Skip to content

README.md

Latest commit

 

History

History
80 lines (52 loc) · 1.94 KB

README.md

File metadata and controls

80 lines (52 loc) · 1.94 KB

selector-testing

Module purpose

selector-testing contains Cucumber end-to-end tests for the selector-based non-blocking echo implementation. It validates behavior across real classes from:

  • selector-client
  • selector-server

What is covered

Feature files:

  • src/test/resources/features/selector.feature
    • functional scenarios
    • edge-case/format scenarios
    • high-load scenario
  • src/test/resources/features/selector-benchmark.feature
    • selector benchmark scenario (SelectorServer + SelectorClient)
    • persisted summary scenario

Step definition classes:

  • SelectorStepDefinitions - functional and load steps with shared implementation.
  • SelectorBenchmarkStepDefinitions - benchmark-only steps.
  • CommonStepDefinitions - shared @Given server startup and @After cleanup.

Tags

  • @Performance - benchmark scenario
  • @PerformanceSummary - reads persisted benchmark summary and prints final result

Benchmark flow

Benchmark step:

  • runs warmups (not measured)
  • runs measured iterations
  • prints per-run and aggregate metrics
  • persists summary under target/performance-results/selector-selector.properties

Summary step:

  • reads persisted benchmark summary
  • prints [PERF][FINAL] output for this selector case
  • prints total measured time and total benchmark scenario time

How tests are run

Runner:

  • dev.nklip.javacraft.echo.selector.CucumberRunner

Reports:

  • target/cucumber-reports/cucumber.html
  • target/cucumber-reports/cucumber.json

Execute from project root

Run full suite for this module:

mvn -pl echo/selector/selector-testing -am test

Run only benchmark scenario:

mvn -pl echo/selector/selector-testing -am test -Dcucumber.filter.tags='@Performance'

Run only benchmark summary:

mvn -pl echo/selector/selector-testing -am test -Dcucumber.filter.tags='@PerformanceSummary'

If @PerformanceSummary runs before benchmark results exist, it fails by design.