search（16）- elastic4s-內嵌檔案：nested and join

雪川大蟲發表於2020-05-23

原文網址 : https://www.cnblogs.com/tiger-xc/p/12941077.html

從SQL領域來的使用者，對於ES的檔案關係維護方式會感到很不習慣。畢竟，ES是分散式資料庫只能高效處理獨個扁平型別檔案，無法支援關係式資料庫那樣的檔案拼接。但是，任何資料庫應用都無法避免樹型檔案關係，因為這是業務模式需要的表現形式。在ES裡，無論nested或join型別的資料，父-子關係的資料檔案實際上是放在同一個索引index裡的。在ES裡已經沒有資料表（doc_type）的概念。但從操作層面上ES提供了relation型別來支援父-子資料關係操作。所以，nested資料型別一般用來表達比較固定的嵌入資料。因為每次更新都需要重新對檔案進行一次索引。join型別的資料則可以對資料關係的兩頭分別獨立進行更新，方便很多。

下面我們現示範一下nested資料型別的使用。在mapping裡可以申明nested資料型別來代表嵌入檔案，如下：

  val fruitMapping = client.execute(
    putMapping("fruits").fields(
      KeywordField("code"),
      SearchAsYouTypeField("name")
        .fields(KeywordField("keyword")),
      floatField("price"),
      NestedField("location").fields(
        KeywordField("shopid"),
        textField("shopname"),
        longField("qty"))
      )
  ).await

這段程式碼產生了下面的mapping:

{
  "fruits" : {
    "mappings" : {
      "properties" : {
        "code" : {
          "type" : "keyword"
        },
        "location" : {
          "type" : "nested",
          "properties" : {
            "qty" : {
              "type" : "long"
            },
            "shopid" : {
              "type" : "keyword"
            },
            "shopname" : {
              "type" : "text"
            }
          }
        },
        "name" : {
          "type" : "text",
          "fields" : {
            "keyword" : {
              "type" : "keyword"
            }
          }
        },
        "price" : {
          "type" : "float"
        }
      }
    }
  }
}

location是個nested型別欄位，內嵌檔案格式含shopid,shopname,qty各欄位。下面的例子裡向fruits索引新增了幾個包含了location的檔案：

  val f1 = indexInto("fruits").id("f001")
      .fields(
        "code" -> "f001",
        "name" -> "東莞荔枝",
        "price" -> 11.5,
        "location" -> List(Map(
          "shopid" -> "s001",
          "shopname" -> "中心店",
          "qty" -> 500.0
          ),
          Map(
            "shopid" -> "s002",
            "shopname" -> "東門店",
            "qty" -> 0.0
          )
        )
      )
  val f2 = indexInto("fruits").id("f002")
    .fields(
      "code" -> "f002",
      "name" -> "陝西富士蘋果",
      "price" -> 11.5,
      "location" -> List(Map(
        "shopid" -> "s001",
        "shopname" -> "中心店",
        "qty" -> 300.0
      ),
        Map(
          "shopid" -> "s003",
          "shopname" -> "龍崗店",
          "qty" -> 200.0
        )
      )
    )
  val f3 = indexInto("fruits").id("f003")
    .fields(
      "code" -> "f003",
      "name" -> "進口菲律賓香蕉",
      "price" -> 5.3,
      "location" -> List(Map(
        "shopid" -> "s001",
        "shopname" -> "中心店",
        "qty" -> 300.0
      ),
        Map(
          "shopid" -> "s003",
          "shopname" -> "龍崗店",
          "qty" -> 200.0
        ),
        Map(
          "shopid" -> "s002",
          "shopname" -> "東門店",
          "qty" -> 200.0
        )
      )
    )
  val newIndex = for {
     _ <- client.execute(f1)
     _ <- client.execute(f2)
     _ <- client.execute(f3)
  } yield ("成功增添三條記錄")

  newIndex.onComplete {
    case Success(trb) => println(s"${trb}")
    case Failure(err) => println(s"error: ${err.getMessage}")
  }

用elastic4s可以比較方便的進行nested型別資料更新。下面是個更新nested檔案的例子：

  val f002 = client.execute(get("fruits","f002").fetchSourceInclude("location")).await
  val locs: List[Map[String,Any]] = f002.result.source("location").asInstanceOf[List[Map[String,Any]]]
  val newloc = Map("shopid" -> "s004","shopname" -> "寶安店", "qty" -> 23)
  val newlocs = locs.foldLeft(List[Map[String,Any]]()) { (b, m) =>
    if (m("shopid") != newloc("shopid"))
      m :: b
    else b
  }

  val newdoc = updateById("fruits","f002")
    .doc(
      Map(
        "location" -> (newloc :: newlocs)
      )
    )

在上面這個例子裡：需要把一條新的嵌入檔案s004更新到f002檔案裡。我們先把f002裡原來的location取出，去掉s004節點，然後將新節點加入location清單，再更新update f002檔案。

剛才提到過：join型別實際上還是在同一個索引裡實現的。比如我希望記錄每個fruit的進貨歷史，也就是說現在fruit下需要增加一個子檔案purchase_history。這個purchase_history也是在同一個mapping裡定義的：

  val fruitMapping = client.execute(
    putMapping("fruits").fields(
      KeywordField("code"),
      SearchAsYouTypeField("name")
        .fields(KeywordField("keyword")),
      floatField("price"),
      NestedField("location").fields(
        KeywordField("shopid"),
        textField("shopname"),
        longField("qty")),
//purchase_history
    keywordField("supplier_code"),
    textField("supplier_name"),
    dateField("purchase_date")
      .ignoreMalformed(true)
      .format("strict_date_optional_time||epoch_millis"),
    joinField("purchase_history")
      .relation("fruit","purchase")
    )
  ).await

下面是關於上層父檔案的索引indexing操作的例子：

  val f1 = indexInto("fruits").id("f001").routing("f001")
      .fields(
        "code" -> "f001",
        "name" -> "東莞荔枝",
        "price" -> 11.5,
        "location" -> List(Map(
          "shopid" -> "s001",
          "shopname" -> "中心店",
          "qty" -> 500.0
          ),
          Map(
            "shopid" -> "s002",
            "shopname" -> "東門店",
            "qty" -> 0.0
          )
        ),
        "purchase_history" -> "fruit"
      )
  val f2 = indexInto("fruits").id("f002").routing("f002")
    .fields(
      "code" -> "f002",
      "name" -> "陝西富士蘋果",
      "price" -> 11.5,
      "location" -> List(Map(
        "shopid" -> "s001",
        "shopname" -> "中心店",
        "qty" -> 300.0
      ),
        Map(
          "shopid" -> "s003",
          "shopname" -> "龍崗店",
          "qty" -> 200.0
        )
      ),
      "purchase_history" -> "fruit"
    )
  val f3 = indexInto("fruits").id("f003").routing("f003")
    .fields(
      "code" -> "f003",
      "name" -> "進口菲律賓香蕉",
      "price" -> 5.3,
      "location" -> List(Map(
        "shopid" -> "s001",
        "shopname" -> "中心店",
        "qty" -> 300.0
      ),
        Map(
          "shopid" -> "s003",
          "shopname" -> "龍崗店",
          "qty" -> 200.0
        ),
        Map(
          "shopid" -> "s002",
          "shopname" -> "東門店",
          "qty" -> 200.0
        )
      ),
      "purchase_history" -> "fruit"
    )
  val newIndex = for {
     _ <- client.execute(f1)
     _ <- client.execute(f2)
     _ <- client.execute(f3)
  } yield ("成功增添三條記錄")

elastic4s子檔案的索引操作示範如下：

  val h1 = indexInto("fruits").id("h001").routing("f003")
    .fields(
      "supplier_code" -> "v001",
      "supplier_name" -> "百果園",
      "purchase_date" -> "2020-02-09",
      "purchase_history" -> Child("purchase", "f003"))

  val h2 = indexInto("fruits").id("h002").routing("f002")
    .fields(
      "supplier_code" -> "v001",
      "supplier_name" -> "百果園",
      "purchase_date" -> "2019-10-11",
      "purchase_history" -> Child("purchase", "f002"))

  val h3 = indexInto("fruits").id("h003").routing("f002")
    .fields(
      "supplier_code" -> "v002",
      "supplier_name" -> "華南城花果批發市場",
      "purchase_date" -> "2020-01-23",
      "purchase_history" -> Child("purchase", "f002"))


  val childIndex = for {
    _ <- client.execute(h1)
    _ <- client.execute(h2)
    _ <- client.execute(h3)
  } yield ("成功增添三條子記錄")

好了，現在這個fruits索引裡已經包含了nested,join兩種嵌入檔案資料。下面我們就試試各種的讀取方式。首先nested型別資料可以通過nestedQuery讀取：

  val qNested = search("fruits").query(
    nestedQuery("location").query(
      matchQuery("location.shopname","中心")
    )
  )
  println(s"${qNested.show}")
  val nestedResult = client.execute(qNested).await
  if(nestedResult.isSuccess)
    nestedResult.result.hits.hits.foreach(m => println(s"${m.sourceAsMap}"))
  else println(s"Error: ${nestedResult.error.causedBy.getOrElse("unknown")}")

...

POST:/fruits/_search?
StringEntity({"query":{"nested":{"path":"location","query":{"match":{"location.shopname":{"query":"中心"}}}}}},Some(application/json))
HashMap(name -> 東莞荔枝, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 500.0), Map(shopid -> s002, shopname -> 東門店, qty -> 0.0)), price -> 11.5, purchase_history -> fruit, code -> f001)
HashMap(name -> 進口菲律賓香蕉, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 300.0), Map(shopid -> s003, shopname -> 龍崗店, qty -> 200.0), Map(shopid -> s002, shopname -> 東門店, qty -> 200.0)), price -> 5.3, purchase_history -> fruit, code -> f003)
HashMap(name -> 陝西富士蘋果, location -> List(Map(shopname -> 寶安店, qty -> 23, shopid -> s004), Map(shopname -> 龍崗店, qty -> 200.0, shopid -> s003), Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)), price -> 11.5, purchase_history -> fruit, code -> f002)

join型別子檔案可以通過子檔案的ParentID Query讀取：

  val qPid = search("fruits").query(
     ParentIdQuery("purchase","f002")
  )
  println(s"${qPid.show}")

  val pidResult = client.execute(qPid).await
  if(pidResult.isSuccess)
    pidResult.result.hits.hits.foreach(m => println(s"${m.sourceAsMap}"))
  else println(s"Error: ${pidResult.error.causedBy.getOrElse("unknown")}")

...

POST:/fruits/_search?
StringEntity({"query":{"parent_id":{"type":"purchase","id":"f002"}}},Some(application/json))
Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2019-10-11, purchase_history -> Map(name -> purchase, parent -> f002))
Map(supplier_code -> v002, supplier_name -> 華南城花果批發市場, purchase_date -> 2020-01-23, purchase_history -> Map(name -> purchase, parent -> f002))

join型別父輩檔案可以通過搜尋其子檔案hasChild獲取：

  val qHaschild = search("fruits").query(
     hasChildQuery("purchase",
       matchQuery("supplier_name","百果")
     )
  )
  println(s"${qHaschild.show}")
  val haschildResult = client.execute(qHaschild).await
  if(haschildResult.isSuccess)
    haschildResult.result.hits.hits.foreach(m => println(s"${m.sourceAsMap}"))
  else println(s"Error: ${haschildResult.error.causedBy.getOrElse("unknown")}")

...

POST:/fruits/_search?
StringEntity({"query":{"has_child":{"type":"purchase","score_mode":"none","query":{"match":{"supplier_name":{"query":"百果"}}}}}},Some(application/json))
HashMap(name -> 進口菲律賓香蕉, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 300.0), Map(shopid -> s003, shopname -> 龍崗店, qty -> 200.0), Map(shopid -> s002, shopname -> 東門店, qty -> 200.0)), price -> 5.3, purchase_history -> fruit, code -> f003)
HashMap(name -> 陝西富士蘋果, location -> List(Map(shopname -> 寶安店, qty -> 23, shopid -> s004), Map(shopname -> 龍崗店, qty -> 200.0, shopid -> s003), Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)), price -> 11.5, purchase_history -> fruit, code -> f002)

join型別子檔案也可以搜尋其父輩檔案獲取：

 val qHasparent= search("fruits").query(
    hasParentQuery("fruit",
      nestedQuery("location").query(
        matchQuery("location.shopname","中心")
      ),false
    )
  )
  println(s"${qHasparent.show}")
  val hasparentResult = client.execute(qHasparent).await
  if(hasparentResult.isSuccess)
    hasparentResult.result.hits.hits.foreach(m => println(s"${m.sourceAsMap}"))
  else println(s"Error: ${hasparentResult.error.causedBy.getOrElse("unknown")}")

...

OST:/fruits/_search?
StringEntity({"query":{"has_parent":{"parent_type":"fruit","query":{"nested":{"path":"location","query":{"match":{"location.shopname":{"query":"中心"}}}}}}}},Some(application/json))
Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2020-02-09, purchase_history -> Map(name -> purchase, parent -> f003))
Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2019-10-11, purchase_history -> Map(name -> purchase, parent -> f002))
Map(supplier_code -> v002, supplier_name -> 華南城花果批發市場, purchase_date -> 2020-01-23, purchase_history -> Map(name -> purchase, parent -> f002))

上面這個例子稍微複雜一點：我們想得出所有子檔案，它們的父輩檔案裡嵌入nested檔案包含location.shopname match "中心"。

這些例子主要展示瞭如何通過父子關係的一方取獲取另一方的資料，如：通過子檔案搜尋獲取對應的父檔案或通過父檔案獲取對應的子檔案。也就是說搜尋目標和獲取目標：父子、子父，不是同一種檔案。我們可以通過inner_hits來同時獲取符合搜尋條件的檔案。如nestedQuery.inner()：

 val qNested = search("fruits").query(
    nestedQuery("location").query(
      matchQuery("location.shopname","中心")
    ).inner(InnerHit("locations"))
  )
  println(s"${qNested.show}")
  val nestedResult = client.execute(qNested).await
  if(nestedResult.isSuccess) {
    nestedResult.result.hits.hits.foreach{ m =>
        println(s"${m.sourceAsMap}")
        m.innerHits.foreach { i =>
          val n = i._1
          i._2.hits.foreach(h => println(s"$n, ${h.source}"))
        }
    }
  } else println(s"Error: ${nestedResult.error.causedBy.getOrElse("unknown")}")

...

POST:/fruits/_search?
StringEntity({"query":{"nested":{"path":"location","query":{"match":{"location.shopname":{"query":"中心"}}},"inner_hits":{"name":"locations"}}}},Some(application/json))
HashMap(name -> 東莞荔枝, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 500.0), Map(shopid -> s002, shopname -> 東門店, qty -> 0.0)), price -> 11.5, purchase_history -> fruit, code -> f001)
locations, Map(shopid -> s001, shopname -> 中心店, qty -> 500.0)
HashMap(name -> 進口菲律賓香蕉, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 300.0), Map(shopid -> s003, shopname -> 龍崗店, qty -> 200.0), Map(shopid -> s002, shopname -> 東門店, qty -> 200.0)), price -> 5.3, purchase_history -> fruit, code -> f003)
locations, Map(shopid -> s001, shopname -> 中心店, qty -> 300.0)
HashMap(name -> 陝西富士蘋果, location -> List(Map(shopname -> 寶安店, qty -> 23, shopid -> s004), Map(shopname -> 龍崗店, qty -> 200.0, shopid -> s003), Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)), price -> 11.5, purchase_history -> fruit, code -> f002)
locations, Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)

hasChildQuery.innerHit()：

  val qHaschild = search("fruits").query(
     hasChildQuery("purchase",
       matchQuery("supplier_name","百果")
     ).innerHit("purchases")
  )
  println(s"${qHaschild.show}")
  val haschildResult = client.execute(qHaschild).await
  if(haschildResult.isSuccess) {
    haschildResult.result.hits.hits.foreach{m =>
      println(s"${m.sourceAsMap}")
      m.innerHits.foreach { i =>
        val n = i._1
        i._2.hits.foreach(h => println(s"$n, ${h.source}"))
      }
    }
  } else println(s"Error: ${haschildResult.error.causedBy.getOrElse("unknown")}")

...

POST:/fruits/_search?
StringEntity({"query":{"has_child":{"type":"purchase","score_mode":"none","query":{"match":{"supplier_name":{"query":"百果"}}},"inner_hits":{"name":"purchases"}}}},Some(application/json))
HashMap(name -> 進口菲律賓香蕉, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 300.0), Map(shopid -> s003, shopname -> 龍崗店, qty -> 200.0), Map(shopid -> s002, shopname -> 東門店, qty -> 200.0)), price -> 5.3, purchase_history -> fruit, code -> f003)
purchases, Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2020-02-09, purchase_history -> Map(name -> purchase, parent -> f003))
HashMap(name -> 陝西富士蘋果, location -> List(Map(shopname -> 寶安店, qty -> 23, shopid -> s004), Map(shopname -> 龍崗店, qty -> 200.0, shopid -> s003), Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)), price -> 11.5, purchase_history -> fruit, code -> f002)
purchases, Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2019-10-11, purchase_history -> Map(name -> purchase, parent -> f002))
purchases, Map(supplier_code -> v002, supplier_name -> 華南城花果批發市場, purchase_date -> 2020-01-23, purchase_history -> Map(name -> purchase, parent -> f002))

hasParentQuery.innerHit():

  val qHasparent= search("fruits").query(
    hasParentQuery("fruit",
      nestedQuery("location").query(
        matchQuery("location.shopname","中心")
      ),false
    ).innerHit(InnerHit("fruits"))
  )
  println(s"${qHasparent.show}")
  val hasparentResult = client.execute(qHasparent).await
  if(hasparentResult.isSuccess) {
    hasparentResult.result.hits.hits.foreach{m =>
      println(s"${m.sourceAsMap}")
      m.innerHits.foreach { i =>
        val n = i._1
        i._2.hits.foreach(h => println(s"$n, ${h.source}"))
      }
    }
  } else println(s"Error: ${hasparentResult.error.causedBy.getOrElse("unknown")}")

...

POST:/fruits/_search?
StringEntity({"query":{"has_parent":{"parent_type":"fruit","query":{"nested":{"path":"location","query":{"match":{"location.shopname":{"query":"中心"}}}}},"inner_hits":{"name":"fruits"}}}},Some(application/json))
Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2020-02-09, purchase_history -> Map(name -> purchase, parent -> f003))
fruits, HashMap(name -> 進口菲律賓香蕉, location -> List(Map(shopid -> s001, shopname -> 中心店, qty -> 300.0), Map(shopid -> s003, shopname -> 龍崗店, qty -> 200.0), Map(shopid -> s002, shopname -> 東門店, qty -> 200.0)), price -> 5.3, purchase_history -> fruit, code -> f003)
Map(supplier_code -> v001, supplier_name -> 百果園, purchase_date -> 2019-10-11, purchase_history -> Map(name -> purchase, parent -> f002))
fruits, HashMap(name -> 陝西富士蘋果, location -> List(Map(shopname -> 寶安店, qty -> 23, shopid -> s004), Map(shopname -> 龍崗店, qty -> 200.0, shopid -> s003), Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)), price -> 11.5, purchase_history -> fruit, code -> f002)
Map(supplier_code -> v002, supplier_name -> 華南城花果批發市場, purchase_date -> 2020-01-23, purchase_history -> Map(name -> purchase, parent -> f002))
fruits, HashMap(name -> 陝西富士蘋果, location -> List(Map(shopname -> 寶安店, qty -> 23, shopid -> s004), Map(shopname -> 龍崗店, qty -> 200.0, shopid -> s003), Map(shopname -> 中心店, qty -> 300.0, shopid -> s001)), price -> 11.5, purchase_history -> fruit, code -> f002)

Mysql join 的本質，Nested Loop Join 和 Hash Join
2020-09-25
MySqlOOP
search（11）- elastic4s-模糊查詢
2020-05-10
AST
search（5）- elastic4s-構建索引
2020-04-18
AST索引
多表連線的三種方式詳解 hash join、merge join、 nested loop
2021-02-03
OOP
資料庫實現原理#1（Nested Loop Join）
2020-03-31
資料庫OOP
PDF Search for Macpdf檔案搜尋工具
2020-12-03
Mac
pdf檔案搜尋工具：PDF Search Mac
2022-03-27
Mac
PDF檔案搜尋工具PDF Search for Mac
2022-07-22
Mac
PDF Search for Mac(pdf檔案搜尋工具)
2022-07-18
Mac
Nginx 內嵌變數
2019-11-22
Nginx變數
java檔案相關（檔案追加內容、檔案內容清空、檔案內容讀取）
2018-06-29
Java
ElasticSearch多層nested查詢、nested過濾排除非結果內容
2020-11-29
Elasticsearch
可以把 FolkMQ 內嵌到 SpringBoot3 專案裡（可內嵌的訊息中介軟體，純血國產）
2024-06-20
MQSpring Boot
MongoDB查詢內嵌文件
2018-08-27
MongoDB
淺談小程式內嵌網頁及內嵌網頁跳轉分享實現
2019-03-04
網頁
內嵌樣式標記style
2018-10-31
小程式內嵌h5
2018-06-25
H5
Spring Boot 內嵌容器 Tomcat / Undertow
2021-09-09
Spring BootTomcat
Nodejs教程16：POST檔案上傳
2019-03-06
NodeJS
3dTiles 資料規範詳解[3] 內嵌在瓦片檔案中的兩大資料表
2020-07-06
3D大資料
PDF Search for Mac(pdf檔案搜尋工具)13.6啟用版
2023-11-05
Mac
SVG 立方體內嵌路徑拼接
2019-04-06
SVG
客戶端內嵌Vue頁面
2019-07-23
客戶端Vue
Android/iOS內嵌Unity開發示例
2020-07-09
AndroidiOSUnity
C++內嵌彙編教程1
2022-01-20
C++
Python_16 配置檔案與封裝
2023-05-03
Python封裝
2024年6月16日 Python - 檔案
2024-06-17
Python
Ashampoo Backup Pro 16，檔案備份工具
2021-10-13
vim內替換檔案內容
2018-05-09
使用ln同步檔案內容,支援忽略檔案
2020-12-19
WordPress模板層次16：搜尋模板search.php
2018-07-08
PHP
Centos 7 透過 targz 檔案安裝 Elastic Search 服務教程！
2024-03-15
CentOSAST
蘋果電腦上pdf檔案搜尋工具——PDF Search for Mac
2020-04-24
蘋果Mac
檔案內容比較
2024-06-18
AdWhirl：iPhone,Android軟體內嵌廣告
2020-04-07
iPhoneAndroid
學習Tomcat（七）之Spring內嵌Tomcat
2021-10-12
TomcatSpring
mysql INNER JOIN、LEFT JOIN、RIGHT JOIN；內連線（等值連線）、左連線、右連線
2024-11-04
MySql
使用iframe內嵌網頁的時候，如何做到內嵌網頁的高度自適應有大用
2024-03-20
網頁

search（16）- elastic4s-內嵌檔案：nested and join

相關文章