Queries

Querying is how you find records that match certain conditions, for example:

  • Find all starred contacts
  • Find distinct first names in contacts
  • Delete all contacts that don't have the last name defined

Because queries are executed on the database and not in Dart, they're really fast. When you cleverly use indexes, you can improve the query performance even further. In the following, you'll learn how to write queries and how you can make them as fast as possible.

There are two different methods of filtering your records: Filters and where clauses. We'll start by taking a look at how filters work.

Filters

Filters are easy to use and understand. Depending on the type of your properties, there are different filter operations available most of which have self-explanatory names.

Filters work by evaluating an expression for every object in the collection being filtered. If the expression resolves to true, Isar includes the object in the results. Filters do not affect the ordering of the results.

We'll use the following model for the examples below:

@collection
class Shoe {
  Id? id;

  int? size;

  late String model;

  late bool isUnisex;
}

Query conditions

Depending on the type of field, there are different conditions available.

ConditionDescription
.equalTo(value)Matches values that are equal to the specified value.
.between(lower, upper)Matches values that are between lower and upper.
.greaterThan(bound)Matches values that are greater than bound.
.lessThan(bound)Matches values that are less than bound. null values will be included by default because null is considered smaller than any other value.
.isNull()Matches values that are null.
.isNotNull()Matches values that are not null.
.length()List, String and link length queries filter objects based on the number of elements in a list or link.

Let's assume the database contains four shoes with sizes 39, 40, 46 and one with an un-set (null) size. Unless you perform sorting, the values will be returned sorted by id.


isar.shoes.filter()
  .sizeLessThan(40)
  .findAll() // -> [39, null]

isar.shoes.filter()
  .sizeLessThan(40, include: true)
  .findAll() // -> [39, null, 40]

isar.shoes.filter()
  .sizeBetween(39, 46, includeLower: false)
  .findAll() // -> [40, 46]

Logical operators

You can composite predicates using the following logical operators:

OperatorDescription
.and()Evaluates to true if both left-hand and right-hand expressions evaluate to true.
.or()Evaluates to true if either expression evaluates to true.
.xor()Evaluates to true if exactly one expression evaluates to true.
.not()Negates the result of the following expression.
.group()Group conditions and allow to specify order of evaluation.

If you want to find all shoes in size 46, you can use the following query:

final result = await isar.shoes.filter()
  .sizeEqualTo(46)
  .findAll();

If you want to use more than one condition, you can combine multiple filters using logical and .and(), logical or .or() and logical xor .xor().

final result = await isar.shoes.filter()
  .sizeEqualTo(46)
  .and() // Optional. Filters are implicitly combined with logical and.
  .isUnisexEqualTo(true)
  .findAll();

This query is equivalent to: size == 46 && isUnisex == true.

You can also group conditions using .group():

final result = await isar.shoes.filter()
  .sizeBetween(43, 46)
  .and()
  .group((q) => q
    .modelNameContains('Nike')
    .or()
    .isUnisexEqualTo(false)
  )
  .findAll()

This query is equivalent to size >= 43 && size <= 46 && (modelName.contains('Nike') || isUnisex == false).

To negate a condition or group, use logical not .not():

final result = await isar.shoes.filter()
  .not().sizeEqualTo(46)
  .and()
  .not().isUnisexEqualTo(true)
  .findAll();

This query is equivalent to size != 46 && isUnisex != true.

String conditions

In addition to the query conditions above, String values offer a few more conditions you can use. Regex-like wildcards, for example, allow more flexibility in search.

ConditionDescription
.startsWith(value)Matches string values that begins with provided value.
.contains(value)Matches string values that contain the provided value.
.endsWith(value)Matches string values that end with the provided value.
.matches(wildcard)Matches string values that match the provided wildcard pattern.

Case sensitivity
All string operations have an optional caseSensitive parameter that defaults to true.

Wildcards:
A wildcard string expressionopen in new window is a string that uses normal characters with two special wildcard characters:

  • The * wildcard matches zero or more of any character
  • The ? wildcard matches any character.
    For example, the wildcard string "d?g" matches "dog", "dig", and "dug", but not "ding", "dg", or "a dog".

Query modifiers

Sometimes it is necessary to build a query based on some conditions or for different values. Isar has a very powerful tool for building conditional queries:

ModifierDescription
.optional(cond, qb)Extends the query only if the condition is true. This can be used almost anywhere in a query for example to conditionally sort or limit it.
.anyOf(list, qb)Extends the query for each value in values and combines the conditions using logical or.
.allOf(list, qb)Extends the query for each value in values and combines the conditions using logical and.

In this example, we build a method that can find shoes with an optional filter:

Future<List<Shoe>> findShoes(Id? sizeFilter) {
  return isar.shoes.filter()
    .optional(
      sizeFilter != null, // only apply filter if sizeFilter != null
      (q) => q.sizeEqualTo(sizeFilter!),
    ).findAll();
}

If you want to find all shoes that have one of multiple shoe sizes, you can either write a conventional query or use the anyOf() modifier:

final shoes1 = await isar.shoes.filter()
  .sizeEqualTo(38)
  .or()
  .sizeEqualTo(40)
  .or()
  .sizeEqualTo(42)
  .findAll();

final shoes2 = await isar.shoes.filter()
  .anyOf(
    [38, 40, 42],
    (q, int size) => q.sizeEqualTo(size)
  ).findAll();

// shoes1 == shoes2

Query modifiers are especially useful when you want to build dynamic queries.

Lists

Even lists can be queried:

class Tweet {
  Id? id;

  String? text;

  List<String> hashtags = [];
}

You can query based on the list length:

final tweetsWithoutHashtags = await isar.tweets.filter()
  .hashtagsIsEmpty()
  .findAll();

final tweetsWithManyHashtags = await isar.tweets.filter()
  .hashtagsLengthGreaterThan(5)
  .findAll();

These are equivalent to the Dart code tweets.where((t) => t.hashtags.isEmpty); and tweets.where((t) => t.hashtags.length > 5);. You can also query based on list elements:

final flutterTweets = await isar.tweets.filter()
  .hashtagsElementEqualTo('flutter')
  .findAll();

This is equivalent to the Dart code tweets.where((t) => t.hashtags.contains('flutter'));.

Embedded objects

Embedded objects are one of Isar's most useful features. They can be queried very efficiently using the same conditions available for top-level objects. Let's assume we have the following model:

@collection
class Car {
  Id? id;

  Brand? brand;
}

@embedded
class Brand {
  String? name;

  String? country;
}

We want to query all cars that have a brand with the name "BMW" and the country "Germany". We can do this using the following query:

final germanCars = await isar.cars.filter()
  .brand((q) => q
    .nameEqualTo('BMW')
    .and()
    .countryEqualTo('Germany')
  ).findAll();

Always try to group nested queries. The above query is more efficient than the following one. Even though the result is the same:

final germanCars = await isar.cars.filter()
  .brand((q) => q.nameEqualTo('BMW'))
  .and()
  .brand((q) => q.countryEqualTo('Germany'))
  .findAll();

If your model contains links or backlinks you can filter your query based on the linked objects or the number of linked objects.

Warning

Keep in mind that link queries can be expensive because Isar needs to look up linked objects. Consider using embedded objects instead.

@collection
class Teacher {
  Id? id;

  late String subject;
}

@collection
class Student {
  Id? id;

  late String name;

  final teachers = IsarLinks<Teacher>();
}

We want to find all students that have a math or English teacher:

final result = await isar.students.filter()
  .teachers((q) {
    return q.subjectEqualTo('Math')
      .or()
      .subjectEqualTo('English');
  }).findAll();

Link filters evaluate to true if at least one linked object matches the conditions.

Let's search for all students that have no teachers:

final result = await isar.students.filter().teachersLengthEqualTo(0).findAll();

or alternatively:

final result = await isar.students.filter().teachersIsEmpty().findAll();

Where clauses

Where clauses are a very powerful tool, but it can be a little challenging to get them right.

In contrast to filters where clauses use the indexes you defined in the schema to check the query conditions. Querying an index is a lot faster than filtering each record individually.

➡️ Learn more: Indexes

Tip

As a basic rule, you should always try to reduce the records as much as possible using where clauses and do the remaining filtering using filters.

You can only combine where clauses using logical or. In other words, you can sum multiple where clauses together, but you can't query the intersection of multiple where clauses.

Let's add indexes to the shoe collection:

@collection
class Shoe with IsarObject {
  Id? id;

  @Index()
  Id? size;

  late String model;

  @Index(composite: [CompositeIndex('size')])
  late bool isUnisex;
}

There are two indexes. The index on size allows us to use where clauses like .sizeEqualTo(). The composite index on isUnisex allows where clauses like .isUnisexSizeEqualTo(). But also .isUnisexEqualTo() because you can always use any prefix of an index.

We can now rewrite the query from before that finds unisex shoes in size 46 using the composite index. This query will be a lot faster than the previous one:

final result = isar.shoes.where()
  .isUnisexSizeEqualTo(true, 46)
  .findAll();

Where clauses have two more superpowers: They give you "free" sorting and a super fast distinct operation.

Combining where clauses and filters

Remember the shoes.filter() queries? It's actually just a shortcut for shoes.where().filter(). You can (and should) combine where clauses and filters in the same query to use the benefits of both:

final result = isar.shoes.where()
  .isUnisexEqualTo(true)
  .filter()
  .modelContains('Nike')
  .findAll();

The where clause is applied first to reduce the number of objects to be filtered. Then the filter is applied to the remaining objects.

Sorting

You can define how the results should be sorted when executing the query using the .sortBy(), .sortByDesc(), .thenBy() and .thenByDesc() methods.

To find all shoes sorted by model name in ascending order and size in descending order without using an index:

final sortedShoes = isar.shoes.filter()
  .sortByModel()
  .thenBySizeDesc()
  .findAll();

Sorting many results can be expensive, especially since sorting happens before offset and limit. The sorting methods above never make use of indexes. Luckily, we can again use where clause sorting and make our query lightning-fast even if we need to sort a million objects.

Where clause sorting

If you use a single where clause in your query, the results are already sorted by the index. That's a big deal!

Let's assume we have shoes in sizes [43, 39, 48, 40, 42, 45] and we want to find all shoes with a size greater than 42 and also have them sorted by size:

final bigShoes = isar.shoes.where()
  .sizeGreaterThan(42) // also sorts the results by size
  .findAll(); // -> [43, 45, 48]

As you can see, the result is sorted by the size index. If you want to reverse the where clause sort order, you can set sort to Sort.desc:

final bigShoesDesc = await isar.shoes.where(sort: Sort.desc)
  .sizeGreaterThan(42)
  .findAll(); // -> [48, 45, 43]

Sometimes you don't want to use a where clause but still benefit from the implicit sorting. You can use the any where clause:

final shoes = await isar.shoes.where()
  .anySize()
  .findAll(); // -> [39, 40, 42, 43, 45, 48]

If you use a composite index, the results are sorted by all fields in the index.

Tip

If you need the results to be sorted, consider using an index for that purpose. Especially if you work with offset() and limit().

Sometimes it's not possible or useful to use an index for sorting. For such cases, you should use indexes to reduce the number of resulting entries as much as possible.

Unique values

To return only entries with unique values, use the distinct predicate. For example, to find out how many different shoe models you have in your Isar database:

final shoes = await isar.shoes.filter()
  .distinctByModel()
  .findAll();

You can also chain multiple distinct conditions to find all shoes with distinct model-size combinations:

final shoes = await isar.shoes.filter()
  .distinctByModel()
  .distinctBySize()
  .findAll();

Only the first result of each distinct combination is returned. You can use where clauses and sort operations to control it.

Where clause distinct

If you have a non-unique index, you may want to get all of its distinct values. You could use the distinctBy operation from the previous section, but it's performed after sorting and filters, so there is some overhead.
If you only use a single where clause, you can instead rely on the index to perform the distinct operation.

final shoes = await isar.shoes.where(distinct: true)
  .anySize()
  .findAll();

Tip

In theory, you could even use multiple where clauses for sorting and distinct. The only restriction is that those where clauses are not overlapping and use the same index. For correct sorting, they also need to be applied in sort order. Be very careful if you rely on this!

Offset & Limit

It's often a good idea to limit the number of results from a query for lazy list views. You can do so by setting a limit():

final firstTenShoes = await isar.shoes.where()
  .limit(10)
  .findAll();

By setting an offset() you can also paginate the results of your query.

final firstTenShoes = await isar.shoes.where()
  .offset(20)
  .limit(10)
  .findAll();

Since instantiating Dart objects is often the most expensive part of executing a query, it is a good idea only to load the objects you need.

Execution order

Isar executes queries always in the same order:

  1. Traverse primary or secondary index to find objects (apply where clauses)
  2. Filter objects
  3. Sort results
  4. Apply distinct operation
  5. Offset & limit results
  6. Return results

Query operations

In the previous examples, we used .findAll() to retrieve all matching objects. There are more operations available, however:

OperationDescription
.findFirst()Retrieve only the first matching object or null if none matches.
.findAll()Retrieve all matching objects.
.count()Count how many objects match the query.
.deleteFirst()Delete the first matching object from the collection.
.deleteAll()Delete all matching objects from the collection.
.build()Compile the query to reuse it later. This saves the cost to build a query if you want to execute it multiple times.

Property queries

If you are only interested in the values of a single property, you can use a property query. Just build a regular query and select a property:

List<String> models = await isar.shoes.where()
  .modelProperty()
  .findAll();

List<int> sizes = await isar.shoes.where()
  .sizeProperty()
  .findAll();

Using only a single property saves time during deserialization. Property queries also work for embedded objects and lists.

Aggregation

Isar supports aggregating the values of a property query. The following aggregation operations are available:

OperationDescription
.min()Finds the minimum value or null if none matches.
.max()Finds the maximum value or null if none matches.
.sum()Sums all values.
.average()Calculates the average of all values or NaN if none matches.

Using aggregations is vastly faster than finding all matching objects and performing the aggregation manually.

Dynamic queries

Danger

This section is most likely not relevant to you. It is discouraged to use dynamic queries unless you absolutely need to (and you rarely do).

All the examples above used the QueryBuilder and the generated static extension methods. Maybe you want to create dynamic queries or a custom query language (like the Isar Inspector). In that case, you can use the buildQuery() method:

ParameterDescription
whereClausesThe where clauses of the query.
whereDistinctWhether where clauses should return distinct values (only useful for single where clauses).
whereSortThe traverse order of the where clauses (only useful for single where clauses).
filterThe filter to apply to the results.
sortByA list of properties to sort by.
distinctByA list of properties to distinct by.
offsetThe offset of the results.
limitThe maximum number of results to return.
propertyIf non-null, only the values of this property are returned.

Let's create a dynamic query:

final shoes = await isar.shoes.buildQuery(
  whereClauses: [
    WhereClause(
      indexName: 'size',
      lower: [42],
      includeLower: true,
      upper: [46],
      includeUpper: true,
    )
  ],
  filter: FilterGroup.and([
    FilterCondition(
      type: ConditionType.contains,
      property: 'model',
      value: 'nike',
      caseSensitive: false,
    ),
    FilterGroup.not(
      FilterCondition(
        type: ConditionType.contains,
        property: 'model',
        value: 'adidas',
        caseSensitive: false,
      ),
    ),
  ]),
  sortBy: [
    SortProperty(
      property: 'model',
      sort: Sort.desc,
    )
  ],
  offset: 10,
  limit: 10,
).findAll();

The following query is equivalent:

final shoes = await isar.shoes.where()
  .sizeBetween(42, 46)
  .filter()
  .modelContains('nike', caseSensitive: false)
  .not()
  .modelContains('adidas', caseSensitive: false)
  .sortByModelDesc()
  .offset(10).limit(10)
  .findAll();