uws examples are misleading

[bompus]

While using uws yields some good performance benefits, trying to actually use it results in some issues.

The first issue I came across, is a fairly important one. Trying to add response headers in both Express and Koa do nothing. All that comes back is the Content-length header, nothing else.

const http = require("uws").http;
// const http = require("http");

Koa:
ctx.set("Cache-Control", "no-cache");
ctx.body = "Hello World";

Express:
res.set("Cache-Control", "no-cache");
res.send("Hello World");

If you use the native http require, it works just fine. Because of this behavior, I don't think you can do a simple 1:1 comparison and have uws claim victory, when it's doing less work.

Perhaps I am doing something wrong, so please check for yourself and let me know if you see any way around this. I know that from digging around, the author of Total.js was going to try uws as well, but I believe realized it had this same issue ( https://blog.totaljs.com/blogs/news/20170530-new-total-js-framework-version-2-6/ - A support for another HTTP module )

[hbakhtiyor]

the uws module is not fully compatible with native http native, you'll be face some missing properties/methods/events.

you can open issue at the repo if not working the above examples.

about the above blog, yeah, no has the property yet, but you can use alternative approach,
e.g.

'use strict';

const express = require('express');
const http = require('uws').http;

const app = http.getExpressApp(express);

app.post('/', (req, res) => {
  const body = [];
  req.on('data', (chunk) => {
    body.push(Buffer.from(chunk));
  }).on('end', () => {
    res.end('You posted me this: ' + Buffer.concat(body).toString());
  });
});

http.createServer(app).listen(8000, () => {
  console.log('Example app listening on port 8000!');
});

[bompus]

Interesting. So we end up parsing our own post body, trying to figure out how to send response headers, etc. Yeah, I see how that would cause issues with some frameworks that are expecting http.

Hopefully uws moves along and shims everything up to match, then I'd be curious to see some real 1:1 testing. I'm sure it is still way faster, but not being able to simply drop it in as a replacement halts me from using it at this time.

Thanks for putting together the benchmarks. Didn't want the issue to seem like a bash, just caught my eye immediately when I started testing the idea out :)

[hbakhtiyor]

Anytime ;)