How do I do a partial match in Elasticsearch?

In Elasticsearch, performing partial matching typically involves several different query types, such as the match query, wildcard query, prefix query, and more complex n-gram tokenizer or edge n-gram tokenizer. I will provide a detailed explanation of these methods along with specific examples.

1. Match Query

The match query is the most common query type in Elasticsearch for handling full-text search and supports partial matching. When using the match query to search a text field, Elasticsearch tokenizes the input search text and then searches for each token.

Example: Suppose we have an index containing product information, with one field being description. If we want to search for products where the description contains "apple", we can use the following query:

json
{
  "query": {
    "match": {
      "description": "apple"
    }
  }
}

This will return all documents where the description field contains "apple", regardless of whether "apple" is a standalone word or part of a phrase.

2. Wildcard Query

The wildcard query allows searching using wildcards, such as * (representing any sequence of characters) and ? (representing any single character). This is a straightforward method for pattern matching during search.

Example: If we want to find all description fields starting with "app":

json
{
  "query": {
    "wildcard": {
      "description": "app*"
    }
  }
}

3. Prefix Query

The prefix query is a specialized query type used to find text with a specific prefix. This query is commonly employed in autocomplete scenarios.

Example: To find all documents where description starts with "app", we can use the following query:

json
{
  "query": {
    "prefix": {
      "description": "app"
    }
  }
}

4. Using N-Gram and Edge N-Gram

By utilizing the n-gram or edge n-gram tokenizer to create sub-terms during indexing, more flexible partial matching searches can be achieved. These tokenizers break down text into a series of n-grams.

Example: Suppose during index setup, we use the edge_ngram tokenizer for the description field with a minimum length of 2 and maximum length of 10. This way, the word "apple" is indexed as ["ap", "app", "appl", "apple"].

json
{
  "query": {
    "match": {
      "description": "app"
    }
  }
}

The above query will match all documents containing the term "app" and its extensions, such as "apple" or "application".

Conclusion

Different partial matching query methods have distinct use cases and performance considerations. For instance, wildcard and prefix queries may perform poorly on large datasets, while n-gram methods, though resulting in larger indexes, offer faster query responses and greater flexibility. The choice depends on specific requirements and dataset characteristics. In practical applications, query optimization and indexing strategies should also be considered to achieve optimal search performance and results.