org.avaje.metric

Interface TimedMetric

All Superinterfaces:

Metric

public interface TimedMetric
extends Metric

A TimedMetric for measuring execution time for methods and events.

Method Summary

Modifier and Type	Method and Description
void	addEventDuration(boolean success, long durationNanos) Add an event duration in nanoseconds noting if it was a success or failure result.
void	addEventSince(boolean success, long startNanos) Add an event based on a startNanos (determined by System.nanoTime()).
Map<String,String>	attributes() Return extra attributes that can be included in the request logging.
void	decrementCollectionCount() Decrement the request timing collection count.
String	getBucketRange() Return the bucket range or empty string if not a bucket.
int	getRequestTimingCollection() Return the number of remaining requests to collect detailed timing on.
boolean	isActiveThreadContext() Return true if this TimedMetric has been pushed onto an active context for this thread.
boolean	isBucket() Return true if this timed metric is part of a bucket range (and hence only hold statistics for the bucket range returned by bucketRange().
void	operationEnd(int opCode, long startNanos) Add an event duration with opCode indicating success or failure.
void	operationEnd(int opCode, long startNanos, boolean activeThreadContext) Add an event duration with opCode indicating success or failure.
void	setRequestTimingCollection(int collectionCount) Specify to collect per request detailed timing collection.
TimedEvent	startEvent() Start an event.

Methods inherited from interface org.avaje.metric.Metric

clear, collect, getName

Method Detail

startEvent

TimedEvent startEvent()

Start an event.

At the completion of the event one of TimedEvent.endWithSuccess(), TimedEvent.endWithError() or TimedEvent.end(boolean) is called to record the event duration and success or otherwise.

This is an alternative to using addEventSince(boolean, long) or addEventDuration(boolean, long). Note that this startEvent() method has slightly higher overhead as it instantiates a TimedEvent object which must be later GC'ed. In this sense generally addEventSince() is the preferred method to use.

addEventSince

void addEventSince(boolean success,
                   long startNanos)

Add an event based on a startNanos (determined by System.nanoTime()).

Success and failure statistics are kept separately.

This is an alternative to using startEvent(). Note that using startEvent() has slightly higher overhead as it instantiates a TimedEvent object which must be later GC'ed. In this sense generally addEventSince() is the preferred method to use.

addEventDuration

void addEventDuration(boolean success,
                      long durationNanos)

Add an event duration in nanoseconds noting if it was a success or failure result.

Success and failure statistics are kept separately.

This is an alternative to using addEventSince(boolean, long) where you pass in the duration rather than the start nanoseconds.

isBucket

boolean isBucket()

Return true if this timed metric is part of a bucket range (and hence only hold statistics for the bucket range returned by bucketRange().

getBucketRange

String getBucketRange()

Return the bucket range or empty string if not a bucket.

operationEnd

void operationEnd(int opCode,
                  long startNanos,
                  boolean activeThreadContext)

Add an event duration with opCode indicating success or failure. This is intended for use by enhanced code and not general use.

Although this looks redundant and a little ugly having this method means that for enhancement the added byte code is minimised. In the case where metric collection is turned off overhead is limited to a System.nanoTime() call and a noop method call.

operationEnd

void operationEnd(int opCode,
                  long startNanos)

Add an event duration with opCode indicating success or failure. This is intended for use by enhanced code and not general use.

isActiveThreadContext

boolean isActiveThreadContext()

Return true if this TimedMetric has been pushed onto an active context for this thread.

This means that the current thread is actively collecting timing entries and this metric has been pushed onto the nested context.

setRequestTimingCollection

void setRequestTimingCollection(int collectionCount)

Specify to collect per request detailed timing collection. The collectionCount is the number of requests to collect detailed timing for and then automatically turn off.

This is expected to only be explicitly called on 'top level' metrics such as web endpoints. Once a request timing context has been created by the top level metric then 'nested metrics' (typically service and data access layers) can append to that existing context. In this way detailed per request level timing entries can be collected for only selected endpoints.

getRequestTimingCollection

int getRequestTimingCollection()

Return the number of remaining requests to collect detailed timing on.

This value starts out as the value set by #setRequestTimingCollection and then decrements after a request timing is reported until it reaches 0 at which point request timing automatically turns off for this metric.

decrementCollectionCount

void decrementCollectionCount()

Decrement the request timing collection count.

This is typically called internally when a request timing is reported and generally not expected to be called by application code.

attributes

Map<String,String> attributes()

Return extra attributes that can be included in the request logging.