09. Compound Query

Compound Query

 

앞서 이미 query 의 종류에 대해 살펴보았고 지금까지 본 모든 query 는 leaf query  로 

 

이는 Elasticsearch query 의 기본이 되며 주어진 fields 에 대해

 

값을 검색시 하나의 operation 만 허용한다. ( n field + 1 operation )

 

이에 반해 compound query 란 이런 leaf query 의 조합으로

 

다수의 operation 을 조합하는 것이다.

 

 

 

Query type context vs Filter type context

 

Compound query 는 query type context 와 filter type context 를 구분하여 조합하게 되는데

 

context 는 한글로 표현하자면 "절" 이다.

 

바꿔말하면 query 절과 filter 절이 있는데 이런 절 안에는

 

Elasticsearch 에서 제공하는 query 를 채울 수 있다.

 

query ( queries ) 

 

두 절의 다른 점은 query 절에는 relevant score 를 계산하는 full-text query 로만 구성해야 하며

 

"How well does this document match this query clause"

 

filter 절에 넣은 query 는 작동시 Releavant Score 를 사용하지 않고 단순히 일치 / 불일치만 계산하므로

 

“Does this document match this query clause?” 

 

filter 절에는 이런 특성을 가지는 term level query 로 구성해야 한다.

 

예를 들어 아래 compound query 는 must 라는 query 절과 filter 라는 filter 절을 포함하며

 

must query 절에는 match 라는 full-text query  ( relevant score 고려한 결과를 내는 쿼리 ) 를 사용하였고

 

filter 절에는 term , range 처럼 term level query 를 사용하였다.

 

GET /my_index/default/_search {   "query": {     "bool": {       "must": [         { "match": { "title":   "Search"        }},         { "match": { "content": "Elasticsearch" }}         ],       "filter": [         { "term":  { "status": "published" }},         { "range": { "publish_date": { "gte": "2015-01-01" }}}       ]     }   } }

 

https://www.elastic.co/guide/en/elasticsearch/reference/5.5/query-filter-context.html

 

 

 

bool query ( bool )

 

bool query 는 총 4가지 옵션 필터를 제공하는데 must, must_not, should, filter 옵션이다.

 

이중 must, must_not, should 는 query type context 로 구성해야 하며

 

filter 옵션은 filter type context 로 구성해야 한다.

 

각 옵션의 의미는 아래와 같다.

 

must : 해당 절에 지정된 모든 query 결과가  true 가 되어야 가 일치 document 으로 간주한다.

 

must_not : 해당 절에 지정된 모든 query 결과가 false 가 되어야 일치 document 으로 간주한다.

 

should  : 해당 절에 지정된 query 가 많이 일치할 수록 Relevance Score 를 높혀 검색 순위를 높힌다.

 

filter  : 해당 절에 지정된 모든 query 가 true 가 되어야 일치 document 로 간주한다.

          단 이 절은 relevance score 계산에 포함되지 않는 filter type context 이다.

 

 

must, filter

 

이중 먼저 가장 많이 사용되는 must , filter 옵션을 사용해 보자.

 

아래와 같이 테스트 해보면 검색 결과로 ingredients.name 이 parmesan 이 포함되고 

 

perparation_time_minutes 가 15 이하인 행들이 검색될 것이다.

 

GET /recipe/default/_search {   "query": {     "bool": {       "must": [         {           "match":{             "ingredients.name":"parmesan"           }         }       ],       "filter": [         {           "range":{             "preparation_time_minutes": {               "lte":15             }           }         }       ]     }   } }

{   "took" : 10,   "timed_out" : false,   "_shards" : {     "total" : 5,     "successful" : 5,     "skipped" : 0,     "failed" : 0   },   "hits" : {     "total" : 3,     "max_score" : 1.447008,     "hits" : [       {         "_index" : "recipe",         "_type" : "default",         "_id" : "1",         "_score" : 1.447008,         "_source" : {           "title" : "....",           "preparation_time_minutes" : 12,           "servings" : {             "min" : 4,             "max" : 6           },           "steps" : [             "...."           ],           "ingredients" : [            ...             {               "name" : "Parmesan cheese"             }           ],           ...         }       },       {         "_index" : "recipe",         "_type" : "default",         "_id" : "10",         "_score" : 1.0727434,         "_source" : {           "title" : "Penne With Hot-As-You-Dare Arrabbiata Sauce",           "description" : "....",           "preparation_time_minutes" : 15,           "servings" : {             "min" : 4,             "max" : 4           },           "steps" : [             "....            ],           "ingredients" : [             ...             {               "name" : "Finely grated Parmesan cheese",               "quantity" : "60g"             },             {               "name" : "Minced flat-leaf parsley leaves",               "quantity" : "Small handful"             }           ],           ...         }       },       {         "_index" : "recipe",         "_type" : "default",         "_id" : "11",         "_score" : 0.44000342,         "_source" : {           "title" : "Spaghetti Puttanesca (Pasta or Spaghetti With Capers, Olives, and Anchovies)",           "description" : "....",           "preparation_time_minutes" : 15,           "servings" : {             "min" : 2,             "max" : 3           },           "steps" : [             "...."           ],           "ingredients" : [             ...             {               "name" : "Finely grated Pecorino Romano or Parmesan cheese",               "quantity" : "30g"             },             {               "name" : "Freshly ground black pepper"             },             {               "name" : "Can oil-packed tuna",               "quantity" : "140g"             }           ],           ...         }       }     ]   } }

 

 

must_not

 

여기에 더해 이번에는 tuna 가 포함되지 않는 결과만 보고 싶다고 하면 아래와 같이 하면 되겠다.

 

GET /recipe/default/_search {   "query": {     "bool": {       "must": [         {           "match":{             "ingredients.name":"parmesan"           }         }       ],       "must_not": [         {           "match":{             "ingredients.name":"tuna"           }         }       ],       "filter": [         {           "range":{             "preparation_time_minutes": {               "lte":15             }           }         }       ]     }   } }

{   "took" : 10,   "timed_out" : false,   "_shards" : {     "total" : 5,     "successful" : 5,     "skipped" : 0,     "failed" : 0   },   "hits" : {     "total" : 2,     "max_score" : 1.447008,     "hits" : [       {         "_index" : "recipe",         "_type" : "default",         "_id" : "1",         "_score" : 1.447008,         "_source" : {           "title" : "Fast and Easy Pasta With Blistered Cherry Tomato Sauce",           "description" : "..."           "preparation_time_minutes" : 12,           "servings" : {             "min" : 4,             "max" : 6           },           "steps" : [             "..."           ],           "ingredients" : [             {               "name" : "Dry pasta",               "quantity" : "450g"             },             {               "name" : "Kosher salt"             },             {               "name" : "Cloves garlic",               "quantity" : "4"             },             {               "name" : "Extra-virgin olive oil",               "quantity" : "90ml"             },             {               "name" : "Cherry tomatoes",               "quantity" : "750g"             },             {               "name" : "Fresh basil leaves",               "quantity" : "30g"             },             {               "name" : "Freshly ground black pepper"             },             {               "name" : "Parmesan cheese"             }           ],           "created" : "2017/03/29",           "ratings" : [             4.5,             5.0,             3.0,             4.5           ]         }       },       {         "_index" : "recipe",         "_type" : "default",         "_id" : "10",         "_score" : 1.0727434,         "_source" : {           "title" : "Penne With Hot-As-You-Dare Arrabbiata Sauce",           "description" : "..."           "preparation_time_minutes" : 15,           "servings" : {             "min" : 4,             "max" : 4           },           "steps" : [              "..."           ],           "ingredients" : [             {               "name" : "Kosher salt"             },             {               "name" : "Penne pasta",               "quantity" : "450g"             },             {               "name" : "Extra-virgin olive oil",               "quantity" : "3 tablespoons"             },             {               "name" : "Clove garlic",               "quantity" : "1"             },             {               "name" : "Crushed red pepper"             },             {               "name" : "Can whole peeled tomatoes",               "quantity" : "400g"             },             {               "name" : "Finely grated Parmesan cheese",               "quantity" : "60g"             },             {               "name" : "Minced flat-leaf parsley leaves",               "quantity" : "Small handful"             }           ],           ...         }       }     ]   } }

 

 

should

 

마지막으로 파슬리를 선호한다고 가정하면 아래와 같이 적용해 보자. 

 

기존 2 개 결과가 서로 순서가 바뀌는 것을 확인할 수 있게 된다.

 

GET /recipe/default/_search {   "query": {     "bool": {       "must": [         {           "match":{             "ingredients.name":"parmesan"           }         }       ],       "must_not": [         {           "match":{             "ingredients.name":"tuna"           }         }       ],       "should": [         {           "match":{             "ingredients.name":"parsley"           }         }       ],       "filter": [         {           "range":{             "preparation_time_minutes": {               "lte":15             }           }         }       ]     }   } }

{   "took" : 9,   "timed_out" : false,   "_shards" : {     "total" : 5,     "successful" : 5,     "skipped" : 0,     "failed" : 0   },   "hits" : {     "total" : 2,     "max_score" : 1.7949219,     "hits" : [       {         "_index" : "recipe",         "_type" : "default",         "_id" : "10",         "_score" : 1.7949219,         "_source" : {           "title" : "Penne With Hot-As-You-Dare Arrabbiata Sauce",           "description" : "..."           "preparation_time_minutes" : 15,           "servings" : {             "min" : 4,             "max" : 4           },           "steps" : [             "..."           ],           "ingredients" : [             {               "name" : "Kosher salt"             },             {               "name" : "Penne pasta",               "quantity" : "450g"             },             {               "name" : "Extra-virgin olive oil",               "quantity" : "3 tablespoons"             },             {               "name" : "Clove garlic",               "quantity" : "1"             },             {               "name" : "Crushed red pepper"             },             {               "name" : "Can whole peeled tomatoes",               "quantity" : "400g"             },             {               "name" : "Finely grated Parmesan cheese",               "quantity" : "60g"             },             {               "name" : "Minced flat-leaf parsley leaves",               "quantity" : "Small handful"             }           ],           ...       },       {         "_index" : "recipe",         "_type" : "default",         "_id" : "1",         "_score" : 1.447008,         "_source" : {           "title" : "Fast and Easy Pasta With Blistered Cherry Tomato Sauce",           "description" : "..."           "preparation_time_minutes" : 12,           "servings" : {             "min" : 4,             "max" : 6           },           "steps" : [             "..."           ],           "ingredients" : [             {               "name" : "Dry pasta",               "quantity" : "450g"             },             {               "name" : "Kosher salt"             },             {               "name" : "Cloves garlic",               "quantity" : "4"             },             {               "name" : "Extra-virgin olive oil",               "quantity" : "90ml"             },             {               "name" : "Cherry tomatoes",               "quantity" : "750g"             },             {               "name" : "Fresh basil leaves",               "quantity" : "30g"             },             {               "name" : "Freshly ground black pepper"             },             {               "name" : "Parmesan cheese"             }           ],           ...         }       }     ]   } }

 

디버깅을 위해서는 _name 필드를 사용하면 된다.

 

GET /recipe/default/_search {   "query": {     "bool": {       "must": [         {           "match":{             "ingredients.name":{               "query":"parmesan",               "_name":"parmesan_must"             }           }         }       ],       "must_not": [         {           "match":{             "ingredients.name":{               "query":"tuna",               "_name":"tuna_must_not"             }           }         }       ],       "should": [         {           "match":{             "ingredients.name":{               "query":"parsley",               "_name":"parsley_should"             }           }         }       ],       "filter": [         {           "range":{             "preparation_time_minutes": {               "lte":15,               "_name":"preparation_time_filter"             }           }         }       ]     }   } }

{   "took" : 10,   "timed_out" : false,   "_shards" : {     "total" : 5,     "successful" : 5,     "skipped" : 0,     "failed" : 0   },   "hits" : {     "total" : 2,     "max_score" : 1.7949219,     "hits" : [       {         "_index" : "recipe",         "_type" : "default",         "_id" : "10",         "_score" : 1.7949219,         "_source" : {           "title" : "Penne With Hot-As-You-Dare Arrabbiata Sauce",           "description" : "..."           "preparation_time_minutes" : 15,           "servings" : {             "min" : 4,             "max" : 4           },           "steps" : [             "..."           ],           "ingredients" : [             {               "name" : "Kosher salt"             },             {               "name" : "Penne pasta",               "quantity" : "450g"             },             {               "name" : "Extra-virgin olive oil",               "quantity" : "3 tablespoons"             },             {               "name" : "Clove garlic",               "quantity" : "1"             },             {               "name" : "Crushed red pepper"             },             {               "name" : "Can whole peeled tomatoes",               "quantity" : "400g"             },             {               "name" : "Finely grated Parmesan cheese",               "quantity" : "60g"             },             {               "name" : "Minced flat-leaf parsley leaves",               "quantity" : "Small handful"             }           ],           "created" : "2017/04/27",           "ratings" : [             1.5,             2.0,             4.0,             3.5,             3.0,             5.0,             1.5           ]         },         "matched_queries" : [           "parmesan_must",           "preparation_time_filter",           "parsley_should"         ]       },       {         "_index" : "recipe",         "_type" : "default",         "_id" : "1",         "_score" : 1.447008,         "_source" : {           "title" : "Fast and Easy Pasta With Blistered Cherry Tomato Sauce",           "description" : "..."           "preparation_time_minutes" : 12,           "servings" : {             "min" : 4,             "max" : 6           },           "steps" : [             "..."             ],           "ingredients" : [             {               "name" : "Dry pasta",               "quantity" : "450g"             },             {               "name" : "Kosher salt"             },             {               "name" : "Cloves garlic",               "quantity" : "4"             },             {               "name" : "Extra-virgin olive oil",               "quantity" : "90ml"             },             {               "name" : "Cherry tomatoes",               "quantity" : "750g"             },             {               "name" : "Fresh basil leaves",               "quantity" : "30g"             },             {               "name" : "Freshly ground black pepper"             },             {               "name" : "Parmesan cheese"             }           ],           "created" : "2017/03/29",           "ratings" : [             4.5,             5.0,             3.0,             4.5           ]         },         "matched_queries" : [           "parmesan_must",           "preparation_time_filter"         ]       }     ]   } }

 

 

 

weight

 

boost 옵션으로 각 쿼리마다 가산점 비율을 조절할 수 있다. ( 기본값은 1 )

 

1보다 크면 점수가 높게 오르고 1보다 적으면 점수가 낮게 오른다.

 

하지만 2를 준다고해서 점수가 2배로 오르는 것은 아니며, 정규화를 거친 후 적용된다.

 

GET /_search {   "query": {     "bool": {       "should": [         {           "match": {             "title": {               "query": "test1",               "boost": 1.2             }           }         },         {           "match": {             "keyword": {               "query": "test2",               "boost": 2             }           }         }       ]     }   } }