如何使用 NEST 进行过滤聚合？

Question

I have a requirement for filter aggregations using NEST. But since I don't know much about this, I have made the below:

class Program
{
        static void Main(string[] args)
        {

            ISearchResponse<TestReportModel> searchResponse =
                            ConnectionToES.EsClient()
                            .Search<TestReportModel>
                            (s => s
                                .Index("feedbackdata")
                                .From(0)
                                .Size(50000)
                                .Query(q =>q.MatchAll())
                            );
                            
            var testRecords = searchResponse.Documents.ToList<TestReportModel>();
                
            result = ComputeTrailGap(testRecords);

        }
        
        
        private static List<TestModel> ComputeTrailGap(List<TestReportModel> testRecords)
        {
            var objTestModel = new List<TestModel>();          
            
            var ptpDispositionCodes = new string[] { "PTP" };
            var bptpDispositionCodes = new string[] { "BPTP","SBPTP" };
                  

            int gapResult = testRecords.Where(w => w.trailstatus == "Gap").Count();           

            var ptpResult = testRecords.Where(w => ptpDispositionCodes.Contains(w.lastdispositioncode)).ToList().Count();

            var bptpResult = testRecords.Where(w => bptpDispositionCodes.Contains(w.lastdispositioncode)).ToList().Count();
           

            objTestModel.Add(new TestModel { TrailStatus = "Gap", NoOfAccounts = gapResult });           
            objTestModel.Add(new TestModel { TrailStatus = "PTP", NoOfAccounts = ptpResult });
            objTestModel.Add(new TestModel { TrailStatus = "BPTP", NoOfAccounts = bptpResult });         

            return objTestModel;
        }
}

DTO

public class TestReportModel
{
   
    public string trailstatus { get; set; }        
    public string lastdispositioncode { get; set; }        
}

public class TestOutputAPIModel
{
    public List<TestModel> TestModelDetail { get; set; }    
}

public class TestModel
{   
    public string TrailStatus { get; set; }
    public int NoOfAccounts { get; set; }
    
}

This program works but as can be figure out that we are only accessing the Elastic Search via NEST and the rest of the Aggregations /Filter are done using Lambda.

I would like to perform the entire operation (Aggregations /Filter etc) using NEST framework and put it in the TestModel using filter aggregation.

How can I construct the DSL query inside NEST?

Update

I have been able to make the below so far but the count is zero. What is wrong in my query construction?

var ptpDispositionCodes = new TermsQuery
            {
                IsVerbatim = true,
                Field = "lastdispositioncode",
                Terms = new string[] { "PTP" },
            };
var bptpDispositionCodes = new TermsQuery
{
    IsVerbatim = true,
    Field = "lastdispositioncode",
    Terms = new string[] { "BPTP" },
};

 ISearchResponse<TestReportModel> searchResponse =
                ConnectionToES.EsClient()
                .Search<TestReportModel>
                (s => s
                    .Index("feedbackdata")
                    .From(0)
                    .Size(50000)
                    .Query(q =>q.MatchAll())
                    .Aggregations(fa => fa

                        .Filter("ptp_aggs", f => f.Filter(fd => ptpDispositionCodes))
                        .Filter("bptp_aggs", f => f.Filter(fd => bptpDispositionCodes))

                        )
               );

Result

Answer 1

I see that you are trying to perform a search on the type of TestReportModel . The overall structure of your approach seems good enough. However, there is a trouble with the queries that are being attached to your filter containers.

Your TestReportModel contains two properties trialStatus and lastdispositioncode . You are setting the Field property as description inside your terms query. This is the reason that you are seeing the counts as zero. The model on which you are performing the search on (in turn the index that you are performing the search on) does not have a property description and hence the difference. NEST or Elasticsearch, in this case does not throw any exception. Instead, it returns count zero. Field value should be modified to lastdispositioncode .

// Type on which the search is being performed. 
// Response is of the type ISearchResponse<TestReportModel>
public class TestReportModel
{
    public string trailstatus { get; set; }        
    public string lastdispositioncode { get; set; }        
}

Modified terms queries are as follows

// Field is "lastdispositioncode" and not "description"
// You may amend the Terms field as applicable
var ptpDispositionCodes = new TermsQuery
{
    IsVerbatim = true,
    Field = "lastdispositioncode",
    Terms = new string[] { "PTP" },
};

var bptpDispositionCodes = new TermsQuery
{
    IsVerbatim = true,
    Field = "lastdispositioncode",
    Terms = new string[] { "BPTP", "SBPTP" },
};

Since it seems that the values to lastdispositioncode seem to take a single word value (PTP or BPTP from your examples), I believe, it is not going to matter if the field in the doc is analyzed or not. You can further obtain the counts from the ISearchResponse<T> type as shown below

var ptpDocCount = ((Nest.SingleBucketAggregate)response.Aggregations["ptp_aggs"]).DocCount;
var bptpDocCount = ((Nest.SingleBucketAggregate)response.Aggregations["bptp_aggs"]).DocCount;

Edit: Adding an approach for keyword search

QueryContainer qc1 = new QueryContainerDescriptor<TestReportModel>()
    .Bool(b => b.Must(m => m.Terms(t => t.Field(f => f.lastdispositioncode.Suffix("keyword"))
        .Terms(new string[]{"ptp"}))));

QueryContainer qc2 = new QueryContainerDescriptor<TestReportModel>()
    .Bool(b => b.Must(m => m.Terms(t => t.Field(f => f.lastdispositioncode.Suffix("keyword"))
        .Terms(new string[]{"bptp", "sbptp"}))));

Now these query containers can be hooked to your aggregation as shown below

.Aggregations(aggs => aggs
    .Filter("f1", f => f.Filter(f => qc1))
    .Filter("f2", f => f.Filter(f => qc2)))

The queries for aggregations that are generated by the NEST client in this case look like below

"f1": {
      "filter": {
        "bool": {
          "must": [
            {
              "terms": {
                "lastdispositioncode.keyword": [
                  "bptp"
                ]
              }
            }
          ]
        }
      }
    }

Also, coming back to the case of search being case-insensitive, Elasticsearch deals with search in a case-insensitive fashion. However, it varies depending on analyzed vs non-analyzed fields. Analyzed fields are tokenized and text fields are by default tokenized. We use the suffix extension method of NEST on analyzed fields ideally and to get an exact match on the analyzed field. More about them here