The JavaScript Site Generator Review, 2023
It’s time again for every framework author’s favorite: The JavaScript Site Generator Review, February 2023 edition.
Here is the list of site generators reviewed in alphabetical order:
- Astro
2.0.15
- Eleventy
2.0.0
- Enhance
1.4.6
(added 27 February 2023) - Gatsby
5.7.0
- Next.js
13.2.1
- Nuxt
3.2.2
- Remix
1.13.0
- SvelteKit
1.8.3
Each example site was created using a fresh run of the Quick Start or Getting Started tutorial on each respective generator’s web site.
Previous Reviews
- Next.js v13 review—October 2022
- Who Pays for Web Frameworks?—Last updated October 2022
- Markdown benchmark and
npm install
Benchmark—July 2022 - Remix v1.5.1 review—June 2022
- Next.js review—December 2021
npm install
Times
Times shown are in seconds. Lower is better.
Expand to show table of results
Framework | npm install Time |
---|---|
Astro 2.0.15 |
12.52 |
Eleventy 2.0.0 |
5.81 |
Enhance 1.4.6 |
29.71 |
Gatsby 5.7.0 |
43.36 |
Next.js 13.2.1 |
3.72 |
Nuxt 3.2.2 |
14.77 |
Remix 1.13.0 |
40.14 |
SvelteKit 1.8.3 |
6.78 |
Notes:
- Each tool was installed five times and the average of those five runs is shown above.
npm cache clean --force
was used to clean the cache before each install.- npm auditing was enabled for these tests.
- Next bundles all of its dependencies (save for
react
andreact-dom
) and as such bypasses for-free minor dependency releases and most of the auditing functionality provided by npm.
Client JavaScript Baseline
Weights shown are in kilobytes of uncompressed resource size (not transfer size).
Expand to show table of results
Framework | Client JavaScript Baseline (kB) |
---|---|
Astro 2.0.15 |
0 |
Eleventy 2.0.0 |
0 |
Enhance 1.4.6 |
0 |
Gatsby 5.7.0 |
210 |
Next.js 13.2.1 |
248 |
Nuxt 3.2.2 |
191 |
Remix 1.13.0 |
228 |
SvelteKit 1.8.3 |
53 |
node_modules
Weight
Weights shown are in megabytes.
Expand to show table of results
Framework | node_modules Weight (MB) |
---|---|
Astro 2.0.15 |
169 |
Eleventy 2.0.0 |
34 |
Enhance 1.4.6 |
210 |
Gatsby 5.7.0 |
583 |
Next.js 13.2.1 |
158 |
Nuxt 3.2.2 |
164 |
Remix 1.13.0 |
497 |
SvelteKit 1.8.3 |
111 |
npm Auditing
Some frameworks use create-*
packages (sometimes with npm init
) and include a step that helpfully installs your npm
dependencies for you. However, some of these tools also hide npm auditing results (including those about high severity security vulnerabilities).
These tools hide npm audit
reports:
- Astro ⚠️
- Gatsby ⚠️
- Next ⚠️ (via bundling its dependencies)
These tools show standard npm audit
reports during project creation:
- Eleventy ✅
- Enhance ✅
- Nuxt ✅
- Remix ✅
- SvelteKit ✅
Telemetry
Telemetry is the practice of collecting anonymous data on usage of the tool. Some tools have this enabled (opt-in) by default and require an opt-out. This isn’t necessarily bad but some folks are very surprised by this default (and education here is important as this default may not comply with your organization’s security policies).
These tools require you to opt-out of Telemetry:
- Astro ⚠️
- Next.js ⚠️
- Gatsby ⚠️
Has Telemetry but it prompts you to ask:
- Nuxt (see additional nuance below)
No known Telemetry or data collection:
- Eleventy ✅
- Enhance ✅
- Remix ✅
- SvelteKit ✅
February 28, 2023 this note was updated with more accurate information on Nuxt with help from Nuxt maintainers.
Notably and delightfully, Nuxt prompts to ask if you want to participate in Telemetry when you first run the development server. This is a global preference and you will only be prompted once; it applies to all Nuxt projects on your machine (stored in a ~/.nuxtrc
file).
This prompt is still technically opt-out when using the development server. The default answer to the prompt is Yes, to participate in telemetry. But the prompt does not execute on deployment servers and CI pipelines, so it’s actually opt-in in those contexts. To me this seems like a well-executed middle ground.
Feedback
Let me know if you have feedback—I’d also welcome recommendations on aspects of these tools that you’d like to see reviewed for next time!
14 Comments
Zach Leatherman :11ty:
(it has evolved to its final form: the blog post)
Matthias Ott
@zachleat (good)
Alex Russell
@zachleat per other thread, would love to see gzip sizes for default JS, and also, is this a script I can run? If not, would you like for one to exist?
Zach Leatherman :11ty:
@slightlyoff whew, yes—I would love to have this automated.I don’t think it would be maintainable to do so, unfortunately, I’d expect a fair bit of churn in framework “Getting Started” happy paths?Certainly not worth it for how often I do these (once or twice a year)
Alex Russell
@zachleat will noodle on it
Steffen
@zachleat Hmmm I guess I have some objections regarding the client side JS part. I think those which ship client side bundles here shouldn’t be compared to those which don’t, just like that. They represent a different architectural style.Some words about that … that it’s not appl… Truncated
Steffen
@zachleat Apart from that it’s nice to see those comparisons nevertheless! Thanks!
Zach Leatherman :11ty:
@tevo ah—just to be clear I don’t accept “sites versus apps” as a useful dichotomy when building things for the web—and further I would argue that this divide has done great damage!I might also say it like this, a bit more cheekily: https://www.zachleat.com/twitter/15393323225495… Truncated
Brian LeRoux 💚
@andy if remix counts why would enhance.dev not count ? Idk fwiw (cc @zachleat )
Andrew Walpole
@brianleroux @andy @zachleat Too many level heads, not enough #hotdrama twitter marketing 🌶️ hotdrama
Brian LeRoux 💚
@walpolea @andy @zachleat we all like html here now I think if I read the drama correct 💀
Zach Leatherman :11ty:
@krinkle @slightlyoff I did this with my markdown benchmark last year since it was doing build performance. If folks are interested in another repo of starters I can open source it but I didn’t do it this time because I wasn’t running any build performance tests? Maybe that separ… Truncated
gotofritz
@zachleat is "node_modules WEIGHT" really something worth measuring and worrying about?
@cvekaso
lol, this made me recheck who wrote the article zachleat.com/web/site-gener… 😅