All Downloads are FREE. Search and download functionalities are using the official Maven repository.

net.openhft.chronicle.wire.domestic.stream.README.adoc Maven / Gradle / Ivy

The newest version!
= Chronicle Stream Bridge
Per Minborg
:css-signature: demo
:toc: macro
:toclevels: 3
:icons: font

toc::[]

== About Chronicle Stream Bridge

Chronicle Queue Stream Bridge acts as a bridge between `MarshallableIn` capable objects such as Chronicle Queue, Chronicle Network and Chronicle Wire and standard `java.util.stream.Stream` objects.

== Stream Bridge Overview

The Stream Bridge allows standard `Stream`, `LongStream` and `DoubleStream` objects to be created directly from `MarshallableIn` objects such as a Chronicle Queues.
Here is an example how a Chronicle Queue that contains `MarketData` objects can be converted and used as a Stream:

[source,java]
----
String s = Streams.of(
               queue.createTailer(),                                <1>
               DocumentExtractor.builder(MarketData.class).build()  <2>
         )
        .skip(100)
        .limit(50)
        .map(Object::toString)
        .collect(Collectors.joining(","));
----

<1> The `DocumentExtractor` to use when reading excerpts from the Queue.
The tailer can be set to a certain configuration and/or position before being submitted to the method above.

<2> The `DocumentExtractor` used to convert an excerpt to a Java object.
In the example above, an extractor of type `MarketData` is used which is obtained via a builder function `DocumentExtractor.builder()`.

The `Stream` above shows an example of a pagination service where the third page of events from the Queue is shown (each page has 50 rows).
Once the `Stream` is created, it behaves like any `Stream` and can take advantage of the whole `Stream` ecosystem.

The snippet above might produce the following output (shortened):

[source,text]
----
!MarketData {
  symbol: MSFT,
  last: 100.0,
  high: 110.0,
  low: 90.0
}
,!.MarketData {
  symbol: APPL,
  last: 200.0,
  high: 220.0,
  low: 180.0
}
,!MarketData {
  symbol: MSFT,
  last: 101.0,
  high: 110.0,
  low: 90.0
}
...
----

== Stream Bridge Details

This chapter contains a more detailed description of the Stream Bridge functionality.
First, a number of examples are shown so that the overall use cases can be more easily understood.
After that, a more formal description is presented.

=== Examples

This chapter contains a number of Stream Bridge examples.

==== LongStream: Finding the Highest Index

This example is using a `LongStream` eliminating object creation:

[source,java]
----
long last = Streams.ofLong(
                q.createTailer(),
                ToLongDocumentExtractor.extractingIndex() <1>
         )
        .max()
        .orElse(-1);                                     <2>

----

<1> Convenience method equivalent to `(wire, index) -> index`.
<2> If no index is present, return `-1`.

==== LongStream: Computing Statistics

This other example is also using a `LongStream` but with a custom extractor that will read a `long` directly from the queue.

[source,java]
----
LongSummaryStatistics stat = Streams.ofLong(
               q.createTailer(),
               (wire, index) -> wire.getValueIn().readLong() <1>
        )
        .summaryStatistics();
----

<1> Custom `ToLongDocumentExtractor` lambda.

==== Stream: Objects of a Certain Type

This example shows an example of creating a `Stream`.

[source,java]
----
Stream stream = Streams.of(queue.createTailer(), builder(MarketData.class).build());
----

The `DocumentExtractor.builder(Class type).build()` construct is equivalent to:

[source,java]
----
(wire, index) -> wire
    .getValueIn()
    .object(type);
----

==== Stream: Objects from a Queue written via MethodWriters

If a queue was written using a method writer, the queue will likely contain messages of different types.
Assuming we have used a method writer implementing:

[source,java]
----
public interface Messages {

    void shares(Shares shares);

    void news(News news);

    void greeting(String greeting);

}
----

we can extract messages of a certain type like so:

[source,java]
----
List newsList = Streams.of(
                    q.createTailer(),
                    builder(News.class)
                        .withMethod(Messages.class, Messages::news)
                        .build()                                           <1>
            )
            .sorted(Comparator.comparing(News::symbol))                    <2>
            .collect(toList());
----

<1> Creates a `DocumentExtractor` that will extract `News` messages that was previously written to the queue using a method writer's `Messages::news` method.

<2> Standard `Stream` operation that will return a `Stream` sorted in `symbol` order.

The extractor will only extract messages of the specified type and method and not other messages.

==== Iterators and Spliterators

Streams, Spliterators and Iterators are related.
The Stream Bridge feature also supports creating various Spliterators and Iterators.
Here is an example:

[source,java]
----
Iterator iterator = Streams.iterator(
                queue.createTailer(),
                builder(MarketData.class).build()      <1>
);
----

<1> Extract messages of this type while iterating.

As can be seen, the procedure here is similar to creating a `Stream`.

==== Parallel Streams

Streams handle thread-safety issues with `MarshallableIn` objects but the provided `MarshallableIn` must be able to run on different threads (e.g. cannot have `ThreadLocal` variables) or else the result is undefined.

=== Constrains and Best Practices

This chapter contains tips for using the Stream Bridge feature.

==== Object Reuse

It is possible to create a `DocumentExtractor` that is reusing objects.
Care must be taken if such an extractor is used so that, for example, reused objects are not exposed and/or not stored internally in the Streams pipeline.
If in doubt, make a copy or extract an immutable value from the object at hand.

[source,java]
----
OptionalDouble max = Streams.of(queue.createTailer(),
               builder(MarketData.class)
                        .withReusing(MarketData::new)    <1>
                        .build())
        .mapToDouble(MarketData::last)                   <2>
        .max();
----

<1> This supplier is used to provide objects that are reused when successively extracting a plurality of elements.
<2> A primitive `double` value is extracted directly making reuse safe.

Here is an example of object reuse that is *not allowed*:

[source,java]
----
List list = Streams.of(queue.createTailer(),
                builder(MarketData.class)
                        .withReusing(MarketData::new)  <1>
                        .build())
        .collect(toList());                            <2>

System.out.println("list = " + list);
----

<1> This supplier is used to provide objects that are reused when successively extracting a plurality of elements.
<2> As objects are reused, the list will be populated with identical objects that will reflect the latest value of the reused object.

This might print something like this:

[source,text]
----
list = [!MarketData {
  symbol: MSFT,
  last: 101.0,
  high: 110.0,
  low: 90.0
}
, !MarketData {
  symbol: MSFT,
  last: 101.0,
  high: 110.0,
  low: 90.0
}
, !MarketData {
  symbol: MSFT,
  last: 101.0,
  high: 110.0,
  low: 90.0
}
]
----

==== Closing an ExcerptTailer

In the examples above, an `ExcerptTailer` was created on demand but was not properly closed.
In memory sensitive applications, it is recommended that this is taken care of as tailers may have allocated internal resources.
Here is an example:

[source,java]
----
Map> groups;
try (ExcerptTailer tailer = queue.createTailer()) {
    groups = Streams.of(tailer, builder(MarketData.class).build())
            .collect(groupingBy(MarketData::symbol));
}    <1>

groups...
----

<1> The tailer is auto-closed here

==== GC and Object creation

Streams are likely to create objects during construction and use.
Therefore, the Stream Bridge features are not recommended in the same JVM as deterministic low-latency applications.
Once these objects are reclaimed by the Garbage Collector, jitter may be incurred on executing Threads.

== To be Documented or todo

* Parallel streams (Thread-safe ExcerptTailer)
* DocumentExtractor.ofType() error handling (what if there is another message on the queue?)
* Extractors' use of `null` and `Long.MIN_VALUE`
* Maybe provide a Supplier so that streams can close the tailer after use?




© 2015 - 2024 Weber Informatics LLC | Privacy Policy