannels++
}
2021-07-06 02:20:38 +02:00
if r.CensorType != 0 && !channelsSet[r.CensoringChannelId] {
channelsSet[r.CensoringChannelId] = true
nmget := elastic.NewMultiGetItem().Id(r.CensoringChannelId).Index(searchIndex)
2021-06-17 18:15:05 +02:00
mget = mget.Add(nmget)
totalChannels++
}
2021-06-12 00:18:17 +02:00
}
}
if totalChannels == 0 {
return []*pb.Output{}, make(map[string]*pb.Output)
}
res, err := mget.Do(ctx)
if err != nil {
metrics.ErrorsCounter.With(prometheus.Labels{"error_type": "get_unique_channels"}).Inc()
log.Println(err)
2021-06-12 00:18:17 +02:00
return []*pb.Output{}, make(map[string]*pb.Output)
}
channelTxos := make([]*pb.Output, totalChannels)
//repostedRecords := make([]*record, totalReposted)
2021-07-06 02:20:38 +02:00
//log.Println("total channel", totalChannels)
2021-06-12 00:18:17 +02:00
for i, doc := range res.Docs {
var r record
err := json.Unmarshal(doc.Source, &r)
if err != nil {
metrics.ErrorsCounter.With(prometheus.Labels{"error_type": "json"}).Inc()
log.Println(err)
2021-06-12 00:18:17 +02:00
return []*pb.Output{}, make(map[string]*pb.Output)
}
2021-06-12 00:18:17 +02:00
channelTxos[i] = r.recordToOutput()
channels[r.ClaimId] = channelTxos[i]
//log.Println(r)
//repostedRecords[i] = &r
}
2021-06-12 00:18:17 +02:00
return channelTxos, channels
}
// getClaimsForReposts takes the record results from the es query and returns
// an array and map of the reposted records as well as an array of those
// records.
2021-09-30 21:03:00 +02:00
func (s *Server) getClaimsForReposts(ctx context.Context, client *elastic.Client, records []*record, searchIndices []string) ([]*pb.Output, []*record, map[string]*pb.Output) {
var totalReposted = 0
var mget = client.Mget() //.StoredFields("_id")
2021-06-12 00:18:17 +02:00
/*
var nmget = elastic.NewMultiGetItem()
for _, index := range searchIndices {
nmget = nmget.Index(index)
}
*/
for _, r := range records {
2021-06-17 18:15:05 +02:00
for _, searchIndex := range searchIndices {
if r.RepostedClaimId != "" {
var nmget = elastic.NewMultiGetItem().Id(r.RepostedClaimId).Index(searchIndex)
//nmget = nmget.Id(r.RepostedClaimId)
mget = mget.Add(nmget)
totalReposted++
}
}
}
2021-06-12 00:18:17 +02:00
//mget = mget.Add(nmget)
if totalReposted == 0 {
return []*pb.Output{}, []*record{}, make(map[string]*pb.Output)
}
res, err := mget.Do(ctx)
if err != nil {
metrics.ErrorsCounter.With(prometheus.Labels{"error_type": "mget"}).Inc()
log.Println(err)
2021-06-12 00:18:17 +02:00
return []*pb.Output{}, []*record{}, make(map[string]*pb.Output)
}
claims := make([]*pb.Output, totalReposted)
repostedRecords := make([]*record, totalReposted)
2021-06-12 00:18:17 +02:00
respostedMap := make(map[string]*pb.Output)
2021-07-06 02:20:38 +02:00
//log.Println("reposted records", totalReposted)
for i, doc := range res.Docs {
var r record
err := json.Unmarshal(doc.Source, &r)
if err != nil {
metrics.ErrorsCounter.With(prometheus.Labels{"error_type": "json"}).Inc()
log.Println(err)
2021-06-12 00:18:17 +02:00
return []*pb.Output{}, []*record{}, make(map[string]*pb.Output)
}
claims[i] = r.recordToOutput()
repostedRecords[i] = &r
2021-06-12 00:18:17 +02:00
respostedMap[r.ClaimId] = claims[i]
}
2021-06-12 00:18:17 +02:00
return claims, repostedRecords, respostedMap
}
// serializeSearchRequest takes a search request and serializes it into a key
// for use in the internal cache for the hub.
2021-10-01 19:54:03 +02:00
func (s *Server) serializeSearchRequest(request *pb.SearchRequest) string {
2021-10-03 04:49:49 +02:00
// Save the offest / limit and set to zero, cache hits should happen regardless
// and they're used in post processing
//offset, limit := request.Offset, request.Limit
//request.Offset = 0
//request.Limit = 0
2021-10-01 19:54:03 +02:00
bytes, err := protojson.Marshal(request)
if err != nil {
return ""
}
str := string((*s.S256).Sum(bytes))
2021-10-03 04:49:49 +02:00
// log.Println(str)
//request.Offset = offset
//request.Limit = limit
2021-10-01 19:54:03 +02:00
return str
}
// searchAhead takes an array of record results, the pageSize and
// perChannelPerPage value and returns the hits for this page.
func searchAhead(searchHits []*record, pageSize int, perChannelPerPage int) []*record {
finalHits := make([]*record, 0, len(searchHits))
var channelCounters map[string]int
channelCounters = make(map[string]int)
nextPageHitsMaybeCheckLater := deque.New()
searchHitsQ := deque.New()
for _, rec := range searchHits {
searchHitsQ.PushRight(rec)
}
for !searchHitsQ.Empty() || !nextPageHitsMaybeCheckLater.Empty() {
if len(finalHits) > 0 && len(finalHits)%pageSize == 0 {
channelCounters = make(map[string]int)
} else if len(finalHits) != 0 {
// means last page was incomplete and we are left with bad replacements
break
}
for i := 0; i < nextPageHitsMaybeCheckLater.Size(); i++ {
rec := nextPageHitsMaybeCheckLater.PopLeft().(*record)
if perChannelPerPage > 0 && channelCounters[rec.ChannelId] < perChannelPerPage {
finalHits = append(finalHits, rec)
channelCounters[rec.ChannelId] = channelCounters[rec.ChannelId] + 1
}
}
for !searchHitsQ.Empty() {
hit := searchHitsQ.PopLeft().(*record)
if hit.ChannelId == "" || perChannelPerPage < 0 {
finalHits = append(finalHits, hit)
} else if channelCounters[hit.ChannelId] < perChannelPerPage {
finalHits = append(finalHits, hit)
channelCounters[hit.ChannelId] = channelCounters[hit.ChannelId] + 1
if len(finalHits)%pageSize == 0 {
break
}
} else {
nextPageHitsMaybeCheckLater.PushRight(hit)
}
}
}
return finalHits
}
2021-06-09 04:07:59 +02:00
// recordToOutput is a function on a record struct to turn it into a pb.Output
// struct.
func (r *record) recordToOutput() *pb.Output {
return &pb.Output{
TxHash: internal.TxIdToTxHash(r.Txid),
Nout: r.Nout,
Height: r.Height,
Meta: &pb.Output_Claim{
Claim: &pb.ClaimMeta{
//Channel:
//Repost:
ShortUrl: r.ShortUrl,
CanonicalUrl: r.CanonicalUrl,
IsControlling: r.IsControlling,
TakeOverHeight: r.TakeOverHeight,
CreationHeight: r.CreationHeight,
ActivationHeight: r.ActivationHeight,
ExpirationHeight: r.ExpirationHeight,
ClaimsInChannel: r.ClaimsInChannel,
2021-07-06 02:20:38 +02:00
Reposted: r.RepostCount,
EffectiveAmount: r.EffectiveAmount,
SupportAmount: r.SupportAmount,
TrendingScore: r.TrendingScore,
},
},
}
}
// getHitId is a function on the record struct to get the id for the search
// hit.
2021-06-09 04:07:59 +02:00
func (r *record) getHitId() string {
if r.RepostedClaimId != "" {
return r.RepostedClaimId
} else {
return r.ClaimId
}
}
// removeDuplicates takes an array of record results and remove duplicates.
2021-06-09 04:07:59 +02:00
func removeDuplicates(searchHits []*record) []*record {
dropped := make(map[*record]bool)
// claim_id -> (creation_height, hit_id), where hit_id is either reposted claim id or original
knownIds := make(map[string]*record)
for _, hit := range searchHits {
hitHeight := hit.Height
hitId := hit.getHitId()
if knownIds[hitId] == nil {
knownIds[hitId] = hit
} else {
prevHit := knownIds[hitId]
if hitHeight < prevHit.Height {
knownIds[hitId] = hit
2021-06-09 04:07:59 +02:00
dropped[prevHit] = true
} else {
dropped[hit] = true
}
}
}
deduped := make([]*record, len(searchHits)-len(dropped))
2021-06-09 04:07:59 +02:00
var i = 0
for _, hit := range searchHits {
if !dropped[hit] {
deduped[i] = hit
i++
}
}
return deduped
}
2021-06-10 02:04:06 +02:00
// removeBlocked takes an array of record results from the es search
// and removes blocked records.
2021-06-12 00:18:17 +02:00
func removeBlocked(searchHits []*record) ([]*record, []*record, map[string]*pb.Blocked) {
2021-06-10 02:04:06 +02:00
newHits := make([]*record, 0, len(searchHits))
2021-06-12 00:18:17 +02:00
blockedHits := make([]*record, 0, len(searchHits))
2021-06-10 02:04:06 +02:00
blockedChannels := make(map[string]*pb.Blocked)
for _, r := range searchHits {
if r.CensorType != 0 {
2021-07-06 02:20:38 +02:00
if blockedChannels[r.CensoringChannelId] == nil {
2021-06-10 02:04:06 +02:00
blockedObj := &pb.Blocked{
Count: 1,
2021-06-12 00:18:17 +02:00
Channel: nil,
2021-06-10 02:04:06 +02:00
}
2021-07-06 02:20:38 +02:00
blockedChannels[r.CensoringChannelId] = blockedObj
2021-06-12 00:18:17 +02:00
blockedHits = append(blockedHits, r)
2021-06-10 02:04:06 +02:00
} else {
2021-07-06 02:20:38 +02:00
blockedChannels[r.CensoringChannelId].Count += 1
2021-06-10 02:04:06 +02:00
}
} else {
newHits = append(newHits, r)
}
}
2021-06-12 00:18:17 +02:00
return newHits, blockedHits, blockedChannels
}