Elastic search match phrase query with single token - elasticsearch

Question

So I am trying to understand how the match_phrase query works under certain circumstances
with elastic search [We have version 6.8 set up as of now ] . When I give it a string with multiple tokens it shows while profiling its running a phrase query but when I run it with a single token while profiling it shows its running a termsquery internally . I am trying to understand shouldn't it be independent of the input and if the positioning of terms is not correct fail to return a match ? Attaching queries and o/p -
Query with multiple tokens -
GET potato_testv3/_search
{"profile": "true",
"query": {
"bool": {
"must": [
{ "match_phrase": { "skill_set": {"query":"potato farmer"} }}
]
}
}
}
Output of the above -
{
"took" : 2,
"timed_out" : false,
"_shards" : {
"total" : 1,
"successful" : 1,
"skipped" : 0,
"failed" : 0
},
"hits" : {
"total" : 1,
"max_score" : 0.5753642,
"hits" : [
{
"_index" : "potato_testv3",
"_type" : "recruiterinsightsv11",
"_id" : "4RShdnkBc8OOeUFVkncD",
"_score" : 0.5753642,
"_source" : {
"skill_set" : [
"silly webdriver",
"uft",
"uft/qtp",
"potato farmer"
]
}
}
]
},
"profile" : {
"shards" : [
{
"id" : "[5QVxJbTCSU-ruYT9EHsujA][potato_testv3][0]",
"searches" : [
{
"query" : [
{
"type" : "PhraseQuery",
"description" : """skill_set:"potato farmer"""",
"time_in_nanos" : 338986,
"breakdown" : {
"score" : 15362,
"build_scorer_count" : 2,
"match_count" : 1,
"create_weight" : 55661,
"next_doc" : 74248,
"match" : 39624,
"create_weight_count" : 1,
"next_doc_count" : 2,
"score_count" : 1,
"build_scorer" : 154084,
"advance" : 0,
"advance_count" : 0
}
}
],
"rewrite_time" : 3932,
"collector" : [
{
"name" : "CancellableCollector",
"reason" : "search_cancelled",
"time_in_nanos" : 48431,
"children" : [
{
"name" : "SimpleTopScoreDocCollector",
"reason" : "search_top_hits",
"time_in_nanos" : 19840
}
]
}
]
}
],
"aggregations" : [ ]
}
]
}
}
Query with single token -
GET potato_testv3/_search
{"profile": "true",
"query": {
"bool": {
"must": [
{ "match_phrase": { "skill_set": {"query":"potato"} }}
]
}
}
}
Output of above -
{
"took" : 2,
"timed_out" : false,
"_shards" : {
"total" : 1,
"successful" : 1,
"skipped" : 0,
"failed" : 0
},
"hits" : {
"total" : 1,
"max_score" : 0.2876821,
"hits" : [
{
"_index" : "potato_testv3",
"_type" : "recruiterinsightsv11",
"_id" : "4RShdnkBc8OOeUFVkncD",
"_score" : 0.2876821,
"_source" : {
"skill_set" : [
"silly webdriver",
"uft",
"uft/qtp",
"potato farmer"
]
}
}
]
},
"profile" : {
"shards" : [
{
"id" : "[TeKxvYLJQfG_GVtD3bmpiw][potato_testv3][0]",
"searches" : [
{
"query" : [
{
"type" : "TermQuery",
"description" : "skill_set:potato",
"time_in_nanos" : 52214,
"breakdown" : {
"score" : 11310,
"build_scorer_count" : 2,
"match_count" : 0,
"create_weight" : 30974,
"next_doc" : 1314,
"match" : 0,
"create_weight_count" : 1,
"next_doc_count" : 2,
"score_count" : 1,
"build_scorer" : 8610,
"advance" : 0,
"advance_count" : 0
}
}
],
"rewrite_time" : 3761,
"collector" : [
{
"name" : "CancellableCollector",
"reason" : "search_cancelled",
"time_in_nanos" : 20912,
"children" : [
{
"name" : "SimpleTopScoreDocCollector",
"reason" : "search_top_hits",
"time_in_nanos" : 15758
}
]
}
]
}
],
"aggregations" : [ ]
}
]
}
}
In case if it helps , schema of the index used -
{
"potato_testv3" : {
"mappings" : {
"recruiterinsightsv11" : {
"dynamic" : "false",
"properties" : {
"skill_set" : {
"type" : "text",
"norms" : false,
"fielddata" : true
}
}
}
}
}
}

Answer 1

You are executing the same match_phrase query, once with a search string made up of multiple terms, once with a search string of a single token.
When executing an Elasticsearch query, Elasticsearch will optimise the query and translate it to the relevant queries on Lucene level. A phrase-query is more expensive to execute as
all terms of the search string need to match, and on top of that
the positions of the terms in a matching document need to be in the very same order as in the search string
If your search string only consist of a single term Elasticsearch can skip all of that extra effort and simply query for documents matching that single search term. What you observe therefore, is making perfect sense. It shows you how Elasticsearch is optimising the query while executing it.