I haven’t done any serious work with either, but I’ve been following both closely (and have contributed a bit to Astro early on). So this isn’t the hands-on response you specifically asked for, and may not contain new information to you. I’m posting anyway in case it adds context for others.
Qwik City[1] is probably more directly analogous to Astro, Qwik being more analogous to Astro’s integrated renderers. But that highlights one of the key differences.
Astro’s compiler mostly focuses on server rendering of static content (.astro templates, MDX) and bundling client resources along with the logic necessary to hydrate islands. Astro defers to those renderers (and in some cases their own compilers) for any further optimization of the client bundle.
Qwik’s compiler optimizes the component code directly, serializing state into the HTML it renders server-side, for the client bundle to resume from that state. Its output is conceptually similar to Phoenix LiveView (which was mentioned in another sub-thread).
Both are compelling approaches. I think Qwik’s will probably (eventually) have an optimization advantage because that’s a core focus of the client library. Astro will likely have an adoption advantage because it’s client-library-agnostic.
Another framework in the space often gets passed over: Marko[2], which has been doing partial hydration for years at eBay. Marko is probably more similar in approach to Qwik (and as I understand it, getting more similar as they’re going resumable too), but like Astro has its own templating language which enables its compiler optimizations.
Also worth watching SolidJS[3] (whose creator has also worked on Marko), which is tracking partial hydration/resumability on its roadmap. I’m not sure what their approach will look like but there’s quite a lot of insight both in the issue and the creator’s tweets/replies on the topic.
Personally I think there’s a gap between all of these approaches which could leverage type-level analysis to go much further. But that isn’t really feasible when types being available or accurate isn’t a safe assumption.
Qwik City[1] is probably more directly analogous to Astro, Qwik being more analogous to Astro’s integrated renderers. But that highlights one of the key differences.
Astro’s compiler mostly focuses on server rendering of static content (.astro templates, MDX) and bundling client resources along with the logic necessary to hydrate islands. Astro defers to those renderers (and in some cases their own compilers) for any further optimization of the client bundle.
Qwik’s compiler optimizes the component code directly, serializing state into the HTML it renders server-side, for the client bundle to resume from that state. Its output is conceptually similar to Phoenix LiveView (which was mentioned in another sub-thread).
Both are compelling approaches. I think Qwik’s will probably (eventually) have an optimization advantage because that’s a core focus of the client library. Astro will likely have an adoption advantage because it’s client-library-agnostic.
Another framework in the space often gets passed over: Marko[2], which has been doing partial hydration for years at eBay. Marko is probably more similar in approach to Qwik (and as I understand it, getting more similar as they’re going resumable too), but like Astro has its own templating language which enables its compiler optimizations.
Also worth watching SolidJS[3] (whose creator has also worked on Marko), which is tracking partial hydration/resumability on its roadmap. I’m not sure what their approach will look like but there’s quite a lot of insight both in the issue and the creator’s tweets/replies on the topic.
Personally I think there’s a gap between all of these approaches which could leverage type-level analysis to go much further. But that isn’t really feasible when types being available or accurate isn’t a safe assumption.
1: https://qwik.builder.io/qwikcity/overview
2: https://markojs.com/
3: https://www.solidjs.com/