在弹性搜索中将子过滤器应用于术语的正确方法是什么？

Question

I have a query I am trying to fetch results via Elastic Search 6.4.2.

It is working. But when I apply the Tags part which should be under the NewsArticle type, it brings back results for the CaseStudyPage type.

Am I doing this correctly?

Basically I want to sub filter tags on the NewsArticle type but it doesn't seem to work correctly.

I have tried many different formats which I have found on StackOverflow and various web pages.

   "highlight":{  
      "pre_tags":[  
         ""
      ],
      "post_tags":[  
         "<\/strong>"
      ],
      "fields":{  
         "*":{  

         }
      },
      "require_field_match":false,
      "fragment_size":100,
      "number_of_fragments":3,
      "highlight_query":{  
         "query_string":{  
            "query":"",
            "analyze_wildcard":true,
            "default_operator":"AND"
         }
      }
   },
   "sort":[{"PublishedDate":"desc"}],
   "size":5000,
   "query":{  
      "bool":{  
         "filter":{  
            "bool":{  
               "should":[
                  {
                     "terms":{  
                        "ClassName":[  
                           "CaseStudyPage"
                        ]
                     }

                  },
                  {
                     "bool":{ 
                        "must": [
                           {
                              "terms":{  
                                 "ClassName":[
                                    "NewsArticle"
                                 ]
                              }
                           },
                           {
                              "terms":{  
                                 "Tags.ID":[
                                    "9"
                                 ]
                              }
                           }
                        ]
                     }
                  }
               ]
            }
         }
      }
   }
}

No error messages. It brings back results for both CaseStudyPage and NewsArticle which both have the Tag.ID = 9. But it should only bring back CaseStudyPage (Full results) and NewsArticle results that only have the Tag.ID = 9.

The results return tagged content from the CaseStudyPage type and the NewsArticle type, but it should only display the tagged content from the NewsArticle type and all the CaseStudyPage type.

Answer 1

First of all, for debuging purposes, I'd remove the "highlight" section & also the "sort" (I'm not sure if a document gets filtered out if the sort field - in your case 'PublishedDate' is empty? probably not)

Now, focusing on the query itself; it appears that your problem (If i undrestood correctly) is that the "Tags flter term query" is not working, because you are receiving in your result all CaseStudyPage and NewsArticle, even tough for those last ones you want specifically the ones with a Tags.ID = 9 (right?).

I believe your Tags.ID is an integer type right?, if so please remove the quotes arround the 9 (if you didn't create a mapping before indexing specifying your Tags.ID was indeed an "int" type, then the elasticsearch created a type mapping for the field based on your first insert, please verify that you Tags.ID is either a "not analyzed String - KEYWORD" or an integer; this is necessary for filter queries & terms queries to work properly).

Another possibility could be that the 'className' field contains several values? for example CaseStudyPage and NewsArticle both like an array? if so; then your should block will pick all those documents containing "CaseStudyPage" independently on the "NewsArticle" being prsent, or the TAGS.ID value. But this is rather unlikely.

I'd create a small filter query where I'd only test that you are correctly filtering the documents with the Tags.ID = 9; and after you've got that working, the you can put that filter again in the must block. The rest of your query looks fine.

EDIT: btw if for the 'CaseStudyPage' filter, you explicitly want entries that are not tagged, then you also need to add a' must not' block for that, or must with !=

Hope it helps.