การสร้างดัชนีข้อความบนฟิลด์สตริง

A text index is a special MongoDB index type that tokenises and stems the words in string fields so you can run full-text keyword searches. Unlike a regular index that stores exact field values, a text index breaks each string into individual words, removes stop words (like 'the', 'is'), and stores the resulting tokens in a B-tree. This makes it possible to search for 'mongodb tutorial' and match documents containing 'mongodb tutorials'.

Pass the string 'text' as the index direction to tell MongoDB to create a text index on that field. You can only have one text index per collection, but it can span multiple fields. The index builds in the background and maintains the token store automatically as documents are inserted or updated.

A single text index can cover multiple string fields, allowing $text queries to search across all of them simultaneously. Each field can be assigned a different weight to influence relevance scoring—a match in a high-weight field like title counts more than a match in a lower-weight field like body.

If you want to search across every string field in a document without listing them all, you can use a wildcard text index with the special '$**' key. MongoDB will automatically tokenise all string-valued fields at any nesting level. This is convenient but creates a larger index than a targeted multi-field text index, so use it thoughtfully.

When MongoDB indexes a string like 'Learning MongoDB databases', it tokenises it into individual words (learning, mongodb, databases), removes stop words, and then stems each token to its root form (e.g., databases → databas). Stemming means a search for 'database' matches documents containing 'databases', 'database', or 'databasing' without needing wildcards.

Stop words are common words like 'the', 'is', 'at', 'which', and 'on' that carry little meaning and are excluded from the text index to keep it lean. If your search term consists entirely of stop words, the $text query returns no results. Stop word lists are language-specific and controlled by the default_language option on the index.

MongoDB's text index supports many languages including english, french, german, spanish, portuguese, italian, dutch, and more. Each language has its own stop word list and stemming rules. You can also set the language to 'none' to disable stop word removal and stemming, treating every token as a literal string.

MongoDB enforces a hard limit of one text index per collection. This means you must plan all the string fields you want searchable and include them in a single multi-field text index definition. Trying to create a second text index on the same collection will throw an error. If you need to add a field to an existing text index, you must drop and recreate the index.

Text indexes can be significantly larger than regular indexes because they store one entry per unique token per document rather than one entry per document. A document with a 500-word description might add hundreds of index entries. Monitor text index size with db.collection.stats().indexSizes and consider whether a dedicated search service (Atlas Search, Elasticsearch) would be more efficient for very large corpora.

A text index can be combined with a regular field in a compound index. For example, you can index { category: 1, description: 'text' } to allow filtering by category alongside the text search. In this case the category equality filter dramatically reduces the number of token entries the planner has to examine, making the text query much faster.

After creating a text index, use db.collection.getIndexes() to confirm it was created with the correct fields and weights, and run a simple $text query with .explain() to verify the query planner uses a TEXT stage. A TEXT stage in the plan means the text index is actively being used for keyword matching.

Test your understanding of MongoDB text indexes from this lesson.

In this lesson you learned: text indexes tokenise and stem string fields to enable full-text keyword searches, only one text index is allowed per collection but it can span multiple fields with custom weights, and language determines stop words and stemming rules. Next up we learn to run $text queries with phrases and negation.