Elasticsearch nGram search query containing blanks - elasticsearch

Question

I have created the following index:
{
"settings":{
"number_of_shards":1,
"number_of_replicas":0,
"blocks":{
"read_only_allow_delete":false,
"read_only":false
},
"analysis":{
"filter":{
"autocomplete_filter":{
"type":"ngram",
"min_gram":3,
"max_gram":30
}
},
"analyzer":{
"autocomplete":{
"type":"custom",
"tokenizer":"standard",
"filter":[
"lowercase",
"autocomplete_filter"
]
}
}
}
},
"mappings":{
"movie":{
"properties":{
"title":{
"type":"text"
},
"actors":{
"type":"nested",
"include_in_all":true,
"properties":{
"name":{
"type":"text",
"analyzer":"autocomplete",
"search_analyzer": "standard"
},
"age":{
"type":"long",
"index":"false"
}
}
}
}
}
}
}
And I have inserted the following data via _bulk endpoint:
{"index":{"_index":"movies","_type":"movie","_id":1}}
{"title":"Ocean's 11", "actors":[{"name":"Brad Pitt","age":54}, {"name":"George Clooney","age":56}, {"name":"Julia Roberts","age":50}, {"name":"Andy Garcia","age":61}]}
{"index":{"_index":"movies","_type":"movie","_id":2}}
{"title":"Usual suspects", "actors":[{"name":"Kevin Spacey","age":58}, {"name":"Benicio del Toro","age":50}]}
{"index":{"_index":"movies","_type":"movie","_id":3}}
{"title":"Fight club", "actors":[{"name":"Brad Pitt","age":54}, {"name":"Edward Norton","age":48}, {"name":"Helena Bonham Carter","age":51}, {"name":"Jared Leto","age":46}]}
{"index":{"_index":"movies","_type":"movie","_id":24}
{"title":"Fight club", "actors":[{"name":"Brad Garrett","age":57}, {"name":"Ben Stiller","age":52}, {"name":"Robin Williams","age":63}]}
Now I want to search the index by actor name. For instance, when I search for brad I get all movies having an actor named brad, which is good.
But when I search for rad p I want only the movies with Brad Pitt, and not Brad Garrett, but I get Brad Garrett.
This is my search query :
{
"query":{
"nested":{
"path":"actors",
"query":{
"match":{
"actors.name":{
"query":"rad p",
"analyzer":"standard"
}
}
},
"inner_hits":{
}
}
}
}
Endpoint I am calling is
/movies/movie/_search?pretty
My question is, how to correctly implement the mentioned feature?
Thanks
BTW elasticsearch version is 6.1.0.

Answer 1

This is because of the standard tokenizer which will split the input into tokens based on whitespaces and punctuation, so Brad Pitt becomes brad and pitt and hence you will not have a token with rad p in it.
What you need to do is to change the tokenizer to (e.g.) keyword so that you consider the full input as one token which you can then apply ngram on.
Or easier, you can simply use the ngram tokenizer and not the token filter

Answer 2

As Val has said, you have to use the nGram tokenizer to do this, and I also had to change my search query to:
{
"query":{
"nested":{
"path":"searchable",
"query":{
"bool":{
"must":{
"match":{
"searchable.searchKeyword":{
"query":"%1$s"
}
}
}
}
},
"inner_hits":{
}
}
}
}
My new index with the nGram tokenizer:
{
"number_of_shards":1,
"number_of_replicas":0,
"blocks":{
"read_only_allow_delete":false,
"read_only":false
},
"analysis":{
"analyzer":{
"autocomplete":{
"tokenizer":"search_tokenizer",
"filter":[
"lowercase",
"asciifolding"
]
}
},
"tokenizer":{
"search_tokenizer":{
"type":"ngram",
"token_chars":[
"letter",
"digit",
"whitespace",
"punctuation",
"symbol"
],
"min_gram":3,
"max_gram":30
}
}
}
}