Type converters

Store more complex data in columns with type converters

Drift supports a variety of types out of the box, but sometimes you need to store more complex data. You can achieve this by using TypeConverters. In this example, we'll use the the json_serializable package to store a custom object in a text column. Drift supports any Dart type for which you provide a TypeConverter, we're using that package here to make the example simpler.

Using converters in Dart

import 'dart:convert';

import 'package:drift/drift.dart';
import 'package:json_annotation/json_annotation.dart' as j;

part 'converters.g.dart';

@j.JsonSerializable()
class Preferences {
  bool receiveEmails;
  String selectedTheme;

  Preferences(this.receiveEmails, this.selectedTheme);

  factory Preferences.fromJson(Map<String, dynamic> json) =>
      _$PreferencesFromJson(json);

  Map<String, dynamic> toJson() => _$PreferencesToJson(this);
}

Next, we have to tell drift how to store a Preferences object in the database. We write a TypeConverter for that:

// stores preferences as strings
class PreferenceConverter extends TypeConverter<Preferences, String> {
  const PreferenceConverter();

  @override
  Preferences fromSql(String fromDb) {
    return Preferences.fromJson(json.decode(fromDb) as Map<String, dynamic>);
  }

  @override
  String toSql(Preferences value) {
    return json.encode(value.toJson());
  }
}

Finally, we can use that converter in a table declaration:

class Users extends Table {
  IntColumn get id => integer().autoIncrement()();
  TextColumn get name => text()();

  TextColumn get preferences =>
      text().map(const PreferenceConverter()).nullable()();
}

The generated User class will then have a preferences column of type Preferences. Drift will automatically take care of storing and loading the object in select, update and insert statements. This feature also works with compiled custom queries.

Since applying type converters for JSON conversion is so common, drift provides a helper for that. For instance, we could declare the type converter as a field in the Preferences class:

static TypeConverter<Preferences, String> converter = TypeConverter.json(
  fromJson: (json) => Preferences.fromJson(json as Map<String, Object?>),
  toJson: (pref) => pref.toJson(),
);

Implicit enum converters

A common scenario for type converters is to map between enums and integers by representing enums as their index. Since this is so common, drift has the integrated intEnum column type to make this easier.

enum Status {
   none,
   running,
   stopped,
   paused
}

class Tasks extends Table {
  IntColumn get id => integer().autoIncrement()();
  IntColumn get status => intEnum<Status>()();
}

If you prefer to store the enum as a text, you can use textEnum instead.

enum Status {
   none,
   running,
   stopped,
   paused
}

class Tasks extends Table {
  IntColumn get id => integer().autoIncrement()();
  TextColumn get status => textEnum<Status>()();
}

Also note that you can't apply another type converter on a column declared with an enum converter.

Using converters in drift

Type converters can also be used inside drift files. Assuming that the Preferences and PreferenceConverter are contained in preferences.dart, that file can imported into drift for the type converter to be available.

import 'preferences.dart';

CREATE TABLE users (
  id INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT,
  name TEXT,
  preferences TEXT MAPPED BY `const PreferenceConverter()`
);

When using type converters in drift files, we recommend the apply_converters_on_variables build option. This will also apply the converter from Dart to SQL, for instance if used on variables: SELECT * FROM users WHERE preferences = ?. With that option, the variable will be inferred to Preferences instead of String.

Drift files also have special support for implicit enum converters:

import 'status.dart';

CREATE TABLE tasks (
  id INTEGER NOT NULL PRIMARY KEY,
  status ENUM(Status)
);

Of course, the warning about automatic enum converters also applies to drift files.

Type converters and json serialization

By default, type converters only apply to the conversion from Dart to the database. They don't impact how values are serialized to and from JSON. If you want to apply the same conversion to JSON as well, make your type converter mix-in the JsonTypeConverter class. You can also override the toJson and fromJson methods to customize serialization as long as the types stay the compatible. The type converter returned by TypeConverter.json already implements JsonTypeConverter, so it will apply to generated row classes as well.

If the JSON type you want to serialize to is different to the SQL type you're mapping to, you can mix-in JsonTypeConverter2 instead. For instance, say you have a type converter mapping to a complex Dart type MyObject. In SQL, you might want to store this as an String. But when serializing to JSON, you may want to use a Map<String, Object?>. Here, simply add the JsonTypeConverter2<MyObject, String, Map<String, Object?>> mixin to your type converter.

As an alternative to using JSON type converters, you can use a custom ValueSerializer and pass it to the serialization methods.

Type converters and SQL

In database rows, columns to which a type converter has been applied are storing the result of toSql(). Drift will apply the type converter automatically when reading or writing rows, but they are not applied automatically when creating your own expressions. For example, filtering for values with column.equals will compare not apply the type converter, you'd be comparing the underlying database values.

On columns with type converters, equalsValue can be used instead - unlike equals, equasValue will apply the converter before emtting a comparison in SQL. If you need to apply the converter for other comparisons as well, you can do that manually with column.converter.toSql.

For variables used in queries that are part of a drift file, type converters will be applied by default if the apply_converters_on_variables builder option is enabled (which it is by default).