-
-
Notifications
You must be signed in to change notification settings - Fork 689
Closed
Labels
bugSomething isn't workingSomething isn't working
Description
Bug Description
parseHeaders in the utils.js seems too slow.
Reproducible By
const dispatcher = new Pool('http://127.0.0.1:8090' /* an 4worker nginx */, {
pipelining: 30,
connections: 100,
});
const proxy = http.createServer((req, res) => {
dispatcher.request({ // the constructor of RequestHandler in undici/lib/api/api-request.js
method: req.method,
path: req.url,
headers: req.headers,
body: req,
}, (err, { statusCode, headers, body /** a Readable */ }) => {
if (err) {
throw err
}
res.writeHead(statusCode, headers);
pipe(body, res);
});
});
function pipe(src, dst) {
src.on('readable', () => {
let chunk;
while (null !== (chunk = src.read(65536))) {
dst.write(chunk);
}
});
src.on('end', () => {
dst.end()
});
}Expected Behavior
Less CPU time spent in parseHeaders
Logs & Screenshots
Environment
Node 22
linux
Additional context
I open this issue because I've spotted some optimization for lowercasing the headers in node's built-in HTTP module, https://github.com/nodejs/node/blob/2e458d973638d01fcb6a0d7d611e0120a94f4d35/lib/_http_incoming.js#L279C3-L279C3
basically, it matches the known header field and use the cached string if found. Maybe we can adapt the same trick here to eliminate the toString() and toLowercase() overhead in parseHeaders.
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working