Geospatial Index Restrictions

Collation Option

2d indexes do not support the collation option, only binary comparison. Binary comparison compares the numeric Unicode value of each character in each string, and does not account for letter case or accent marks.

To create a 2d index on a collection that has a non-simple collation, you must explicitly specify { collation: { locale: "simple" } } when you create the index.

For example, consider a collection named collationTest with a collation of { locale: "en" }:

db.createCollection(
   "collationTest",
   {
      collation: { locale: "en" }
   }
)

To create a 2d index on the collationTest collection, you must specify { collation: { locale: "simple" } }. This command creates a 2d index on the loc field:

db.collationTest.createIndex(
   {
      loc: "2d"
   },
   {
      collation: { locale: "simple" }
   }
)

Covered Queries

Geospatial indexes cannot cover a query.

Shard Key

You cannot use a geospatial index as a shard key. However, you can create a geospatial index on a sharded collection by using a different field as the shard key.

Multiple Geospatial Indexes with $geoNear

If your collection has multiple geospatial indexes, when you run the $geoNear pipeline stage, you must specify the $geoNear key option. The key option specifies which index to use to support the query.

Supported Data Types

A field indexed with a 2dsphere index must contain geometry data. Geometry data can either be:

• GeoJSON data

• Legacy coordinate pairs

You cannot:

• Insert a document with non-geometry data into a field that is indexed with a 2dsphere index.

• Build a 2dsphere index on a field that contains non-geometry data.

Number of Index Keys

When you create a 2dsphere index, mongod maps GeoJSON shapes to an internal representation. The resulting internal representation may be a large array of values.

The indexMaxNumGeneratedKeysPerDocument setting limits the maximum number of keys generated for a single document to prevent out of memory errors. If an operation requires more keys than the indexMaxNumGeneratedKeysPerDocument parameter specifies, the operation fails.

By default, the server allows up to 100,000 index keys per document. To allow more index keys, raise the indexMaxNumGeneratedKeysPerDocument value.

Exact Matches on a Flat Surface

A 2d index cannot improve performance for exact matches on a coordinate pair.

For example, consider a contacts collection with these documents:

db.contacts.insertMany( [
   {
      name: "Evander Otylia",
      phone: "202-555-0193",
      address: [ 55.5, 42.3 ]
   },
   {
      name: "Georgine Lestaw",
      phone: "714-555-0107",
      address: [ -74, 44.74 ]
   }
] )

A 2d index on the address field does not improve performance for the following query:

db.contacts.find( { address: [ 55.5, 42.3 ] } )

To improve performance for this query, create either an ascending or descending index on the address field:

db.contacts.createIndex( { address: 1 } )

Learn More

• 2d Index Internals

• Index Properties