Elasticsearch by Example: Part 4
We refactor our solution for faceted queries to make it more scalable.
This article is part of a series, starting with Elasticsearch by Example: Part 1, exploring the Elasticsearch database / search engine.
The Document Structure (Revisited)
For our shirts example, we will update our representative document.
{
"entity": {
"name": "tshirt"
},
"keyword_facets": [
{
"facet_name": "size",
"facet_value": "S"
},
{
"facet_name": "color",
"facet_value": "black"
},
{
"facet_name": "fabric",
"facet_value": "cotton"
}
],
"long_facets": [
{
"facet_name": "price",
"facet_value": 1000
}
]
}Observations:
- Moved the descriptive name property under an entity property; this is where I anticipate placing other non-facet related properties.
- We divide the facet data based on types: keyword and long.
Create (Revisited)
Assuming that we are using the same cluster environment, we need to delete the existing shirts index.
DELETE: ENDPOINT/shirts
and then recreate it as follows:
PUT: ENDPOINT/shirts
{
"settings": {
"number_of_shards": 1,
"number_of_replicas": 1
},
"mappings": {
"shirt": {
"properties": {
"entity": {
"properties": {
"name": {
"type": "text"
}
}
},
"keyword_facets": {
"type": "nested",
"properties": {
"facet_name": {
"type": "keyword"
},
"facet_value": {
"type": "keyword"
}
}
},
"long_facets": {
"type": "nested",
"properties": {
"facet_name": {
"type": "keyword"
},
"facet_value": {
"type": "long"
}
}
}
}
}
}
}Observations:
- The mapping does not change as you add additional facets; exact same mapping if you have 50 facets.
- The nested type allows you to effectively use an array as a property.
As in our previous example, we can create a single document with the following URL:
POST: ENDPOINT/shirts/shirt
and in bulk with the URL:
POST: ENDPOINT/shirts/shirt/_bulk
note: Omitted the detail as there is nothing really new here.
Search (Revisited)
As the facet values are now buried in a list, searches are a bit more complex. Here we find small black shirts.
POST: ENDPOINT/shirts/shirt/_search
{
"query": {
"bool": {
"filter": [
{
"nested": {
"path": "keyword_facets",
"query": {
"bool": {
"filter": [
{ "term": { "keyword_facets.facet_name": "size" } },
{ "term": { "keyword_facets.facet_value": "S" } }
]
}
}
}
},
{
"nested": {
"path": "keyword_facets",
"query": {
"bool": {
"filter": [
{ "term": { "keyword_facets.facet_name": "color" } },
{ "term": { "keyword_facets.facet_value": "black" } }
]
}
}
}
}
]
}
}
}Aggregation (Revisited)
Say we wanted to obtain all of the sizes and colors (including counts) of cotton shirts.
POST: ENDPOINT/shirts/shirt/_search
{
"query": {
"nested": {
"path": "keyword_facets",
"query": {
"bool": {
"filter": [
{ "term": { "keyword_facets.facet_name": "fabric" } },
{ "term": { "keyword_facets.facet_value": "cotton" } }
]
}
}
}
},
"aggs": {
"agg_keyword_facet": {
"nested": {
"path": "keyword_facets"
},
"aggs": {
"facet_name": {
"terms": {
"field": "keyword_facets.facet_name"
},
"aggs": {
"facet_value": {
"terms": {
"field": "keyword_facets.facet_value"
}
}
}
}
}
}
}
}Observation: The big win here is that the aggregations query no longer needs to explicitly list the facets.
The aggregations section below the query results looking like:
...
"aggregations": {
"agg_keyword_facet": {
"doc_count": 18,
"facet_name": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "color",
"doc_count": 9,
"facet_value": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "black",
"doc_count": 6
},
{
"key": "red",
"doc_count": 3
}
]
}
},
{
"key": "size",
"doc_count": 9,
"facet_value": {
"doc_count_error_upper_bound": 0,
"sum_other_doc_count": 0,
"buckets": [
{
"key": "L",
"doc_count": 3
},
{
"key": "M",
"doc_count": 3
},
{
"key": "S",
"doc_count": 3
}
]
}
}
]
}
}
}
}
...Next Steps
We now have a scalable strategy for doing faceted searches using the Elasticsearch database / search engine. Or do we?
In the last article in this series, Elasticsearch by Example: Part 5, we refactor our solution to accommodate facet queries that allow for multiple selections on a facet.
