🗓️ 04102025 0034
📎
Elasticsearch Text Search Cheatsheet
A quick reference for Elasticsearch search capabilities and when to use each type.
Core Concepts
Full-Text vs Exact Match
| Type | Use Case | Example |
|---|---|---|
| Full-Text | Natural language search, fuzzy matching | "searching for articles" |
| Exact Match | IDs, status codes, categories | status: "active" |
Key difference: Full-text is analyzed (lowercased, stemmed, tokenized), exact match is not.
Search Types Overview
| Search Type | What It Does | Best For |
|---|---|---|
| Match | Full-text search with analysis | General search, user queries |
| Multi-Match | Search across multiple fields | Searching title + description |
| Match Phrase | Exact phrase in order | "New York", "machine learning" |
| Term | Exact value match | IDs, enums, keywords |
| Fuzzy | Typo-tolerant search | Handling misspellings |
| Prefix | Starts with... | Autocomplete, type-ahead |
| Wildcard | Pattern matching | user*, *@gmail.com |
| Regexp | Regex patterns | Complex pattern matching |
| Range | Numeric/date ranges | Price filters, date ranges |
| Bool | Combine queries | Complex filtering logic |
match vs term
- match: Analyzed (lowercased, stemmed) - "Searching" → "search"
- term: Exact value - "Searching" ≠ "searching"
match vs match_phrase
- match: Words can be in any order
- match_phrase: Words must be in exact order
must vs filter (in bool)
- must: Affects relevance score (use for search)
- filter: Yes/no match, faster (use for filtering)
should vs must
- should: Optional, boosts score if matched (OR logic)
- must: Required (AND logic)
Common Patterns
Search with Filters
Combine search queries (must) with filters (filter) - filters don't affect relevance score and are cached for performance.
Fuzzy Search with Minimum Match
Use fuzziness: "AUTO" to handle typos, and minimum_should_match (e.g., "75%") to require a percentage of search terms to match.
Wildcard Search
Pattern matching with * and ? - useful but can be slow, especially with leading wildcards.
Performance Tips
- Use filters over queries when you don't need scoring
- Term queries are faster than match queries
- Wildcard with
*at start is slow - avoid if possible - Use prefix/edge_ngram for autocomplete, not wildcard
- Cache filters - ES automatically caches filter results
- Limit fields in multi_match - don't search everything
References
- elasticsearch_scoring_relevance - How results are ranked and scored
- elasticsearch_analyzers - How text is processed and tokenized
- Elasticsearch Query DSL
- Match Query
- Bool Query