ElasticSearch 5 Sort by Keyword Field Case Insensitive - elasticsearch

Question

We are using ElasticSearch 5.
I have a field city using a custom analyzer and the following mapping.
Analyzer
"analysis": {
"analyzer": {
"lowercase_analyzer": {
"filter": [
"standard",
"lowercase",
"trim"
],
"type": "custom",
"tokenizer": "keyword"
}
}
Mapping
"city": {
"type": "text",
"analyzer": "lowercase_analyzer"
}
I am doing this so that I can do a case insensitive sort on the city field. Here is an example query that I am trying to run
{
"query": {
"term": {
"email": {
"value": "some_email#test.com"
}
}
},
"sort": [
{
"city": {
"order": "desc"
}
}
]
}
Here is the error I am getting:
"Fielddata is disabled on text fields by default. Set fielddata=true
on [city] in order to load fielddata in memory by uninverting the
inverted index. Note that this can however use significant memory."
I don't want to turn on FieldData and incur a performance hit in ElasticSearch. I would like to have a Keyword field that is not case sensitive, so that I can perform more meaningful aggregations and sorts on it. Is there no way to do this?

Answer 1

Yes, there is a way to do this, using multi_fields.
In Elasticsearch 5.0 onwards, string field types were split out into two separate types, text field types that are analyzed and can be used for search, and keyword field types that are not analyzed and are suited to use for sorting, aggregations and exact value matches.
With dynamic mapping in Elasticsearch 5.0 (i.e. let Elasticsearch infer the type that a document property should be mapped to), a json string property is mapped to a text field type, with a sub-field of "keyword" that is mapped as a keyword field type and the setting ignore_above:256.
With NEST 5.x automapping, a string property on your POCO will be automapped in the same way as dynamic mapping in Elasticsearch maps it as per above e.g. given the following document
public class Document
{
public string Property { get; set; }
}
automapping it
var pool = new SingleNodeConnectionPool(new Uri("http://localhost:9200"));
var defaultIndex = "default-index";
var connectionSettings = new ConnectionSettings(pool)
.DefaultIndex(defaultIndex);
var client = new ElasticClient(connectionSettings);
client.CreateIndex(defaultIndex, c => c
.Mappings(m => m
.Map<Document>(mm => mm
.AutoMap()
)
)
);
produces
{
"mappings": {
"document": {
"properties": {
"property": {
"fields": {
"keyword": {
"ignore_above": 256,
"type": "keyword"
}
},
"type": "text"
}
}
}
}
}
You can now use property for sorting using Field(f => f.Property.Suffix("keyword"). Take a look at Field Inference for more examples.
keyword field types have doc_values enabled by default, which means that a columnar data structure is built at index time and this is what provides efficient sorting and aggregations.
To add a custom analyzer at index creation time, we can automap as before, but then provide overrides for fields that we want to control the mapping for with .Properties()
client.CreateIndex(defaultIndex, c => c
.Settings(s => s
.Analysis(a => a
.Analyzers(aa => aa
.Custom("lowercase_analyzer", ca => ca
.Tokenizer("keyword")
.Filters(
"standard",
"lowercase",
"trim"
)
)
)
)
)
.Mappings(m => m
.Map<Document>(mm => mm
.AutoMap()
.Properties(p => p
.Text(t => t
.Name(n => n.Property)
.Analyzer("lowercase_analyzer")
.Fields(f => f
.Keyword(k => k
.Name("keyword")
.IgnoreAbove(256)
)
)
)
)
)
)
);
which produces
{
"settings": {
"analysis": {
"analyzer": {
"lowercase_analyzer": {
"type": "custom",
"filter": [
"standard",
"lowercase",
"trim"
],
"tokenizer": "keyword"
}
}
}
},
"mappings": {
"document": {
"properties": {
"property": {
"type": "text",
"fields": {
"keyword": {
"type": "keyword",
"ignore_above": 256
}
},
"analyzer": "lowercase_analyzer"
}
}
}
}
}