Ubench Laravel Package

Getting Started

Start by installing via Composer: composer require --dev devster/ubench. Create a Benchmark instance at the top of your test script or inline in code you want to profile. Use start() before the block and stop() after — the class stores the interval and memory delta automatically. For the first real use, benchmark a trivial operation (e.g., string concatenation vs. implode()) and call getReport() to see formatted output. Check the examples/ folder in the repo (if available) or inspect Benchmark::start(), Benchmark::stop(), and Benchmark::getReport() signatures in the source.

Implementation Patterns

• Inline development benchmarks: Wrap suspected hot paths with start/stop blocks to compare refactorings or algorithms during development.
• Regression guard: Add a simple Benchmark in your test suite’s development/test runner to assert performance thresholds (e.g., ensure a utility function never exceeds 2ms for typical inputs).
• Comparison matrix: Run multiple variants of a function in a loop, collecting reports into an array, then output a table manually or via implode() for quick side-by-side analysis.
• Logging + monitoring: Hook getReport() output into your dev-mode logging (e.g., info() with Monolog) to catch accidental slowdowns in local dev.
• Test-driven performance: Use ->getElapsedMicroseconds() and ->getMemoryDelta() in assertions to prevent regressions (e.g., $this->assertLessThan(0.001, $bench->getElapsedSeconds())).

Gotchas and Tips

• Warm-up matters: The first run of a function is often slower due to autoloading/JIT (if applicable). Run benchmarks multiple times and discard the first sample when comparing.
• Memory delta can be negative: PHP’s memory management sometimes reclaims memory during execution — use getPeakMemoryUsage() instead of raw delta for stable measurements.
• Micro-benchmarks ≠ real-world: This tool measures isolated code; avoid over-interpreting results without application-level profiling (e.g., Blackfire/Xdebug).
• No built-in statistics: You must manage repetitions and averaging yourself (e.g., in a for loop). Consider wrapping Benchmark in a simple BenchmarkRunner utility.
• No PHP 8.1+ deprecation warnings in tests: Given the last release (2019), test for strict type compatibility — some microtime(true) usages may need explicit float casting in newer PHP versions.
• Reset is manual: If reusing a Benchmark instance, call reset() to clear previous runs — don’t rely on re-instantiation in tight loops.