[[search-suggesters-completion]] === Completion Suggester NOTE: In order to understand the format of suggestions, please read the <> page first. The `completion` suggester is a so-called prefix suggester. It does not do spell correction like the `term` or `phrase` suggesters but allows basic `auto-complete` functionality. IMPORTANT: This feature is marked as experimental. This means, that the API is not considered stable, and that you might need to reindex your data after an upgrade in order to get suggestions up and running again. Please keep this in mind. ==== Why another suggester? Why not prefix queries? The first question which comes to mind when reading about a prefix suggestion is, why you should use it all, if you have prefix queries already. The answer is simple: Prefix suggestions are fast. The data structures are internally backed by Lucenes `AnalyzingSuggester`, which uses FSTs to execute suggestions. Usually these data structures are costly to create, stored in-memory and need to be rebuilt every now and then to reflect changes in your indexed documents. The `completion` suggester circumvents this by storing the FST as part of your index during index time. This allows for really fast loads and executions. [[completion-suggester-mapping]] ==== Mapping In order to use this feature, you have to specify a special mapping for this field, which enables the special storage of the field. [source,js] -------------------------------------------------- curl -X PUT localhost:9200/music curl -X PUT localhost:9200/music/song/_mapping -d '{ "song" : { "properties" : { "name" : { "type" : "string" }, "suggest" : { "type" : "completion", "index_analyzer" : "simple", "search_analyzer" : "simple", "payloads" : true } } } }' -------------------------------------------------- Mapping supports the following parameters: `index_analyzer`:: The index analyzer to use, defaults to `simple`. `search_analyzer`:: The search analyzer to use, defaults to `simple`. In case you are wondering why we did not opt for the `standard` analyzer: We try to have easy to understand behaviour here, and if you index the field content `At the Drive-in`, you will not get any suggestions for `a`, nor for `d` (the first non stopword). `payloads`:: Enables the storing of payloads, defaults to `false` `preserve_separators`:: Preserves the separators, defaults to `true`. If disabled, you could find a field starting with `Foo Fighters`, if you suggest for `foof`. `preserve_position_increments`:: Enables position increments, defaults to `true`. If disabled and using stopwords analyzer, you could get a field starting with `The Beatles`, if you suggest for `b`. *Note*: You could also achieve this by indexing two inputs, `Beatles` and `The Beatles`, no need to change a simple analyzer, if you are able to enrich your data. `max_input_len`:: Limits the length of a single input, defaults to `50` UTF-16 code points. This limit is only used at index time to reduce the total number of characters per input string in order to prevent massive inputs from bloating the underlying datastructure. The most usecases won't be influenced by the default value since prefix completions hardly grow beyond prefixes longer than a handful of characters. [[indexing]] ==== Indexing [source,js] -------------------------------------------------- curl -X PUT 'localhost:9200/music/song/1?refresh=true' -d '{ "name" : "Nevermind", "suggest" : { "input": [ "Nevermind", "Nirvana" ], "output": "Nirvana - Nevermind", "payload" : { "artistId" : 2321 }, "weight" : 34 } }' -------------------------------------------------- The following parameters are supported: `input`:: The input to store, this can be a an array of strings or just a string. This field is mandatory. `output`:: The string to return, if a suggestion matches. This is very useful to normalize outputs (i.e. have them always in the format `artist - songname`). The result is de-duplicated if several documents have the same output, i.e. only one is returned as part of the suggest result. This is optional. `payload`:: An arbitrary JSON object, which is simply returned in the suggest option. You could store data like the id of a document, in order to load it from elasticsearch without executing another search (which might not yield any results, if `input` and `output` differ strongly). `weight`:: A positive integer, which defines a weight and allows you to rank your suggestions. This field is optional. NOTE: Even though you are losing most of the features of the completion suggest, you can opt in for the shortest form, which even allows you to use inside of multi_field. But keep in mind, that you will not be able to use several inputs, an output, payloads or weights. [source,js] -------------------------------------------------- { "suggest" : "Nirvana" } -------------------------------------------------- NOTE: The suggest data structure might not reflect deletes on documents immediately. You may need to do an <> for that. You can call optimize with the `only_expunge_deletes=true` to only cater for deletes or alternatively call a <> operation. [[querying]] ==== Querying Suggesting works as usual, except that you have to specify the suggest type as `completion`. [source,js] -------------------------------------------------- curl -X POST 'localhost:9200/music/_suggest?pretty' -d '{ "song-suggest" : { "text" : "n", "completion" : { "field" : "suggest" } } }' { "_shards" : { "total" : 5, "successful" : 5, "failed" : 0 }, "song-suggest" : [ { "text" : "n", "offset" : 0, "length" : 4, "options" : [ { "text" : "Nirvana - Nevermind", "score" : 34.0, "payload" : {"artistId":2321} } ] } ] } -------------------------------------------------- As you can see, the payload is included in the response, if configured appropriately. If you configured a weight for a suggestion, this weight is used as `score`. Also the `text` field uses the `output` of your indexed suggestion, if configured, otherwise the matched part of the `input` field. [[fuzzy]] ==== Fuzzy queries The completion suggester also supports fuzzy queries - this means, you can actually have a typo in your search and still get results back. [source,js] -------------------------------------------------- curl -X POST 'localhost:9200/music/_suggest?pretty' -d '{ "song-suggest" : { "text" : "n", "completion" : { "field" : "suggest", "fuzzy" : { "edit_distance" : 2 } } } }' -------------------------------------------------- The fuzzy query can take specific fuzzy parameters. The following parameters are supported: [horizontal] `edit_distance`:: Maximum edit distance, defaults to `1` `transpositions`:: Sets if transpositions should be counted as one or two changes, defaults to `true` `min_length`:: Minimum length of the input before fuzzy suggestions are returned, defaults `3` `prefix_length`:: Minimum length of the input, which is not checked for fuzzy alternatives, defaults to `1` `unicode_aware`:: Sets all are measurements (like edit distance, transpositions and lengths) in unicode code points (actual letters) instead of bytes. NOTE: If you want to stick with the default values, but still use fuzzy, you can either use `fuzzy: {}` or `fuzzy: true`.