Text vs keyword field types in Elasticsearch - Visual Side-by-Side Comparison

Concept Flow - Text vs keyword field types

Input Document

↓

Field Type?

↓

text

↓

Analyzed

↓

Tokenized

Documents have fields that are either 'text' or 'keyword'. Text fields are analyzed and tokenized for full-text search. Keyword fields are stored as-is for exact matching and sorting.

Execution Sample

Elasticsearch

PUT /my_index
{
  "mappings": {
    "properties": {
      "title": {"type": "text"},
      "status": {"type": "keyword"}
    }
  }
}

Defines an index with 'title' as a text field for full-text search and 'status' as a keyword field for exact matching.

Execution Table

Step	Field	Type	Analyzed?	Stored As	Use Case
1	title	text	Yes	Tokens (words)	Full-text search, relevance scoring
2	status	keyword	No	Exact value	Filtering, sorting, aggregations
3	Search 'title:quick fox'	title	Analyzed query	Matches tokens 'quick' and 'fox'	Find documents with those words
4	Filter 'status:active'	status	Exact match	Matches exact 'active'	Filter documents with status active
5	Sort by 'status'	status	Exact value	Sorts documents alphabetically by status	Ordering results
6	Exit	-	-	-	No more steps, mapping and usage clear

💡 All fields processed with their types and usage explained.

Variable Tracker

Variable	Start	After Step 1	After Step 2	After Step 3	Final
title	undefined	text type set	analyzed tokens created	tokens matched in search	used for full-text search
status	undefined	keyword type set	stored as exact value	exact match filter applied	used for filtering and sorting

Key Moments - 3 Insights

Why is the 'text' field analyzed but the 'keyword' field is not?

Because 'text' fields are meant for full-text search, they are broken into tokens (words) to match parts of the text. 'Keyword' fields store exact values for filtering and sorting, so they are not broken down. See execution_table rows 1 and 2.

Can you use a 'keyword' field for full-text search?

No, 'keyword' fields are stored as exact values and do not support tokenization or relevance scoring. Full-text search requires 'text' fields. Refer to execution_table row 3 for how 'title' is searched.

Why do we sort on 'keyword' fields and not on 'text' fields?

'Keyword' fields store exact values, so sorting is straightforward. 'Text' fields are tokenized and analyzed, making sorting ambiguous. See execution_table row 5.

Visual Quiz - 3 Questions

Look at the execution table, what is the 'status' field stored as after step 2?

A Exact value

B Analyzed tokens

C Lowercase words

D Stemmed words

💡 Hint

Check execution_table row 2 under 'Stored As' column.

At which step does the 'title' field get tokenized for search?

A Step 1

B Step 2

C Step 3

D Step 5

💡 Hint

Look at execution_table row 3 describing search on 'title'.

If you want to filter documents by exact status, which field type should you use?

A text

B keyword

C integer

D date

💡 Hint

Refer to execution_table rows 2 and 4 about filtering on 'status'.

Concept Snapshot

Text vs Keyword fields in Elasticsearch:
- text: analyzed, tokenized for full-text search
- keyword: not analyzed, stored as exact value
- Use text for searching words inside text
- Use keyword for exact match, filtering, sorting
- Mapping defines field type for proper usage

Full Transcript

In Elasticsearch, fields can be either 'text' or 'keyword' types. Text fields are analyzed and broken into tokens to support full-text search, allowing you to find documents by words inside the text. Keyword fields are stored as exact values without analysis, making them perfect for filtering, sorting, and aggregations. For example, a 'title' field as text lets you search for words like 'quick' or 'fox', while a 'status' field as keyword lets you filter documents exactly matching 'active'. Understanding these types helps you design your index for efficient search and filtering.

Practice

1. Which Elasticsearch field type should you use if you want to perform full-text search inside sentences?

easy

A. text

B. keyword

C. integer

D. date

Solution

1. Step 1: Understand the purpose of text fields

   Text fields are designed for full-text search, allowing analysis and tokenization of sentences.

2. Step 2: Compare with keyword fields

   Keyword fields are for exact matches and do not analyze the text, so they are not suitable for searching inside sentences.

3. Final Answer:

   text -> Option A

4. Quick Check:

   Full-text search = text [OK]

Hint: Full-text search needs analyzed text fields [OK]

Common Mistakes:

• Choosing keyword for full-text search
• Confusing integer or date types with text
• Thinking keyword fields tokenize text

2. Which of the following is the correct way to define a keyword field in an Elasticsearch mapping?

easy

A. "name": { "type": "text" }

B. "name": { "type": "keyword" }

C. "name": { "type": "string" }

D. "name": { "type": "fulltext" }

Solution

1. Step 1: Recall Elasticsearch field types

   Elasticsearch uses "keyword" as the type for exact match fields.

2. Step 2: Identify the correct syntax

   "keyword" is the correct type name; "text" is for analyzed fields, "string" and "fulltext" are invalid types.

3. Final Answer:

   "name": { "type": "keyword" } -> Option B

4. Quick Check:

   Keyword field syntax = "name": { "type": "keyword" } [OK]

Hint: Keyword type is exactly "keyword" in mapping [OK]

Common Mistakes:

• Using "string" instead of "keyword"
• Confusing "text" with "keyword"
• Using invalid type names like "fulltext"

3. Given this Elasticsearch mapping snippet:

{
  "properties": {
    "title": { "type": "text" },
    "category": { "type": "keyword" }
  }
}

Which query will correctly filter documents where category is exactly "books"?

medium

A. { "term": { "category": "books" } }

B. { "match": { "category": "books" } }

C. { "match": { "title": "books" } }

D. { "term": { "title": "books" } }

Solution

1. Step 1: Identify field types and query types

   "category" is a keyword field, so exact match queries use "term" query.

2. Step 2: Match query usage

   "match" queries are for full-text search on text fields, not exact matches on keyword fields.

3. Final Answer:

   { "term": { "category": "books" } } -> Option A

4. Quick Check:

   Exact filter on keyword = term query [OK]

Hint: Use term query for keyword fields exact match [OK]

Common Mistakes:

• Using match query on keyword fields
• Filtering text fields with term query expecting full-text search
• Confusing field names in queries

4. You have this mapping:

{
  "properties": {
    "username": { "type": "keyword" },
    "bio": { "type": "text" }
  }
}

A developer tries to filter users with username containing "john" using this query:

{ "match": { "username": "john" } }

Why does this query not work as expected?

medium

A. Because match query only works on numeric fields

B. Because username field should be of type text for exact matches

C. Because match query does not work on keyword fields for partial matches

D. Because bio field should be used instead of username

Solution

1. Step 1: Understand keyword field behavior

   Keyword fields are not analyzed and do not support partial matching with match query.

2. Step 2: Understand match query limitations

   Match query analyzes input and is meant for text fields, not keyword fields which require term or wildcard queries for partial matches.

3. Final Answer:

   Because match query does not work on keyword fields for partial matches -> Option C

4. Quick Check:

   Match query + keyword field = no partial match [OK]

Hint: Match query needs text fields, not keyword [OK]

Common Mistakes:

• Thinking keyword fields support partial match with match query
• Confusing field types for filtering
• Using match query on numeric or date fields

5. You want to store user tags in Elasticsearch so you can both filter by exact tag and search inside tag text. Which mapping setup is best?

hard

A. "tags": { "type": "text" }

B. "tags": { "type": "keyword" }

C. "tags": { "type": "keyword", "fields": { "analyzed": { "type": "text" } } }

D. "tags": { "type": "text", "fields": { "raw": { "type": "keyword" } } }

Solution

1. Step 1: Understand the need for both full-text search and exact filtering

   Full-text search requires text type; exact filtering requires keyword type.

2. Step 2: Use multi-fields to support both

   Mapping with "text" type and a subfield "raw" as "keyword" allows searching inside tags and filtering/sorting by exact tag.

3. Final Answer:

   "tags": { "type": "text", "fields": { "raw": { "type": "keyword" } } } -> Option D

4. Quick Check:

   Multi-fields = text + keyword for search and filter [OK]

Hint: Use multi-fields for both text search and exact filter [OK]

Common Mistakes:

• Using only text type and expecting exact filters
• Using only keyword type and expecting full-text search
• Not defining subfields for multi-purpose queries