TopolearnFull-text search with relevance ranking is the reason Elasticsearch exists.
GET /index/_search with a query returns ranked hits. The workhorse full-text query is match: { "match": { "name": "blue pens" } } analyzes the query text into terms (blue, pens, lower-cased) and finds documents containing any of them, scoring documents with more and rarer matches higher.
The response contains:
hits.total: the number of matches (counted exactly up to 10,000 by default, then reported as a lower bound with"relation": "gte"),hits.hits: the top results, each with_id,_scoreand_source,tookin milliseconds andtimed_out.
Useful options on match: "operator": "and" to require all terms, "fuzziness": "AUTO" to tolerate typos, minimum_should_match for "at least 2 of 3 words". multi_match searches several fields at once, e.g. ["name^3", "description"] where ^3 boosts name matches.
Control the response with size (default 10), _source filtering ("_source": ["name", "price"]) or fields, and highlight to get matched fragments wrapped in tags.
A search with no query, or { "match_all": {} }, returns everything with score 1.0. GET /index/_count returns just the number of matches.
Example
GET /products/_search
{
"query": {
"match": { "name": { "query": "blue pens", "operator": "or", "fuzziness": "AUTO" } }
},
"_source": ["name", "price"],
"size": 3,
"highlight": { "fields": { "name": {} } }
}
# "hits": {
# "total": { "value": 2, "relation": "eq" },
# "hits": [
# { "_id": "1", "_score": 1.38, "_source": { "name": "Blue pen", "price": 1.2 },
# "highlight": { "name": ["<em>Blue</em> <em>pen</em>"] } },
# { "_id": "4", "_score": 0.54, "_source": { "name": "Blue notebook", "price": 4 } } ] }Remember
- match analyzes the query text and ranks by relevance (_score)
- operator: and, fuzziness and minimum_should_match tune matching
- multi_match with ^boost searches several fields
- size, _source filtering and highlight shape the response
Watch out for
- Reading hits.total as exact for large result sets: beyond 10,000 it is a lower bound unless track_total_hits is true.