Queries and Indexes

Consider whether your application users want to return the content of documents or metadata about your documents:

• If users want document content, use the $search aggregation stage to return documents that match their search criteria.

• If users want metadata about their search results, use the $searchMeta stage to return customizable counts of matching documents and facets.

Identify the specific fields within your collections that users are likely to search so that you know which fields to index. Each Atlas Search query searches a single Atlas Search index, which contains terms that are extracted from one or more specified fields within a collection. When planning your Atlas Search queries, decide whether to index only key fields or every field in your specified collection by enabling static or dynamic mapping. You can query across multiple fields by specifying the query path as an array of fields, or by using the queryString operator.

Your users' common search terms may be exact, similar , or partial matches for the data in your Atlas cluster. For example, users of a movie review application may want to filter for movies from an exact year, or see movie recommendations that are similar to their favorite film.

Determine the type of matches your users are searching for to inform which operators to use in your Atlas Search queries:

• For exact matches, use operators like 等于 or in to match documents that contain terms that are identical to the specified query value. You can also use the text operator to match documents that contain any or all of the strings in the query value.

• For similar matches, use operators like 接近, 更多类似内容, or 短语 to match documents that contain numeric values, documents, or string orderings that are similar to the specified search terms. You can also use the 范围 operator to match documents that contain a value within a specified range of values.

• For partial matches, such as search-as-you-type queries, use operators like 自动补全, 正则表达式(Regex), or 通配符 to implement search-as-you-type functionality or match terms using regular expressions.

• Use the 多个子句 operator to blend multiple matching behaviors in a single query.

For applications that require text normalization, multi-language support, stemming, or more, leverage Atlas Search text analysis tools:

• Choose a built-in analyzer in your index definition to match the language and nature of your text data. Analyzers break text into terms or tokens and can adjust text to remove punctuation and capitalization, convert words to their root form, and more.

• Configure custom analyzers if your application has specific requirements like handling domain-specific jargon or parsing formatted text like email addresses or dash-separated IDs. Custom analyzers enable you to filter text by character, define the number of characters to include in each token chunk, or enable stemming or redaction.

• Define synonyms to improve search accuracy for terms with the same or similar meanings.

You can adjust the presentation of search results based on your users' filtering, sorting, or relevancy demands:

• Use the score query option to modify the relevance score of documents and affect the order in which users view results. Atlas Search queries associate a relevance-based score with every document in the result set, and returns documents in order from the highest to the lowest score.

• Set the sort query option for indexed fields that users are likely to sort in ascending or descending order, such as dates or numeric fields.

• Use the searchBefore or searchAfter query options to display results as a set of pages that users can navigate sequentially or skip through.

• Use the facet collector to allow users to filter results by categories or other dimensions. This can significantly improve the relevance of search results, offering users a more guided search experience.

Atlas Search query performance is affected by your index configuration and the complexity of your queries. Focus on indexing fields that are critical to your application's search functionality and aim for a logical balance between query complexity and speed.

To further optimize performance, consider the following query options:

• Use the concurrent query option to set the number of concurrent search requests that are executed when evaluating a query. This option is useful for complex queries or large datasets.

• Use the returnStoredSource query option in combination with the 存储的源 index option to determine whether to return original source documents, stored as part of the index, alongside the search results. This option is useful for applications where you display summaries or highlights based on search criteria.

• Use the numPartitions index option to partition your index, distributing index objects between sub-indexes in an optimal way.

For more recommendations on how to optimize your query performance, see Atlas Search 查询性能.