Update the how-to section of the docs for 7.0: #37717
Conversation
- new `feature`/`script_score` queries - new `index_phrases`/`index_prefixes` options - disabling `_field_names` doesn't help anymore - adaptive replica selection is on by default
jpountz
added
>docs
|
Pinging @elastic/es-search |
| === Faster phrase queries with `index_phrases` | ||
|
|
||
| The <<text,`text`>> field has an <<index-phrases,`index_phrases`>> option that | ||
| pre-indexes 2-shingles at index-time and is automatically leveraged by query |
There was a problem hiding this comment.
nit: just a matter of taste I guess, but I find "pre-indexes ... at index-time" is a bit redundant (especially the pre). Maybe "indexes 2-shingles" or "already adds 2-shingles at index-time" is enough?
| === Faster prefix queries with `index_prefixes` | ||
|
|
||
| The <<text,`text`>> field has an <<index-phrases,`index_prefixes`>> option that | ||
| pre-indexes prefixes of all terms at index-time and is automatically leveraged |
|
|
||
| The <<mapping-field-names-field,`_field_names` field>> introduces some | ||
| index-time overhead, so you might want to disable it if you never need to | ||
| run `exists` queries. |
There was a problem hiding this comment.
Just to confirm, this is removed because indexing this field doesn't have the index overhead it once did?
There was a problem hiding this comment.
Right. @colings86 made things better in #26930 and we now use doc values or norms to run queries most-of-time, and only index into the _field_names field when doc values are not enabled (either because the user disabled them or because the field doesn't support doc values).
| criteria called <<search-adaptive-replica,adaptive replica selection>> to select | ||
| the best copy of the data based on response time, service time, and queue size | ||
| of the node containing each copy of the shard. This can improve query throughput | ||
| and reduce latency for search-heavy applications. |
There was a problem hiding this comment.
Also just to confirm, this is removed because adaptive replica selection is now the default?
- new `rank_feature`/`script_score` queries - new `index_phrases`/`index_prefixes` options - disabling `_field_names` doesn't help anymore - adaptive replica selection is on by default
- new `rank_feature`/`script_score` queries - new `index_phrases`/`index_prefixes` options - disabling `_field_names` doesn't help anymore - adaptive replica selection is on by default
feature/script_scorequeriesindex_phrases/index_prefixesoptions_field_namesdoesn't help anymore