76 lines
2.8 KiB
ReStructuredText
76 lines
2.8 KiB
ReStructuredText
|
.. _migration-streams-2.0-2.4-java:
|
||
|
|
|
||
|
|
##############################
|
||
|
|
Migration Guide 2.0.x to 2.4.x
|
||
|
|
##############################
|
||
|
|
|
||
|
|
General notes
|
||
|
|
=============
|
||
|
|
|
||
|
|
akka.Done and akka.NotUsed replacing Unit and BoxedUnit
|
||
|
|
-------------------------------------------------------
|
||
|
|
|
||
|
|
To provide more clear signatures and have a unified API for both
|
||
|
|
Java and Scala two new types have been introduced:
|
||
|
|
|
||
|
|
``akka.NotUsed`` is meant to be used instead of ``Unit`` in Scala
|
||
|
|
and ``BoxedUnit`` in Java to signify that the type parameter is required
|
||
|
|
but not actually used. This is commonly the case with ``Source``, ``Flow`` and ``Sink``
|
||
|
|
that do not materialize into any value.
|
||
|
|
|
||
|
|
``akka.Done`` is added for the use case where it is boxed inside another object to signify
|
||
|
|
completion but there is no actual value attached to the completion. It is used to replace
|
||
|
|
occurrences of ``Future<BoxedUnit>`` with ``Future<Done>`` in Java and ``Future[Unit]`` with
|
||
|
|
``Future[Done]`` in Scala.
|
||
|
|
|
||
|
|
All previous usage of ``Unit`` and ``BoxedUnit`` for these two cases in the akka streams APIs
|
||
|
|
has been updated.
|
||
|
|
|
||
|
|
This means that Java code like this::
|
||
|
|
|
||
|
|
Source<String, BoxedUnit> source = Source.from(Arrays.asList("1", "2", "3"));
|
||
|
|
Sink<String, Future<BoxedUnit>> sink = Sink.ignore();
|
||
|
|
|
||
|
|
needs to be changed into::
|
||
|
|
|
||
|
|
Source<String, NotUsed> source = Source.from(Arrays.asList("1", "2", "3"));
|
||
|
|
Sink<String, Future<Done>> sink = Sink.ignore();
|
||
|
|
|
||
|
|
These changes apply to all the places where streams are used, which means that signatures
|
||
|
|
in the persistent query APIs also are affected.
|
||
|
|
|
||
|
|
Changed Operators
|
||
|
|
=================
|
||
|
|
|
||
|
|
``expand()`` is now based on an Iterator
|
||
|
|
----------------------------------------
|
||
|
|
|
||
|
|
Previously the ``expand`` combinator required two functions as input: the first
|
||
|
|
one lifted incoming values into an extrapolation state and the second one
|
||
|
|
extracted values from that, possibly evolving that state. This has been
|
||
|
|
simplified into a single function that turns the incoming element into an
|
||
|
|
Iterator.
|
||
|
|
|
||
|
|
The most prominent use-case previously was to just repeat the previously received value::
|
||
|
|
|
||
|
|
// This no longer works!
|
||
|
|
Flow.of(Integer.class).expand(i -> i)(i -> new Pair<>(i, i));
|
||
|
|
|
||
|
|
In Akka 2.4.x this is simplified to:
|
||
|
|
|
||
|
|
.. includecode:: ../code/docs/stream/MigrationsJava.java#expand-continually
|
||
|
|
|
||
|
|
If state needs to be be kept during the expansion process then this state will
|
||
|
|
need to be managed by the Iterator. The example of counting the number of
|
||
|
|
expansions might previously have looked like::
|
||
|
|
|
||
|
|
// This no longer works!
|
||
|
|
Flow.of(Integer.class).expand(i -> new Pair<>(i, 0))(
|
||
|
|
pair -> new Pair<>(new Pair<>(pair.first(), pair.second()),
|
||
|
|
new Pair<>(pair.first(), pair.second() + 1)));
|
||
|
|
|
||
|
|
In Akka 2.4.x this is formulated like so:
|
||
|
|
|
||
|
|
.. includecode:: ../code/docs/stream/MigrationsJava.java#expand-state
|
||
|
|
|