CODEWITHSUNDEEP
javanetbeansmavenjunitprofiling
IAdapter
IAdapter

Reputation: 64915

How do I profile a JUnit test in a NetBeans Maven project?

I tried to do it, but it seems impossible, since you can only profile main methods. I even added a main method to using test, but it just shows the time of that main method. :(

I'm using mocks and abstract testcases, so I cannot convet it into main method.

Any ideas?

Upvotes: 3

Views: 1787

Answers (4)

Wheezil
Wheezil

Reputation: 3472

I've found the Junit framework to be a convenient place to park ad-hoc performance tests. Sure, there are better ways to manage it, but if all you are after is fixing a hotspot and want a profile while doing so, it will suffice. My approach is:

  • Run the unit test
  • While it is running, attach the profiler

You may want to put add a sleep up front, to allow time for you to attach. Quick and dirty.

Upvotes: 0

user2179737
user2179737

Reputation: 551

I would recommend using JMH + a profiler (spf4j, flight recorder)

a example on how you can do that you can see at : http://zolyfarkas.github.io/spf4j/xref-test/org/spf4j/JmhTest.html

If you use jenkins with the jmh plugin: https://github.com/blackboard/jmh-jenkins you can trend your performance, and you will always be able to look at the profiler data to investigate.

Upvotes: 0

oers
oers

Reputation: 18714

You could use jvisualvm from the java installation.

This tool can monitor and profile any running java application.

Upvotes: 3

Mike
Mike

Reputation: 8100

If you're trying to get performance results of the code that you're testing, that's something that's usually done by profiling the actual application as it's running in a real environment (container, executable file, etc.) Thus, I'd posit that attempting to profile the actual application code through unit tests is the incorrect way to do it, as you might not get true performance details. There are multiple reasons for this:

  • The interactions with other objects that you're mocking out in unit tests actually do take time and consume resources that you're thus not getting performance details on.
  • The Java runtime optimizes the bytecode while it's running such that performance typically improves over time
  • I'm sure there are others; these are what comes to me pre-coffee.

If instead you're actually attempting to profile your unit test cases, I'd have to ask why you'd want to do that. So long as they run "fast enough" (which is a subjective term, but "under 10 minutes" is usually a good proxy value) to get developers feedback on whether they pass or fail in a timely manner, then that should be good enough...

Upvotes: 1

Related Questions