Package brave.propagation

Class Propagation.Factory

java.lang.Object

brave.propagation.Propagation.Factory

Enclosing interface:

Propagation<K>

public abstract static class Propagation.Factory
extends Object

Constructor Summary

Constructors

Constructor	Description
Factory()

Method Summary

Modifier and Type	Method	Description
abstract <K> Propagation<K>	create(Propagation.KeyFactory<K> keyFactory)
TraceContext	decorate(TraceContext context)	Decorates the input such that it can propagate extra data, such as a timestamp or a carrier for extra fields.
boolean	requires128BitTraceId()
boolean	supportsJoin()	Does the propagation implementation support sharing client and server span IDs.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

Factory

public Factory()

Method Detail

supportsJoin

public boolean supportsJoin()

Does the propagation implementation support sharing client and server span IDs. For example, should an RPC server span share the same identifiers extracted from an incoming request? In usual B3 Propagation, the parent span ID is sent across the wire so that the client and server can share the same identifiers. Other propagation formats, like trace-context only propagate the calling trace and span ID, with an assumption that the receiver always starts a new child span. When join is supported, you can assume that when the parent span ID is null, you've been propagated a root span. When join is not supported, you must always fork a new child.

requires128BitTraceId

public boolean requires128BitTraceId()

create

public abstract <K> Propagation<K> create(Propagation.KeyFactory<K> keyFactory)

decorate

public TraceContext decorate(TraceContext context)

Decorates the input such that it can propagate extra data, such as a timestamp or a carrier for extra fields.

Implementations are responsible for data scoping, if relevant. For example, if only global configuration is present, it could suffice to simply ensure that data is present. If data is span-scoped, an implementation might compare the context to its last span ID, copying on write or otherwise to ensure writes to one context don't affect another.

Implementations should be idempotent, returning the same instance instead of re-applying change.

See Also:

TraceContext.extra()