FormidableLabs/react-ssr-prepass

what could we do while ssr-prepass and renderToNodeStream

sayjeyhi opened this issue · 2 comments

sayjeyhi commented

There is a gap between we are waiting to ssrPrepass(element) finish (api calls finish) and renderToNodeStream which increase TTFB, is there any idea about doing sth in this gap.

for example if we could get helmet needs data and stream header before finishing the whole ssr-prepass promise..

kitten commented

So if you don’t have a lot of async tasks you can also try out react-lightyear: https://github.com/Ephem/react-lightyear

It only does a single render pass but comes with other caveats. It will reduce TTFB but that will only be useful depending on where your suspense boundary is.

It may be worth though for you to look into CDN caching if possible, since that’ll always have the greatest impact.

A middle ground may be to move the entire schema execution into the SSR server. That could have event loop implications as your SSR server does more work, but since you’ll execute the schema in the same process, you also have more in-memory caching options. (That depends on this package being shipped urql-graphql/urql#474)

So overall you have several options but they really depend more on your architecture and on finding small workarounds since this is inherently an issue with SSR :)

  • 👍1
  • ❤️2
sayjeyhi commented

Thank you dear @kitten, actually I take a look at react-lightyear and give it a try, I see it is not the thing I am looking for. Specially using throw to handle api call suspense, gets a little bit complicated redux or saga.

I see react-ssr-prepass is doing better my tasks. I got an idea and it improved my TTFB, I write html head with main js and css file to be preloaded, it decreased my TTFB to 100ms. In other hand it helped me to preload my main js file parallel with prepassing SSR needs api calls.

const statsFile = path.resolve(
    `./build/loadable-stats.json`,
);
const chunkExtractor = new ChunkExtractor({ statsFile });
const { stats } = chunkExtractor;

res.set({ 'content-type': 'text/html; charset=utf-8' });
res.write('<!doctype html><html lang="fa"><head>
<link rel="preload" href="${stats.assetsByChunkName.main}" as="script" />
<script async src="${stats.assetsByChunkName.main}" ></script>
');

Maybe some one asks about handling 404,301 cuz I am writing data on response before deciding about route existence or redirection, for that purpose I used matchPath from react-router-dom so I can predict them.

  • 👍1