Reduce build duration by lowering Mockito usage #2707
Description
🤔 What's the problem you've observed?
Mockito has a impact on the unit test execution time (seen using IntelliJ Profiler flame graph):
| project | test duration [s] | mockito contribution |
|---|---|---|
| cucumber-junit | 4.4 | 31% = 1.3 s |
| cucumber-core | 7.5 | 10% = 0.7 s |
| cucumber-java | 2.8 | 24% = 0.6 s |
| cucumber-java8 | 1.8 | 27% = 0.5 s |
| cucumber-spring | 4.5 | 13% = 0.5 s |
| cucumber-testng | 1.5 | 20% = 0.3 s |
| cucumber-guice | 1.8 | 15% = 0.3 s |
| datatable | 0.5 | (unused dependency) 0% = 0 s |
Total build duration: 3min17 = 197 seconds
Total test duration : 23.5 seconds
Total Mockito contribution : 4.2 seconds (18% of tests, 2% of build duration)
By reducing the usage of Mockito, the test duration can be lowered with two advantages:
- this gives the developer a shorter feedback loop.
- the energy consumption is lower (green IT), which is important because the cucumber-jvm project is built often (due to PR and renovate bot). 2% is not a lot, but on the long term, it makes a difference.
✨ Do you have a proposal for making it better?
I didn't check in details, but for example in JavaBackendTest, the @Mock ObjectFactory and Glue could be replaced by fake implementations, which could reduce the class test duration by about 80%.