mirror of
https://codeberg.org/forgejo/forgejo.git
synced 2024-11-30 09:41:11 -05:00
12a1f914f4
* update github.com/alecthomas/chroma v0.8.0 -> v0.8.1 * github.com/blevesearch/bleve v1.0.10 -> v1.0.12 * editorconfig-core-go v2.1.1 -> v2.3.7 * github.com/gliderlabs/ssh v0.2.2 -> v0.3.1 * migrate editorconfig.ParseBytes to Parse * github.com/shurcooL/vfsgen to 0d455de96546 * github.com/go-git/go-git/v5 v5.1.0 -> v5.2.0 * github.com/google/uuid v1.1.1 -> v1.1.2 * github.com/huandu/xstrings v1.3.0 -> v1.3.2 * github.com/klauspost/compress v1.10.11 -> v1.11.1 * github.com/markbates/goth v1.61.2 -> v1.65.0 * github.com/mattn/go-sqlite3 v1.14.0 -> v1.14.4 * github.com/mholt/archiver v3.3.0 -> v3.3.2 * github.com/microcosm-cc/bluemonday 4f7140c49acb -> v1.0.4 * github.com/minio/minio-go v7.0.4 -> v7.0.5 * github.com/olivere/elastic v7.0.9 -> v7.0.20 * github.com/urfave/cli v1.20.0 -> v1.22.4 * github.com/prometheus/client_golang v1.1.0 -> v1.8.0 * github.com/xanzy/go-gitlab v0.37.0 -> v0.38.1 * mvdan.cc/xurls v2.1.0 -> v2.2.0 Co-authored-by: Lauris BH <lauris@nix.lv>
745 lines
20 KiB
Go
Vendored
745 lines
20 KiB
Go
Vendored
// Copyright 2012-present Oliver Eilhard. All rights reserved.
|
|
// Use of this source code is governed by a MIT-license.
|
|
// See http://olivere.mit-license.org/license.txt for details.
|
|
|
|
package elastic
|
|
|
|
import (
|
|
"context"
|
|
"errors"
|
|
"fmt"
|
|
"net/http"
|
|
"net/url"
|
|
"strings"
|
|
)
|
|
|
|
// ReindexService is a method to copy documents from one index to another.
|
|
// It is documented at https://www.elastic.co/guide/en/elasticsearch/reference/7.0/docs-reindex.html.
|
|
type ReindexService struct {
|
|
client *Client
|
|
|
|
pretty *bool // pretty format the returned JSON response
|
|
human *bool // return human readable values for statistics
|
|
errorTrace *bool // include the stack trace of returned errors
|
|
filterPath []string // list of filters used to reduce the response
|
|
headers http.Header // custom request-level HTTP headers
|
|
|
|
refresh string
|
|
timeout string
|
|
waitForActiveShards string
|
|
waitForCompletion *bool
|
|
requestsPerSecond *int
|
|
slices interface{}
|
|
body interface{}
|
|
source *ReindexSource
|
|
destination *ReindexDestination
|
|
conflicts string
|
|
size *int
|
|
script *Script
|
|
}
|
|
|
|
// NewReindexService creates a new ReindexService.
|
|
func NewReindexService(client *Client) *ReindexService {
|
|
return &ReindexService{
|
|
client: client,
|
|
}
|
|
}
|
|
|
|
// Pretty tells Elasticsearch whether to return a formatted JSON response.
|
|
func (s *ReindexService) Pretty(pretty bool) *ReindexService {
|
|
s.pretty = &pretty
|
|
return s
|
|
}
|
|
|
|
// Human specifies whether human readable values should be returned in
|
|
// the JSON response, e.g. "7.5mb".
|
|
func (s *ReindexService) Human(human bool) *ReindexService {
|
|
s.human = &human
|
|
return s
|
|
}
|
|
|
|
// ErrorTrace specifies whether to include the stack trace of returned errors.
|
|
func (s *ReindexService) ErrorTrace(errorTrace bool) *ReindexService {
|
|
s.errorTrace = &errorTrace
|
|
return s
|
|
}
|
|
|
|
// FilterPath specifies a list of filters used to reduce the response.
|
|
func (s *ReindexService) FilterPath(filterPath ...string) *ReindexService {
|
|
s.filterPath = filterPath
|
|
return s
|
|
}
|
|
|
|
// Header adds a header to the request.
|
|
func (s *ReindexService) Header(name string, value string) *ReindexService {
|
|
if s.headers == nil {
|
|
s.headers = http.Header{}
|
|
}
|
|
s.headers.Add(name, value)
|
|
return s
|
|
}
|
|
|
|
// Headers specifies the headers of the request.
|
|
func (s *ReindexService) Headers(headers http.Header) *ReindexService {
|
|
s.headers = headers
|
|
return s
|
|
}
|
|
|
|
// WaitForActiveShards sets the number of shard copies that must be active before
|
|
// proceeding with the reindex operation. Defaults to 1, meaning the primary shard only.
|
|
// Set to `all` for all shard copies, otherwise set to any non-negative value less than or
|
|
// equal to the total number of copies for the shard (number of replicas + 1).
|
|
func (s *ReindexService) WaitForActiveShards(waitForActiveShards string) *ReindexService {
|
|
s.waitForActiveShards = waitForActiveShards
|
|
return s
|
|
}
|
|
|
|
// RequestsPerSecond specifies the throttle to set on this request in sub-requests per second.
|
|
// -1 means set no throttle as does "unlimited" which is the only non-float this accepts.
|
|
func (s *ReindexService) RequestsPerSecond(requestsPerSecond int) *ReindexService {
|
|
s.requestsPerSecond = &requestsPerSecond
|
|
return s
|
|
}
|
|
|
|
// Slices specifies the number of slices this task should be divided into. Defaults to 1.
|
|
// It used to be a number, but can be set to "auto" as of 6.7.
|
|
//
|
|
// See https://www.elastic.co/guide/en/elasticsearch/reference/7.0/docs-reindex.html#docs-reindex-slice
|
|
// for details.
|
|
func (s *ReindexService) Slices(slices interface{}) *ReindexService {
|
|
s.slices = slices
|
|
return s
|
|
}
|
|
|
|
// Refresh indicates whether Elasticsearch should refresh the effected indexes
|
|
// immediately.
|
|
//
|
|
// See https://www.elastic.co/guide/en/elasticsearch/reference/7.0/docs-refresh.html
|
|
// for details.
|
|
func (s *ReindexService) Refresh(refresh string) *ReindexService {
|
|
s.refresh = refresh
|
|
return s
|
|
}
|
|
|
|
// Timeout is the time each individual bulk request should wait for shards
|
|
// that are unavailable.
|
|
func (s *ReindexService) Timeout(timeout string) *ReindexService {
|
|
s.timeout = timeout
|
|
return s
|
|
}
|
|
|
|
// WaitForCompletion indicates whether Elasticsearch should block until the
|
|
// reindex is complete.
|
|
func (s *ReindexService) WaitForCompletion(waitForCompletion bool) *ReindexService {
|
|
s.waitForCompletion = &waitForCompletion
|
|
return s
|
|
}
|
|
|
|
// Source specifies the source of the reindexing process.
|
|
func (s *ReindexService) Source(source *ReindexSource) *ReindexService {
|
|
s.source = source
|
|
return s
|
|
}
|
|
|
|
// SourceIndex specifies the source index of the reindexing process.
|
|
func (s *ReindexService) SourceIndex(index string) *ReindexService {
|
|
if s.source == nil {
|
|
s.source = NewReindexSource()
|
|
}
|
|
s.source = s.source.Index(index)
|
|
return s
|
|
}
|
|
|
|
// Destination specifies the destination of the reindexing process.
|
|
func (s *ReindexService) Destination(destination *ReindexDestination) *ReindexService {
|
|
s.destination = destination
|
|
return s
|
|
}
|
|
|
|
// DestinationIndex specifies the destination index of the reindexing process.
|
|
func (s *ReindexService) DestinationIndex(index string) *ReindexService {
|
|
if s.destination == nil {
|
|
s.destination = NewReindexDestination()
|
|
}
|
|
s.destination = s.destination.Index(index)
|
|
return s
|
|
}
|
|
|
|
// DestinationIndexAndType specifies both the destination index and type
|
|
// of the reindexing process.
|
|
func (s *ReindexService) DestinationIndexAndType(index, typ string) *ReindexService {
|
|
if s.destination == nil {
|
|
s.destination = NewReindexDestination()
|
|
}
|
|
s.destination = s.destination.Index(index)
|
|
s.destination = s.destination.Type(typ)
|
|
return s
|
|
}
|
|
|
|
// Conflicts indicates what to do when the process detects version conflicts.
|
|
// Possible values are "proceed" and "abort".
|
|
func (s *ReindexService) Conflicts(conflicts string) *ReindexService {
|
|
s.conflicts = conflicts
|
|
return s
|
|
}
|
|
|
|
// AbortOnVersionConflict aborts the request on version conflicts.
|
|
// It is an alias to setting Conflicts("abort").
|
|
func (s *ReindexService) AbortOnVersionConflict() *ReindexService {
|
|
s.conflicts = "abort"
|
|
return s
|
|
}
|
|
|
|
// ProceedOnVersionConflict aborts the request on version conflicts.
|
|
// It is an alias to setting Conflicts("proceed").
|
|
func (s *ReindexService) ProceedOnVersionConflict() *ReindexService {
|
|
s.conflicts = "proceed"
|
|
return s
|
|
}
|
|
|
|
// Size sets an upper limit for the number of processed documents.
|
|
func (s *ReindexService) Size(size int) *ReindexService {
|
|
s.size = &size
|
|
return s
|
|
}
|
|
|
|
// Script allows for modification of the documents as they are reindexed
|
|
// from source to destination.
|
|
func (s *ReindexService) Script(script *Script) *ReindexService {
|
|
s.script = script
|
|
return s
|
|
}
|
|
|
|
// Body specifies the body of the request to send to Elasticsearch.
|
|
// It overrides settings specified with other setters, e.g. Query.
|
|
func (s *ReindexService) Body(body interface{}) *ReindexService {
|
|
s.body = body
|
|
return s
|
|
}
|
|
|
|
// buildURL builds the URL for the operation.
|
|
func (s *ReindexService) buildURL() (string, url.Values, error) {
|
|
// Build URL path
|
|
path := "/_reindex"
|
|
|
|
// Add query string parameters
|
|
params := url.Values{}
|
|
if v := s.pretty; v != nil {
|
|
params.Set("pretty", fmt.Sprint(*v))
|
|
}
|
|
if v := s.human; v != nil {
|
|
params.Set("human", fmt.Sprint(*v))
|
|
}
|
|
if v := s.errorTrace; v != nil {
|
|
params.Set("error_trace", fmt.Sprint(*v))
|
|
}
|
|
if len(s.filterPath) > 0 {
|
|
params.Set("filter_path", strings.Join(s.filterPath, ","))
|
|
}
|
|
if s.refresh != "" {
|
|
params.Set("refresh", s.refresh)
|
|
}
|
|
if s.timeout != "" {
|
|
params.Set("timeout", s.timeout)
|
|
}
|
|
if s.requestsPerSecond != nil {
|
|
params.Set("requests_per_second", fmt.Sprintf("%v", *s.requestsPerSecond))
|
|
}
|
|
if s.slices != nil {
|
|
params.Set("slices", fmt.Sprintf("%v", s.slices))
|
|
}
|
|
if s.waitForActiveShards != "" {
|
|
params.Set("wait_for_active_shards", s.waitForActiveShards)
|
|
}
|
|
if s.waitForCompletion != nil {
|
|
params.Set("wait_for_completion", fmt.Sprintf("%v", *s.waitForCompletion))
|
|
}
|
|
return path, params, nil
|
|
}
|
|
|
|
// Validate checks if the operation is valid.
|
|
func (s *ReindexService) Validate() error {
|
|
var invalid []string
|
|
if s.body != nil {
|
|
return nil
|
|
}
|
|
if s.source == nil {
|
|
invalid = append(invalid, "Source")
|
|
} else {
|
|
if len(s.source.request.indices) == 0 {
|
|
invalid = append(invalid, "Source.Index")
|
|
}
|
|
}
|
|
if s.destination == nil {
|
|
invalid = append(invalid, "Destination")
|
|
}
|
|
if len(invalid) > 0 {
|
|
return fmt.Errorf("missing required fields: %v", invalid)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// getBody returns the body part of the document request.
|
|
func (s *ReindexService) getBody() (interface{}, error) {
|
|
if s.body != nil {
|
|
return s.body, nil
|
|
}
|
|
|
|
body := make(map[string]interface{})
|
|
|
|
if s.conflicts != "" {
|
|
body["conflicts"] = s.conflicts
|
|
}
|
|
if s.size != nil {
|
|
body["size"] = *s.size
|
|
}
|
|
if s.script != nil {
|
|
out, err := s.script.Source()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
body["script"] = out
|
|
}
|
|
|
|
src, err := s.source.Source()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
body["source"] = src
|
|
|
|
dst, err := s.destination.Source()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
body["dest"] = dst
|
|
|
|
return body, nil
|
|
}
|
|
|
|
// Do executes the operation.
|
|
func (s *ReindexService) Do(ctx context.Context) (*BulkIndexByScrollResponse, error) {
|
|
// Check pre-conditions
|
|
if err := s.Validate(); err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Get URL for request
|
|
path, params, err := s.buildURL()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Setup HTTP request body
|
|
body, err := s.getBody()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Get HTTP response
|
|
res, err := s.client.PerformRequest(ctx, PerformRequestOptions{
|
|
Method: "POST",
|
|
Path: path,
|
|
Params: params,
|
|
Body: body,
|
|
Headers: s.headers,
|
|
})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Return operation response
|
|
ret := new(BulkIndexByScrollResponse)
|
|
if err := s.client.decoder.Decode(res.Body, ret); err != nil {
|
|
return nil, err
|
|
}
|
|
ret.Header = res.Header
|
|
return ret, nil
|
|
}
|
|
|
|
// DoAsync executes the reindexing operation asynchronously by starting a new task.
|
|
// Callers need to use the Task Management API to watch the outcome of the reindexing
|
|
// operation.
|
|
func (s *ReindexService) DoAsync(ctx context.Context) (*StartTaskResult, error) {
|
|
// Check pre-conditions
|
|
if err := s.Validate(); err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// DoAsync only makes sense with WaitForCompletion set to false
|
|
if s.waitForCompletion != nil && *s.waitForCompletion {
|
|
return nil, fmt.Errorf("cannot start a task with WaitForCompletion set to true")
|
|
}
|
|
f := false
|
|
s.waitForCompletion = &f
|
|
|
|
// Get URL for request
|
|
path, params, err := s.buildURL()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Setup HTTP request body
|
|
body, err := s.getBody()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Get HTTP response
|
|
res, err := s.client.PerformRequest(ctx, PerformRequestOptions{
|
|
Method: "POST",
|
|
Path: path,
|
|
Params: params,
|
|
Body: body,
|
|
Headers: s.headers,
|
|
})
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
// Return operation response
|
|
ret := new(StartTaskResult)
|
|
if err := s.client.decoder.Decode(res.Body, ret); err != nil {
|
|
return nil, err
|
|
}
|
|
ret.Header = res.Header
|
|
return ret, nil
|
|
}
|
|
|
|
// -- Source of Reindex --
|
|
|
|
// ReindexSource specifies the source of a Reindex process.
|
|
type ReindexSource struct {
|
|
request *SearchRequest
|
|
remoteInfo *ReindexRemoteInfo
|
|
}
|
|
|
|
// NewReindexSource creates a new ReindexSource.
|
|
func NewReindexSource() *ReindexSource {
|
|
return &ReindexSource{
|
|
request: NewSearchRequest(),
|
|
}
|
|
}
|
|
|
|
// Request specifies the search request used for source.
|
|
func (r *ReindexSource) Request(request *SearchRequest) *ReindexSource {
|
|
if request == nil {
|
|
r.request = NewSearchRequest()
|
|
} else {
|
|
r.request = request
|
|
}
|
|
return r
|
|
}
|
|
|
|
// SearchType is the search operation type. Possible values are
|
|
// "query_then_fetch" and "dfs_query_then_fetch".
|
|
func (r *ReindexSource) SearchType(searchType string) *ReindexSource {
|
|
r.request = r.request.SearchType(searchType)
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) SearchTypeDfsQueryThenFetch() *ReindexSource {
|
|
r.request = r.request.SearchType("dfs_query_then_fetch")
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) SearchTypeQueryThenFetch() *ReindexSource {
|
|
r.request = r.request.SearchType("query_then_fetch")
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) Index(indices ...string) *ReindexSource {
|
|
r.request = r.request.Index(indices...)
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) Type(types ...string) *ReindexSource {
|
|
r.request = r.request.Type(types...)
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) Preference(preference string) *ReindexSource {
|
|
r.request = r.request.Preference(preference)
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) RequestCache(requestCache bool) *ReindexSource {
|
|
r.request = r.request.RequestCache(requestCache)
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) Scroll(scroll string) *ReindexSource {
|
|
r.request = r.request.Scroll(scroll)
|
|
return r
|
|
}
|
|
|
|
func (r *ReindexSource) Query(query Query) *ReindexSource {
|
|
r.request = r.request.Query(query)
|
|
return r
|
|
}
|
|
|
|
// Sort adds a sort order.
|
|
func (r *ReindexSource) Sort(field string, ascending bool) *ReindexSource {
|
|
r.request = r.request.Sort(field, ascending)
|
|
return r
|
|
}
|
|
|
|
// SortWithInfo adds a sort order.
|
|
func (r *ReindexSource) SortWithInfo(info SortInfo) *ReindexSource {
|
|
r.request = r.request.SortWithInfo(info)
|
|
return r
|
|
}
|
|
|
|
// SortBy adds a sort order.
|
|
func (r *ReindexSource) SortBy(sorter ...Sorter) *ReindexSource {
|
|
r.request = r.request.SortBy(sorter...)
|
|
return r
|
|
}
|
|
|
|
// FetchSource indicates whether the response should contain the stored
|
|
// _source for every hit.
|
|
func (r *ReindexSource) FetchSource(fetchSource bool) *ReindexSource {
|
|
r.request = r.request.FetchSource(fetchSource)
|
|
return r
|
|
}
|
|
|
|
// FetchSourceIncludeExclude specifies that _source should be returned
|
|
// with each hit, where "include" and "exclude" serve as a simple wildcard
|
|
// matcher that gets applied to its fields
|
|
// (e.g. include := []string{"obj1.*","obj2.*"}, exclude := []string{"description.*"}).
|
|
func (r *ReindexSource) FetchSourceIncludeExclude(include, exclude []string) *ReindexSource {
|
|
r.request = r.request.FetchSourceIncludeExclude(include, exclude)
|
|
return r
|
|
}
|
|
|
|
// FetchSourceContext indicates how the _source should be fetched.
|
|
func (r *ReindexSource) FetchSourceContext(fsc *FetchSourceContext) *ReindexSource {
|
|
r.request = r.request.FetchSourceContext(fsc)
|
|
return r
|
|
}
|
|
|
|
// RemoteInfo sets up reindexing from a remote cluster.
|
|
func (r *ReindexSource) RemoteInfo(ri *ReindexRemoteInfo) *ReindexSource {
|
|
r.remoteInfo = ri
|
|
return r
|
|
}
|
|
|
|
// Source returns a serializable JSON request for the request.
|
|
func (r *ReindexSource) Source() (interface{}, error) {
|
|
src, err := r.request.sourceAsMap()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
source, ok := src.(map[string]interface{})
|
|
if !ok {
|
|
return nil, errors.New("unable to use SearchRequest as map[string]interface{}")
|
|
}
|
|
|
|
switch len(r.request.indices) {
|
|
case 1:
|
|
source["index"] = r.request.indices[0]
|
|
default:
|
|
source["index"] = r.request.indices
|
|
}
|
|
switch len(r.request.types) {
|
|
case 0:
|
|
case 1:
|
|
source["type"] = r.request.types[0]
|
|
default:
|
|
source["type"] = r.request.types
|
|
}
|
|
if r.remoteInfo != nil {
|
|
src, err := r.remoteInfo.Source()
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
source["remote"] = src
|
|
}
|
|
return source, nil
|
|
}
|
|
|
|
// ReindexRemoteInfo contains information for reindexing from a remote cluster.
|
|
type ReindexRemoteInfo struct {
|
|
host string
|
|
username string
|
|
password string
|
|
socketTimeout string // e.g. "1m" or "30s"
|
|
connectTimeout string // e.g. "1m" or "30s"
|
|
}
|
|
|
|
// NewReindexRemoteInfo creates a new ReindexRemoteInfo.
|
|
func NewReindexRemoteInfo() *ReindexRemoteInfo {
|
|
return &ReindexRemoteInfo{}
|
|
}
|
|
|
|
// Host sets the host information of the remote cluster.
|
|
// It must be of the form "http(s)://<hostname>:<port>"
|
|
func (ri *ReindexRemoteInfo) Host(host string) *ReindexRemoteInfo {
|
|
ri.host = host
|
|
return ri
|
|
}
|
|
|
|
// Username sets the username to authenticate with the remote cluster.
|
|
func (ri *ReindexRemoteInfo) Username(username string) *ReindexRemoteInfo {
|
|
ri.username = username
|
|
return ri
|
|
}
|
|
|
|
// Password sets the password to authenticate with the remote cluster.
|
|
func (ri *ReindexRemoteInfo) Password(password string) *ReindexRemoteInfo {
|
|
ri.password = password
|
|
return ri
|
|
}
|
|
|
|
// SocketTimeout sets the socket timeout to connect with the remote cluster.
|
|
// Use ES compatible values like e.g. "30s" or "1m".
|
|
func (ri *ReindexRemoteInfo) SocketTimeout(timeout string) *ReindexRemoteInfo {
|
|
ri.socketTimeout = timeout
|
|
return ri
|
|
}
|
|
|
|
// ConnectTimeout sets the connection timeout to connect with the remote cluster.
|
|
// Use ES compatible values like e.g. "30s" or "1m".
|
|
func (ri *ReindexRemoteInfo) ConnectTimeout(timeout string) *ReindexRemoteInfo {
|
|
ri.connectTimeout = timeout
|
|
return ri
|
|
}
|
|
|
|
// Source returns the serializable JSON data for the request.
|
|
func (ri *ReindexRemoteInfo) Source() (interface{}, error) {
|
|
res := make(map[string]interface{})
|
|
res["host"] = ri.host
|
|
if len(ri.username) > 0 {
|
|
res["username"] = ri.username
|
|
}
|
|
if len(ri.password) > 0 {
|
|
res["password"] = ri.password
|
|
}
|
|
if len(ri.socketTimeout) > 0 {
|
|
res["socket_timeout"] = ri.socketTimeout
|
|
}
|
|
if len(ri.connectTimeout) > 0 {
|
|
res["connect_timeout"] = ri.connectTimeout
|
|
}
|
|
return res, nil
|
|
}
|
|
|
|
// -- Destination of Reindex --
|
|
|
|
// ReindexDestination is the destination of a Reindex API call.
|
|
// It is basically the meta data of a BulkIndexRequest.
|
|
//
|
|
// See https://www.elastic.co/guide/en/elasticsearch/reference/7.0/docs-reindex.html
|
|
// for details.
|
|
type ReindexDestination struct {
|
|
index string
|
|
typ string
|
|
routing string
|
|
parent string
|
|
opType string
|
|
version int64 // default is MATCH_ANY
|
|
versionType string // default is "internal"
|
|
pipeline string
|
|
}
|
|
|
|
// NewReindexDestination returns a new ReindexDestination.
|
|
func NewReindexDestination() *ReindexDestination {
|
|
return &ReindexDestination{}
|
|
}
|
|
|
|
// Index specifies name of the Elasticsearch index to use as the destination
|
|
// of a reindexing process.
|
|
func (r *ReindexDestination) Index(index string) *ReindexDestination {
|
|
r.index = index
|
|
return r
|
|
}
|
|
|
|
// Type specifies the Elasticsearch type to use for reindexing.
|
|
func (r *ReindexDestination) Type(typ string) *ReindexDestination {
|
|
r.typ = typ
|
|
return r
|
|
}
|
|
|
|
// Routing specifies a routing value for the reindexing request.
|
|
// It can be "keep", "discard", or start with "=". The latter specifies
|
|
// the routing on the bulk request.
|
|
func (r *ReindexDestination) Routing(routing string) *ReindexDestination {
|
|
r.routing = routing
|
|
return r
|
|
}
|
|
|
|
// Keep sets the routing on the bulk request sent for each match to the routing
|
|
// of the match (the default).
|
|
func (r *ReindexDestination) Keep() *ReindexDestination {
|
|
r.routing = "keep"
|
|
return r
|
|
}
|
|
|
|
// Discard sets the routing on the bulk request sent for each match to null.
|
|
func (r *ReindexDestination) Discard() *ReindexDestination {
|
|
r.routing = "discard"
|
|
return r
|
|
}
|
|
|
|
// Parent specifies the identifier of the parent document (if available).
|
|
func (r *ReindexDestination) Parent(parent string) *ReindexDestination {
|
|
r.parent = parent
|
|
return r
|
|
}
|
|
|
|
// OpType specifies if this request should follow create-only or upsert
|
|
// behavior. This follows the OpType of the standard document index API.
|
|
// See https://www.elastic.co/guide/en/elasticsearch/reference/7.0/docs-index_.html#operation-type
|
|
// for details.
|
|
func (r *ReindexDestination) OpType(opType string) *ReindexDestination {
|
|
r.opType = opType
|
|
return r
|
|
}
|
|
|
|
// Version indicates the version of the document as part of an optimistic
|
|
// concurrency model.
|
|
func (r *ReindexDestination) Version(version int64) *ReindexDestination {
|
|
r.version = version
|
|
return r
|
|
}
|
|
|
|
// VersionType specifies how versions are created.
|
|
func (r *ReindexDestination) VersionType(versionType string) *ReindexDestination {
|
|
r.versionType = versionType
|
|
return r
|
|
}
|
|
|
|
// Pipeline specifies the pipeline to use for reindexing.
|
|
func (r *ReindexDestination) Pipeline(pipeline string) *ReindexDestination {
|
|
r.pipeline = pipeline
|
|
return r
|
|
}
|
|
|
|
// Source returns a serializable JSON request for the request.
|
|
func (r *ReindexDestination) Source() (interface{}, error) {
|
|
source := make(map[string]interface{})
|
|
if r.index != "" {
|
|
source["index"] = r.index
|
|
}
|
|
if r.typ != "" {
|
|
source["type"] = r.typ
|
|
}
|
|
if r.routing != "" {
|
|
source["routing"] = r.routing
|
|
}
|
|
if r.opType != "" {
|
|
source["op_type"] = r.opType
|
|
}
|
|
if r.parent != "" {
|
|
source["parent"] = r.parent
|
|
}
|
|
if r.version > 0 {
|
|
source["version"] = r.version
|
|
}
|
|
if r.versionType != "" {
|
|
source["version_type"] = r.versionType
|
|
}
|
|
if r.pipeline != "" {
|
|
source["pipeline"] = r.pipeline
|
|
}
|
|
return source, nil
|
|
}
|