This can serialize and de-serialize objects to multiple different
formats. It is most useful if you have Dart on both ends, rather
than needing to communicate with an external system. It can handle
cycles, inheritance, getters and setters, private or final fields set
via constructors, objects
serialized in different ways at different times, and other complex
options. It can handle serializing acyclic objects with only public
fields to a simple JSON format, but might be more heavyweight than is
necessary if that’s the only requirement.
This has no privileged access to object representations, so objects
are accessed and created according to their public APIs. As a result,
serializations from older versions where the internal representation
has changed can still be read as long as the public API is still available.
The way an object’s state is read and written is defined by
SerializationRules. These can be implemented in various ways. The
easiest to use is using mirrors to find the members. Rules can also be
hand-written or, for relatively simple classes, generated using a
transformer.
Then, set up the generated rules in a Serialization instance, and then call
write(). The serialization rules will be named as
the name of the model file with _serialization_rules appended to it,
so in the case of stuff.dart, stuff_serialization_rules.dart, in
the same directory as the original.
Normally you won’t ever see these files, because the
transformer creates it on the fly and it is sent directly to pub serve
or to dart2js without ever being written to disk.
To see the generated code, run pub build in debug mode, e.g.
if there is a program in the package’s bin directory to run something
using these files, then
pub build --mode=debug bin
would generate the code for that, and also log which
files were generated, in the form
Generated serialization rules in my_package|lib/stuff_serialization_rules.dart
It’s also possible to run the transformer’s code outside of the
transformer, which is helpful for debugging or to use the code in a
different way. See the `test/transformer/generate_standalone.dart’ for
an example of that.
The bin directory code would look something like.
import 'package:my_package/stuff_serialization_rules.dart' as foo;
...
var serialization = new Serialization();
foo.rules.values.forEach(serialization.addRule);
...
sendToClient(serialization.write(somePerson));
and on the client, do something like
p = readFromServer(personId).then((data) => serialization.read(data));
Alternatively, if using the mirrored rules, just tell the
serialization which classes might be serialized.
var serialization = new Serialization()
..addRuleFor(Person);
..addRuleFor(Address);
serialization.write(address);
For more concrete examples, see the test directory, and particularly
for examples of the transformer it may be useful to look at the
pubspec.yaml for this package, and thetest/transformer directory.
General serialization for Dart objects.
Save and restore objects flexibly.
This can serialize and de-serialize objects to multiple different formats. It is most useful if you have Dart on both ends, rather than needing to communicate with an external system. It can handle cycles, inheritance, getters and setters, private or final fields set via constructors, objects serialized in different ways at different times, and other complex options. It can handle serializing acyclic objects with only public fields to a simple JSON format, but might be more heavyweight than is necessary if that’s the only requirement.
This has no privileged access to object representations, so objects are accessed and created according to their public APIs. As a result, serializations from older versions where the internal representation has changed can still be read as long as the public API is still available.
The way an object’s state is read and written is defined by SerializationRules. These can be implemented in various ways. The easiest to use is using mirrors to find the members. Rules can also be hand-written or, for relatively simple classes, generated using a transformer.
Usage
Import either
or
depending on whether or not you want the mirrored rules. These are more convenient, but cause increased code size in dartj2s.
To use the transformer, include something in the pubspec like
Then, set up the generated rules in a Serialization instance, and then call write(). The serialization rules will be named as the name of the model file with
_serialization_rulesappended to it, so in the case ofstuff.dart,stuff_serialization_rules.dart, in the same directory as the original.Normally you won’t ever see these files, because the transformer creates it on the fly and it is sent directly to pub serve or to dart2js without ever being written to disk. To see the generated code, run pub build in debug mode, e.g. if there is a program in the package’s
bindirectory to run something using these files, thenwould generate the code for that, and also log which files were generated, in the form
It’s also possible to run the transformer’s code outside of the transformer, which is helpful for debugging or to use the code in a different way. See the `test/transformer/generate_standalone.dart’ for an example of that.
The bin directory code would look something like.
and on the client, do something like
Alternatively, if using the mirrored rules, just tell the serialization which classes might be serialized.
For more concrete examples, see the
testdirectory, and particularly for examples of the transformer it may be useful to look at thepubspec.yamlfor this package, and thetest/transformerdirectory.Requests and bugs
Please file feature requests and bugs via the GitHub Issue Tracker. This is licensed under the same license as Dart
Disclaimer
This is not an official Google project.