Elastic nest token_limit for phrase suggester

Question

It seems as though the latest nest client doesn't support the token_limit parameter for the phrase suggester? How can I extend NEST to add this without having to create a raw query?

Answer 1

The latest NEST client is missing token_limit for phrase suggester; I'll open a PR to add it.

In the meantime, you could support it and still use the fluent lambda API by deriving from PhraseSuggesterDescriptor<T> and implementing the token limit on an interface

private static void Main()
{
    var client = new ElasticClient();

    var searchResponse = client.Search<Question>(s => s
        .Size(0)
        .Suggest(su => su
            .Phrase("suggest_phrase", p => new MyPhraseSuggesterDescriptor<Question>()
                .TokenLimit(5)
                .Field(f => f.Title)
                .Highlight(h => h
                    .PreTag("<em>")
                    .PostTag("</em>")
                )
                .Text("dotnot entrity framework")
            )
        )
    );
}

public interface IMyPhraseSuggester : IPhraseSuggester
{
    [PropertyName("token_limit")]
    int? TokenLimit { get; set; }
}

public class MyPhraseSuggesterDescriptor<T> : PhraseSuggesterDescriptor<T>, IMyPhraseSuggester where T : class 
{
    int? IMyPhraseSuggester.TokenLimit { get; set; }

    public MyPhraseSuggesterDescriptor<T> TokenLimit(int tokenLimit)
    {
        ((IMyPhraseSuggester)this).TokenLimit = tokenLimit;
        return this;
    }
}

which serializes to

{
  "size": 0,
  "suggest": {
    "suggest_phrase": {
      "text": "dotnot entrity framework",
      "phrase": {
        "field": "title",
        "highlight": {
          "pre_tag": "<em>",
          "post_tag": "</em>"
        },
        "token_limit": 5
      }
    }
  }
}

NOTE that the TokenLimit() method call is made before any other calls, because other calls will return PhraseSuggesterDescriptor<T> and not MyPhraseSuggesterDescriptor<T>. You could go ahead and redefine all the properties of IPhraseSuggester on MyPhraseSuggesterDescriptor<T> if you wanted to, to avoid this API quirk, but that's probably more effort than it's worth to work around it for the moment :)