thelarkinn / bundler-performance-benchmark

This is a super tiny example of a transparent comparison between parcel and webpack and anyone else who wants to be involved.

bundler-performance-benchmark

Why

Showing numbers means nothing if you can't back it up. This is a joint effort between all bundlers involved to create a fair and unbiased set of benchmarks for build and size performance.

How to Run

• Clone repo
• run yarn
• run yarn build:webpack && yarn build:parcel
• Run again for cache times

Single Entry, Production, Medium Graph, JS Only

bundler	version	build time	build time w/ cache	bundle size	config LOC
webpack	webpack/webpack#next	340ms	184ms	5.72kb	15
parcel-bundler	1.2.1	1.78s	473ms	21.094kb	0

Parcel Recipes (Preact)

bundler	version	build time	build time w/ cache	bundle size	config LOC
webpack	webpack/webpack#next	1172ms	700ms	9.69k	20
parcel-bundler	1.2.1	1270ms	266ms	11.560kb	5

Help Wanted

Need help organizing and setting up multiple real world scenarios for all bundlers to run.

Feel free to checkout existing example or optimize the workflow via PR.

bundler-performance-benchmark's Issues

Add real-world apps instead of tiny benchmark

X nominated: https://github.com/opencollective/frontend

I think this is a great one.

webpack version differences?

In the package.json I see webpack next; what version does that specifically point to? I think for a benchmark, especially in the rapidly evolving world of JS tools, to be specific so that a better comparison can be had.

For example, the computer language shoot out does this for each benchmark suite: https://benchmarksgame.alioth.debian.org/u64q/nbody.html

They show Rust with 3 different programs; I think it makes sense to show webpack at 2 different versions, next and whatever is the currently used version. Same with parcel, maybe pick latest + current production standard version?

Let’s add FuseBox

Hey Sean, nice benchmark, I also feel that’s its not fair to show some numbers out of thin air, claiming to be “blazing fast” which is fuse motto btw xD

You think I can dig in and add fuse benchmark there?

UPD:
I think it's fair to let parcel and fuse-box to use cache, as well as not uglifying the output. Besides, production builds are made differently as opposed to development flow. That's what i got

parcel: 67.12ms | 50.58 KiB
fuse-box: 11.04ms | 44.10 KiB
webpack: 81.93ms | 5.90 KiB

UPD2:
Cache off, for fuse-box and parcel

fuse-box: 42.44ms | 36.74 KiB
parcel: 161.83ms | 38.97 KiB
webpack: 61.63ms | 9.70 KiB

Create readme with install instrauctions and how to run the benchmark

• yarn
• yarn build:webpack
• yarn build:parcel
• yarn build:parcel (for cache)

Add webpack3 to benchmark

Should find a way to be able to run webpack 3 also. Maybe there is a way to leverage yarn aliases.