July contributions
August 5, 2024
The e18e community has been flourishing lately. We've seen a huge amount of contributions across the ecosystem, all the way upstream to some of the deepest dependencies.
Most of us will have been affected by these contributions somewhere down the line, giving us better performance and a smaller footprint.
The people contributing these changes deserve so many thanks for their work, so here is a summary of some of the highlights from the past month!
svelte-preprocess becomes dependency-free
The svelte team have been busy reducing the footprint of svelte-preprocess.
The before and after values speak for themselves:
- Before: 1.5MB / 26 packages
- After: 97KB / 1 package
You can read about how this was achieve through these two PRs in particular:
- https://github.com/sveltejs/svelte-preprocess/pull/640
- https://github.com/sveltejs/svelte-preprocess/pull/645
codemods
One of the biggest steps forward with the replacements project has certainly been the codemods repo.
If you haven't yet seen the replacements project, it basically provides community driven lists of suggested replacements to npm packages. These are then consumed by various tools, such as the eslint plugin.
The codemods project is another consumer of these lists. It provides a place for the community to contribute codemods which can automatically migrate your code to a given replacement package or native functionality.
@passle_ has done great work here starting and maintaining the codemods project, and the community have contributed many codemods already.
The future of the es-tooling organisation will likely involve a CLI which can apply these codemods for you. Keep an eye out for it in future!
Removing is-number from micromatch
While this hasn't been published yet, it is great progress, showing that even some aged projects are open to performance contributions.
This change removed, in this case, an unnecessary dependency often seen in the ongoing cleanup work.
is-number may be small but is often entirely unnecessary. Most consumers can instead use something as simple as typeof n === 'number' and provide a stricter API rather than trying to coerce inputs.
Credit to @talentlessguy for this contribution!
Removing rimraf from node-pre-gyp
Node has had a recursive option for rmdir (or rm in newer versions) for a while now. This can achieve the same functionality rimraf was providing in many cases.
A small but effective win here is the removal of rimraf from node-pre-gyp.
It is worth mentioning, for those cases where you'd still like a CLI, you can use an alternative.
Credit to @benjaminmccann for this one!
picoquery
As part of the cleanup project, we were in need of a fast and lightweight query-string parser/stringifier.
Some excellent solutions exist for part of this:
URLSearchParamsfor general parsing and stringifying of query stringsfast-querystring: a very fast query string parser with support for arrays
Though a gap existed for the speed of fast-querystring, but the functionality of something like qs (object nesting, arrays, custom delimiters, etc).
This is why picoquery was born! A super fast and small configurable query string library.
Our hope is that we see more and more of these alternatives being produced by the community. Many popular, bloated or outdated packages still need alternatives. Join the discussion if you're interested!
neotraverse
Similar to how picoquery came about, we needed a lighter alternative to the popular traverse package.
Big thanks to @puruvjdev for making this happen by publishing neotraverse!
The savings are great:
- traverse: 3.9MB / 67 packages
- neotraverse: 140KB / 1 package
package-size-calculator
Built by @DevMiner, this is a cool little CLI that can provide package size stats and an estimate of the monthly bandwidth usage.
Many people have been using this already in the e18e community to get a rough idea of potential savings by changing dependencies.
Combining this with npmgraph and pkg-size is a great way to investigate how to improve a package.