schemaDefinition : ObjectObject where each key represents the name of an object store and each value represents the primary and secondary indexes

Return Value



Specifies tables to be added, altered or deleted in this version. Each key in the schemaDefinition argument represents a table name and each value represents the primary key followed by the list of indexed properties. NOTE: Unlike SQL, you don’t need to specify all properties but only the one you wish to index.


Never index properties containing images, movies or large (huge) strings. Store them in IndexedDB, yes! but just don’t index them!

  friends: '++id, name, age' // don't index "picture"

  name: 'Camilla',
  age: 25,
  picture: await getBlob('camilla.png') // but store it

Example how the “picture” property is stored without being indexed.

Writing this because there have been some issues on github where people index images or movies without really understanding the purpose of indexing fields. A rule of thumb: Are you going to put your property in a where(‘…’) clause? If yes, index it, if not, dont. Large indexes will affect database performance and in extreme cases make it unstable.

Changing the Schema

Please refer to Database Versioning that explains how to add, alter or remove a table using the Versioning framework in Dexie.

Schema Syntax

++Auto-incremented primary key
&Unique index
*Multi-entry index
[A+B]Compound index or primary key

Detail: Primary keys are implicitly marked as unique.

Indexable Types

Only properties of certain types can be indexed. This includes string, number, Date and Array but NOT boolean, null or undefined. Indexing a property path that turns out to hold a non-indexable type will have no effect. And using orderBy() with that property will NOT list that object.

Detailed Schema Syntax

The first entry in the schema string will always represent the primary key.

Syntax For Primary Key
++keyPathAutoincrement primary keyMeans that the primary key will be auto-incremented. Primary key must always be unique.
++Hidden autoincremented primary keyMeans that primary key is auto-incremented but not visible on the objects.
keyPathDon't autoincrement primary keyMeans that primary key can be any type and we have to provide it ourself
(blank)Hidden primary keyLeaving the first entry blank means that primary key is hidden and not auto-incremented
Syntax For Indexes
keyPathMeans that keyPath is indexed
&keyPathUniqueMeans that keyPath is indexed and keys must be unique
*keyPathMulti-valuedMeans that if key is an array, each array value will be regarded as a key to the object.
[keyPath1+keyPath2]CompoundDefining a compound index for keyPath1 and keyPath2

NOTE: keyPath represents a property name or a dotted path to a nested property.


var db = new Dexie('MyDatabase');
    friends: '++id,name,shoeSize', // Primary Key is auto-incremented (++id)
    pets: 'id, name, kind',        // Primary Key is not auto-incremented (id)
    cars: '++, name',              // Primary Key auto-incremented but not inbound
    enemies: ',name,*weaknesses',  // Primary key is neither inbound nor auto-incr
                                   // 'weaknesses' contains an array of keys (*)
    users: 'meta.ssn,',  // Dotted keypath refers to nested property 
    people: '[name+ssn], &ssn'     // Compound primary key. Unique index ssn

Detailed Sample

This sample shows how to define the schema of a given version:

var db = new Dexie("FriendsAndPetsDatabase");
    users: "++id, name, &username, *email,",
    relations: "++, userId1, userId2, [userId1+userId2], relation"
}); (e) {
    console.error("Open failed: " + e.stack);

db.transaction('rw', db.users, function () {

        name: "Zlatan",
        username: "ibra",
        email: [
        address: {
            city: "Malmö",
            country: "Sweden"

        .or("").anyOf (["Malmö", "Stockholm", "Barcelona"])
        .each(function (user) {
            console.log("Found user: " +;

}).catch (function (e) {

Detailed Sample Explained

  • Table “users” has:
    • an auto-incremented primary key named id.
    • an index on the name property which could be of any type.
    • a unique index on the username property.
    • a multi index on the email property, meaning that it allow multiple emails and the possibility to index each of them and find the single user object. NOTE! This feature lacks support in IE.
    • an index on the nested property ‘’.
  • Table “relations” doesn’t have a “visible” primary key (however, it must have one autoincremented internally).
  • Table “relations” has index on the userId1, userId2 and relation properties.
  • Table “relations” has a compound index of the properties userId1 and userId2 combined NOTE! This feature lacks support in IE.

Queries you could do with these indexes:

  • db.users.get(2) will give you the user with id 2
  • db.users.where('name').startsWithIgnoreCase('da') - will give you all users starting with “da”
  • db.users.where('username').equals('usrname').first() - will give you the user with username ‘usrname’
  • db.users.where('email').startsWith('david@').distinct() - will give you the users that have any of their emails starting with ‘david@’
  • db.users.where('').equalsIgnoreCase('malmö') - will give you all users residing in Malmö.
  • db.relations.where('userId1').equals(2) - will give you all relations that user with id 2 has to other users
  • db.relations.where('relation').anyOf('wife', 'husband', 'son', 'daughter') - will give you all family relations.
  • db.relations.where('userId1').equals(2).or('userId2').equals(2) - will give you all relations that user with id 2 has to other users or other users have to user 2
  • db.relations.where('[userId1+userId2]').equals([2,3]) - will give you all the relations that user 2 has to user 3
  • db.relations.where('[userId1+userId2]').equals([2,3]).or('[userId1+userId2]').equals([3,2]) - will give you all the relations that user 2 has to user 3 or user 3 has to user 2.

Table of Contents