| 1 |
// Copyright 2012-present Oliver Eilhard. All rights reserved. |
| 2 |
// Use of this source code is governed by a MIT-license. |
| 3 |
// See http://olivere.mit-license.org/license.txt for details. |
| 4 |
|
| 5 |
package elastic |
| 6 |
|
| 7 |
import ( |
| 8 |
"context" |
| 9 |
"fmt" |
| 10 |
"io" |
| 11 |
"net/url" |
| 12 |
"strings" |
| 13 |
"sync" |
| 14 |
|
| 15 |
"gopkg.in/olivere/elastic.v5/uritemplates" |
| 16 |
) |
| 17 |
|
| 18 |
const ( |
| 19 |
// DefaultScrollKeepAlive is the default time a scroll cursor will be kept alive. |
| 20 |
DefaultScrollKeepAlive = "5m" |
| 21 |
) |
| 22 |
|
| 23 |
// ScrollService iterates over pages of search results from Elasticsearch. |
| 24 |
type ScrollService struct { |
| 25 |
client *Client |
| 26 |
retrier Retrier |
| 27 |
indices []string |
| 28 |
types []string |
| 29 |
keepAlive string |
| 30 |
body interface{} |
| 31 |
ss *SearchSource |
| 32 |
size *int |
| 33 |
pretty bool |
| 34 |
routing string |
| 35 |
preference string |
| 36 |
ignoreUnavailable *bool |
| 37 |
allowNoIndices *bool |
| 38 |
expandWildcards string |
| 39 |
filterPath []string |
| 40 |
|
| 41 |
mu sync.RWMutex |
| 42 |
scrollId string |
| 43 |
} |
| 44 |
|
| 45 |
// NewScrollService initializes and returns a new ScrollService. |
| 46 |
func NewScrollService(client *Client) *ScrollService { |
| 47 |
builder := &ScrollService{ |
| 48 |
client: client, |
| 49 |
ss: NewSearchSource(), |
| 50 |
keepAlive: DefaultScrollKeepAlive, |
| 51 |
} |
| 52 |
return builder |
| 53 |
} |
| 54 |
|
| 55 |
// Retrier allows to set specific retry logic for this ScrollService. |
| 56 |
// If not specified, it will use the client's default retrier. |
| 57 |
func (s *ScrollService) Retrier(retrier Retrier) *ScrollService { |
| 58 |
s.retrier = retrier |
| 59 |
return s |
| 60 |
} |
| 61 |
|
| 62 |
// Index sets the name of one or more indices to iterate over. |
| 63 |
func (s *ScrollService) Index(indices ...string) *ScrollService { |
| 64 |
if s.indices == nil { |
| 65 |
s.indices = make([]string, 0) |
| 66 |
} |
| 67 |
s.indices = append(s.indices, indices...) |
| 68 |
return s |
| 69 |
} |
| 70 |
|
| 71 |
// Type sets the name of one or more types to iterate over. |
| 72 |
func (s *ScrollService) Type(types ...string) *ScrollService { |
| 73 |
if s.types == nil { |
| 74 |
s.types = make([]string, 0) |
| 75 |
} |
| 76 |
s.types = append(s.types, types...) |
| 77 |
return s |
| 78 |
} |
| 79 |
|
| 80 |
// Scroll is an alias for KeepAlive, the time to keep |
| 81 |
// the cursor alive (e.g. "5m" for 5 minutes). |
| 82 |
func (s *ScrollService) Scroll(keepAlive string) *ScrollService { |
| 83 |
s.keepAlive = keepAlive |
| 84 |
return s |
| 85 |
} |
| 86 |
|
| 87 |
// KeepAlive sets the maximum time after which the cursor will expire. |
| 88 |
// It is "2m" by default. |
| 89 |
func (s *ScrollService) KeepAlive(keepAlive string) *ScrollService { |
| 90 |
s.keepAlive = keepAlive |
| 91 |
return s |
| 92 |
} |
| 93 |
|
| 94 |
// Size specifies the number of documents Elasticsearch should return |
| 95 |
// from each shard, per page. |
| 96 |
func (s *ScrollService) Size(size int) *ScrollService { |
| 97 |
s.size = &size |
| 98 |
return s |
| 99 |
} |
| 100 |
|
| 101 |
// Body sets the raw body to send to Elasticsearch. This can be e.g. a string, |
| 102 |
// a map[string]interface{} or anything that can be serialized into JSON. |
| 103 |
// Notice that setting the body disables the use of SearchSource and many |
| 104 |
// other properties of the SearchService. |
| 105 |
func (s *ScrollService) Body(body interface{}) *ScrollService { |
| 106 |
s.body = body |
| 107 |
return s |
| 108 |
} |
| 109 |
|
| 110 |
// SearchSource sets the search source builder to use with this iterator. |
| 111 |
// Notice that only a certain number of properties can be used when scrolling, |
| 112 |
// e.g. query and sorting. |
| 113 |
func (s *ScrollService) SearchSource(searchSource *SearchSource) *ScrollService { |
| 114 |
s.ss = searchSource |
| 115 |
if s.ss == nil { |
| 116 |
s.ss = NewSearchSource() |
| 117 |
} |
| 118 |
return s |
| 119 |
} |
| 120 |
|
| 121 |
// Query sets the query to perform, e.g. a MatchAllQuery. |
| 122 |
func (s *ScrollService) Query(query Query) *ScrollService { |
| 123 |
s.ss = s.ss.Query(query) |
| 124 |
return s |
| 125 |
} |
| 126 |
|
| 127 |
// PostFilter is executed as the last filter. It only affects the |
| 128 |
// search hits but not facets. See |
| 129 |
// https://www.elastic.co/guide/en/elasticsearch/reference/5.2/search-request-post-filter.html |
| 130 |
// for details. |
| 131 |
func (s *ScrollService) PostFilter(postFilter Query) *ScrollService { |
| 132 |
s.ss = s.ss.PostFilter(postFilter) |
| 133 |
return s |
| 134 |
} |
| 135 |
|
| 136 |
// Slice allows slicing the scroll request into several batches. |
| 137 |
// This is supported in Elasticsearch 5.0 or later. |
| 138 |
// See https://www.elastic.co/guide/en/elasticsearch/reference/5.2/search-request-scroll.html#sliced-scroll |
| 139 |
// for details. |
| 140 |
func (s *ScrollService) Slice(sliceQuery Query) *ScrollService { |
| 141 |
s.ss = s.ss.Slice(sliceQuery) |
| 142 |
return s |
| 143 |
} |
| 144 |
|
| 145 |
// FetchSource indicates whether the response should contain the stored |
| 146 |
// _source for every hit. |
| 147 |
func (s *ScrollService) FetchSource(fetchSource bool) *ScrollService { |
| 148 |
s.ss = s.ss.FetchSource(fetchSource) |
| 149 |
return s |
| 150 |
} |
| 151 |
|
| 152 |
// FetchSourceContext indicates how the _source should be fetched. |
| 153 |
func (s *ScrollService) FetchSourceContext(fetchSourceContext *FetchSourceContext) *ScrollService { |
| 154 |
s.ss = s.ss.FetchSourceContext(fetchSourceContext) |
| 155 |
return s |
| 156 |
} |
| 157 |
|
| 158 |
// Version can be set to true to return a version for each search hit. |
| 159 |
// See https://www.elastic.co/guide/en/elasticsearch/reference/5.2/search-request-version.html. |
| 160 |
func (s *ScrollService) Version(version bool) *ScrollService { |
| 161 |
s.ss = s.ss.Version(version) |
| 162 |
return s |
| 163 |
} |
| 164 |
|
| 165 |
// Sort adds a sort order. This can have negative effects on the performance |
| 166 |
// of the scroll operation as Elasticsearch needs to sort first. |
| 167 |
func (s *ScrollService) Sort(field string, ascending bool) *ScrollService { |
| 168 |
s.ss = s.ss.Sort(field, ascending) |
| 169 |
return s |
| 170 |
} |
| 171 |
|
| 172 |
// SortWithInfo specifies a sort order. Notice that sorting can have a |
| 173 |
// negative impact on scroll performance. |
| 174 |
func (s *ScrollService) SortWithInfo(info SortInfo) *ScrollService { |
| 175 |
s.ss = s.ss.SortWithInfo(info) |
| 176 |
return s |
| 177 |
} |
| 178 |
|
| 179 |
// SortBy specifies a sort order. Notice that sorting can have a |
| 180 |
// negative impact on scroll performance. |
| 181 |
func (s *ScrollService) SortBy(sorter ...Sorter) *ScrollService { |
| 182 |
s.ss = s.ss.SortBy(sorter...) |
| 183 |
return s |
| 184 |
} |
| 185 |
|
| 186 |
// Pretty asks Elasticsearch to pretty-print the returned JSON. |
| 187 |
func (s *ScrollService) Pretty(pretty bool) *ScrollService { |
| 188 |
s.pretty = pretty |
| 189 |
return s |
| 190 |
} |
| 191 |
|
| 192 |
// Routing is a list of specific routing values to control the shards |
| 193 |
// the search will be executed on. |
| 194 |
func (s *ScrollService) Routing(routings ...string) *ScrollService { |
| 195 |
s.routing = strings.Join(routings, ",") |
| 196 |
return s |
| 197 |
} |
| 198 |
|
| 199 |
// Preference sets the preference to execute the search. Defaults to |
| 200 |
// randomize across shards ("random"). Can be set to "_local" to prefer |
| 201 |
// local shards, "_primary" to execute on primary shards only, |
| 202 |
// or a custom value which guarantees that the same order will be used |
| 203 |
// across different requests. |
| 204 |
func (s *ScrollService) Preference(preference string) *ScrollService { |
| 205 |
s.preference = preference |
| 206 |
return s |
| 207 |
} |
| 208 |
|
| 209 |
// IgnoreUnavailable indicates whether the specified concrete indices |
| 210 |
// should be ignored when unavailable (missing or closed). |
| 211 |
func (s *ScrollService) IgnoreUnavailable(ignoreUnavailable bool) *ScrollService { |
| 212 |
s.ignoreUnavailable = &ignoreUnavailable |
| 213 |
return s |
| 214 |
} |
| 215 |
|
| 216 |
// AllowNoIndices indicates whether to ignore if a wildcard indices |
| 217 |
// expression resolves into no concrete indices. (This includes `_all` string |
| 218 |
// or when no indices have been specified). |
| 219 |
func (s *ScrollService) AllowNoIndices(allowNoIndices bool) *ScrollService { |
| 220 |
s.allowNoIndices = &allowNoIndices |
| 221 |
return s |
| 222 |
} |
| 223 |
|
| 224 |
// ExpandWildcards indicates whether to expand wildcard expression to |
| 225 |
// concrete indices that are open, closed or both. |
| 226 |
func (s *ScrollService) ExpandWildcards(expandWildcards string) *ScrollService { |
| 227 |
s.expandWildcards = expandWildcards |
| 228 |
return s |
| 229 |
} |
| 230 |
|
| 231 |
// FilterPath allows reducing the response, a mechanism known as |
| 232 |
// response filtering and described here: |
| 233 |
// https://www.elastic.co/guide/en/elasticsearch/reference/5.6/common-options.html#common-options-response-filtering. |
| 234 |
func (s *ScrollService) FilterPath(filterPath ...string) *ScrollService { |
| 235 |
s.filterPath = append(s.filterPath, filterPath...) |
| 236 |
return s |
| 237 |
} |
| 238 |
|
| 239 |
// ScrollId specifies the identifier of a scroll in action. |
| 240 |
func (s *ScrollService) ScrollId(scrollId string) *ScrollService { |
| 241 |
s.mu.Lock() |
| 242 |
s.scrollId = scrollId |
| 243 |
s.mu.Unlock() |
| 244 |
return s |
| 245 |
} |
| 246 |
|
| 247 |
// Do returns the next search result. It will return io.EOF as error if there |
| 248 |
// are no more search results. |
| 249 |
func (s *ScrollService) Do(ctx context.Context) (*SearchResult, error) { |
| 250 |
s.mu.RLock() |
| 251 |
nextScrollId := s.scrollId |
| 252 |
s.mu.RUnlock() |
| 253 |
if len(nextScrollId) == 0 { |
| 254 |
return s.first(ctx) |
| 255 |
} |
| 256 |
return s.next(ctx) |
| 257 |
} |
| 258 |
|
| 259 |
// Clear cancels the current scroll operation. If you don't do this manually, |
| 260 |
// the scroll will be expired automatically by Elasticsearch. You can control |
| 261 |
// how long a scroll cursor is kept alive with the KeepAlive func. |
| 262 |
func (s *ScrollService) Clear(ctx context.Context) error { |
| 263 |
s.mu.RLock() |
| 264 |
scrollId := s.scrollId |
| 265 |
s.mu.RUnlock() |
| 266 |
if len(scrollId) == 0 { |
| 267 |
return nil |
| 268 |
} |
| 269 |
|
| 270 |
path := "/_search/scroll" |
| 271 |
params := url.Values{} |
| 272 |
body := struct { |
| 273 |
ScrollId []string `json:"scroll_id,omitempty"` |
| 274 |
}{ |
| 275 |
ScrollId: []string{scrollId}, |
| 276 |
} |
| 277 |
|
| 278 |
_, err := s.client.PerformRequestWithOptions(ctx, PerformRequestOptions{ |
| 279 |
Method: "DELETE", |
| 280 |
Path: path, |
| 281 |
Params: params, |
| 282 |
Body: body, |
| 283 |
Retrier: s.retrier, |
| 284 |
}) |
| 285 |
if err != nil { |
| 286 |
return err |
| 287 |
} |
| 288 |
|
| 289 |
return nil |
| 290 |
} |
| 291 |
|
| 292 |
// -- First -- |
| 293 |
|
| 294 |
// first takes the first page of search results. |
| 295 |
func (s *ScrollService) first(ctx context.Context) (*SearchResult, error) { |
| 296 |
// Get URL and parameters for request |
| 297 |
path, params, err := s.buildFirstURL() |
| 298 |
if err != nil { |
| 299 |
return nil, err |
| 300 |
} |
| 301 |
|
| 302 |
// Get HTTP request body |
| 303 |
body, err := s.bodyFirst() |
| 304 |
if err != nil { |
| 305 |
return nil, err |
| 306 |
} |
| 307 |
|
| 308 |
// Get HTTP response |
| 309 |
res, err := s.client.PerformRequestWithOptions(ctx, PerformRequestOptions{ |
| 310 |
Method: "POST", |
| 311 |
Path: path, |
| 312 |
Params: params, |
| 313 |
Body: body, |
| 314 |
Retrier: s.retrier, |
| 315 |
}) |
| 316 |
if err != nil { |
| 317 |
return nil, err |
| 318 |
} |
| 319 |
|
| 320 |
// Return operation response |
| 321 |
ret := new(SearchResult) |
| 322 |
if err := s.client.decoder.Decode(res.Body, ret); err != nil { |
| 323 |
return nil, err |
| 324 |
} |
| 325 |
s.mu.Lock() |
| 326 |
s.scrollId = ret.ScrollId |
| 327 |
s.mu.Unlock() |
| 328 |
if ret.Hits == nil || len(ret.Hits.Hits) == 0 { |
| 329 |
return ret, io.EOF |
| 330 |
} |
| 331 |
return ret, nil |
| 332 |
} |
| 333 |
|
| 334 |
// buildFirstURL builds the URL for retrieving the first page. |
| 335 |
func (s *ScrollService) buildFirstURL() (string, url.Values, error) { |
| 336 |
// Build URL |
| 337 |
var err error |
| 338 |
var path string |
| 339 |
if len(s.indices) == 0 && len(s.types) == 0 { |
| 340 |
path = "/_search" |
| 341 |
} else if len(s.indices) > 0 && len(s.types) == 0 { |
| 342 |
path, err = uritemplates.Expand("/{index}/_search", map[string]string{ |
| 343 |
"index": strings.Join(s.indices, ","), |
| 344 |
}) |
| 345 |
} else if len(s.indices) == 0 && len(s.types) > 0 { |
| 346 |
path, err = uritemplates.Expand("/_all/{typ}/_search", map[string]string{ |
| 347 |
"typ": strings.Join(s.types, ","), |
| 348 |
}) |
| 349 |
} else { |
| 350 |
path, err = uritemplates.Expand("/{index}/{typ}/_search", map[string]string{ |
| 351 |
"index": strings.Join(s.indices, ","), |
| 352 |
"typ": strings.Join(s.types, ","), |
| 353 |
}) |
| 354 |
} |
| 355 |
if err != nil { |
| 356 |
return "", url.Values{}, err |
| 357 |
} |
| 358 |
|
| 359 |
// Add query string parameters |
| 360 |
params := url.Values{} |
| 361 |
if s.pretty { |
| 362 |
params.Set("pretty", "1") |
| 363 |
} |
| 364 |
if s.size != nil && *s.size > 0 { |
| 365 |
params.Set("size", fmt.Sprintf("%d", *s.size)) |
| 366 |
} |
| 367 |
if len(s.keepAlive) > 0 { |
| 368 |
params.Set("scroll", s.keepAlive) |
| 369 |
} |
| 370 |
if len(s.routing) > 0 { |
| 371 |
params.Set("routing", s.routing) |
| 372 |
} |
| 373 |
if len(s.preference) > 0 { |
| 374 |
params.Set("preference", s.preference) |
| 375 |
} |
| 376 |
if s.allowNoIndices != nil { |
| 377 |
params.Set("allow_no_indices", fmt.Sprintf("%v", *s.allowNoIndices)) |
| 378 |
} |
| 379 |
if len(s.expandWildcards) > 0 { |
| 380 |
params.Set("expand_wildcards", s.expandWildcards) |
| 381 |
} |
| 382 |
if s.ignoreUnavailable != nil { |
| 383 |
params.Set("ignore_unavailable", fmt.Sprintf("%v", *s.ignoreUnavailable)) |
| 384 |
} |
| 385 |
if len(s.filterPath) > 0 { |
| 386 |
// Always add "hits._scroll_id", otherwise we cannot scroll |
| 387 |
s.filterPath = append(s.filterPath, "_scroll_id") |
| 388 |
params.Set("filter_path", strings.Join(s.filterPath, ",")) |
| 389 |
} |
| 390 |
|
| 391 |
return path, params, nil |
| 392 |
} |
| 393 |
|
| 394 |
// bodyFirst returns the request to fetch the first batch of results. |
| 395 |
func (s *ScrollService) bodyFirst() (interface{}, error) { |
| 396 |
var err error |
| 397 |
var body interface{} |
| 398 |
|
| 399 |
if s.body != nil { |
| 400 |
body = s.body |
| 401 |
} else { |
| 402 |
// Use _doc sort by default if none is specified |
| 403 |
if !s.ss.hasSort() { |
| 404 |
// Use efficient sorting when no user-defined query/body is specified |
| 405 |
s.ss = s.ss.SortBy(SortByDoc{}) |
| 406 |
} |
| 407 |
|
| 408 |
// Body from search source |
| 409 |
body, err = s.ss.Source() |
| 410 |
if err != nil { |
| 411 |
return nil, err |
| 412 |
} |
| 413 |
} |
| 414 |
|
| 415 |
return body, nil |
| 416 |
} |
| 417 |
|
| 418 |
// -- Next -- |
| 419 |
|
| 420 |
func (s *ScrollService) next(ctx context.Context) (*SearchResult, error) { |
| 421 |
// Get URL for request |
| 422 |
path, params, err := s.buildNextURL() |
| 423 |
if err != nil { |
| 424 |
return nil, err |
| 425 |
} |
| 426 |
|
| 427 |
// Setup HTTP request body |
| 428 |
body, err := s.bodyNext() |
| 429 |
if err != nil { |
| 430 |
return nil, err |
| 431 |
} |
| 432 |
|
| 433 |
// Get HTTP response |
| 434 |
res, err := s.client.PerformRequestWithOptions(ctx, PerformRequestOptions{ |
| 435 |
Method: "POST", |
| 436 |
Path: path, |
| 437 |
Params: params, |
| 438 |
Body: body, |
| 439 |
Retrier: s.retrier, |
| 440 |
}) |
| 441 |
if err != nil { |
| 442 |
return nil, err |
| 443 |
} |
| 444 |
|
| 445 |
// Return operation response |
| 446 |
ret := new(SearchResult) |
| 447 |
if err := s.client.decoder.Decode(res.Body, ret); err != nil { |
| 448 |
return nil, err |
| 449 |
} |
| 450 |
s.mu.Lock() |
| 451 |
s.scrollId = ret.ScrollId |
| 452 |
s.mu.Unlock() |
| 453 |
if ret.Hits == nil || len(ret.Hits.Hits) == 0 { |
| 454 |
return ret, io.EOF |
| 455 |
} |
| 456 |
return ret, nil |
| 457 |
} |
| 458 |
|
| 459 |
// buildNextURL builds the URL for the operation. |
| 460 |
func (s *ScrollService) buildNextURL() (string, url.Values, error) { |
| 461 |
path := "/_search/scroll" |
| 462 |
|
| 463 |
// Add query string parameters |
| 464 |
params := url.Values{} |
| 465 |
if s.pretty { |
| 466 |
params.Set("pretty", "1") |
| 467 |
} |
| 468 |
|
| 469 |
return path, params, nil |
| 470 |
} |
| 471 |
|
| 472 |
// body returns the request to fetch the next batch of results. |
| 473 |
func (s *ScrollService) bodyNext() (interface{}, error) { |
| 474 |
s.mu.RLock() |
| 475 |
body := struct { |
| 476 |
Scroll string `json:"scroll"` |
| 477 |
ScrollId string `json:"scroll_id,omitempty"` |
| 478 |
}{ |
| 479 |
Scroll: s.keepAlive, |
| 480 |
ScrollId: s.scrollId, |
| 481 |
} |
| 482 |
s.mu.RUnlock() |
| 483 |
return body, nil |
| 484 |
} |