Skip to content

bundle-analysis.md

Latest commit

 

History

History
81 lines (54 loc) · 2.44 KB

bundle-analysis.md

File metadata and controls

81 lines (54 loc) · 2.44 KB

Benchmarks and Bundle Analysis

The package includes two separate tools:

  • Runtime benchmarks for key fint-i18n hot paths.
  • Bundle analysis to understand the composition of the built dist.

Runtime Benchmarks

To run:

yarn --cwd fint-i18n bench

What the command does:

  • Runs benchmark files from bench/**/*.bench.ts via vitest bench.
  • Saves machine-readable results to dist/analysis/bench-results.json.
  • Prints ops/sec for each scenario to the console.

Current scenarios in bench/core.bench.ts:

  • Cold compile for compileTemplate().
  • Steady-state translation via i18n.t() after cache warmup.
  • Nested lookup for an already compiled key.
  • setLocale() with already registered used blocks.

What the console output means:

  • name — Name of the benchmark scenario.
  • hz / ops/sec — Operations per second; higher is faster.
  • min, max, mean — Minimum, maximum, and average time for a single operation in milliseconds; lower is better.
  • p75, p99, p995, p999 — Execution time percentiles; showing how expensive the "tails" are.
  • rme — Relative margin of error; lower means more stable results.
  • samples — Number of measurements collected for statistical calculation.

Important:

  • Results should only be compared between runs on the same machine under similar conditions.
  • This is not a microbenchmark of the entire Node runtime, but applied scenarios for fint-i18n itself.

Bundle Composition Analysis

To run:

yarn --cwd fint-i18n build:analyze

What the command does:

  • Builds the library as usual.
  • Enables rollup-plugin-visualizer only during analysis.
  • Places reports in dist/analysis/.

Artifacts:

  • dist/analysis/stats.html — Interactive treemap.
  • dist/analysis/stats.md — Text summary for reading and code review.
  • dist/analysis/stats.json — Raw data report for automated processing or comparison.

How to read the reports:

  • stats.html helps quickly see which modules and chunks take up the most space.
  • stats.md is convenient for use in PRs or local audits when a text summary is needed.
  • stats.json is useful for subsequent machine analysis or custom scripts.

Usage Practice

Minimum cycle after changing the runtime:

yarn --cwd fint-i18n bench
yarn --cwd fint-i18n build:analyze

This answers two different questions:

  • Has the performance of the hot path degraded?
  • Has the composition/size of the built bundle increased without reason?