req & res in http.createServer have not the same map

[welefen]

• Version: v4.4.3
• Platform: Mac os
• Subsystem:

var http = require('http');

var prevReq = null;
var prevRes = null;
var server = http.createServer(function(req, res){
  if(!prevReq){
    prevReq = req;
    prevRes = res;
  }else{
    console.log('req', %HaveSameMap(req, prevReq));
    console.log('res', %HaveSameMap(res, prevRes));
  }
  res.end('hello word');
});

server.listen(3337);

console.log('server running at http://127.0.0.1:3337');

use node --allow-natives-syntax index.js to test.

[evanlucas]

From what I've seen, the _idlePrev and _idleNext properties on both req and res can have different maps.

[welefen]

https://github.com/nodejs/node/blob/master/lib/_http_incoming.js#L71 , This Line add dynamic property for IncomingMessage, It caused IncomingMessage instance have not the same map.

[Fishrock123]

I don't understand how this is a problem...

Closing, let us know if we need to re-open.

[bnoordhuis]

@Fishrock123 It means request and response objects are implemented much less efficiently than they can be. The issue is legitimate, I'll reopen (and try to look at it this week.)

[welefen]

Request and Reponse object in createServer has not same map, So it can not use Hidden Classes optimizing in V8, Hidden Classes is about 2x faster when get class instance. https://developers.google.com/v8/design#fast-property-access

[bnoordhuis]

I had a look and I filed #7003 as a basic sanity check.

It doesn't catch this particular issue but I've come to believe the test case is something of a red herring because it's comparing objects at different phases in their life cycle. If you rewrite the test to only check at the end (like I did in #7003), everything is fine.

That's not to say the http module is currently a paragon of efficiency - it's not - but this particular issue doesn't seem valid. I'll close it for now.