fix: fix response binary content handling

[alacroix]

Hey,

First of all, thanks for the project, it's really great.

We found a bug while upgrading to v3, concerning binary content.

It seems that the handling of binary content as response is broken. In convertRes, the parseHeaders function simply stores headers in a Record<string, string> but doesn't take in account the Uppercase / Lowercase thing that we can found in headers. In our case, the headers received are like Content-Type but in the code, it compares to lowercase headers["content-type"]. (Our infra uses lambdas + ApiGateway on top of that like mentioned in the docs)

Here is a small NextJS API route that reproduce the problem in our case.

const IMAGE_BASE64 =
  'data:image/jpeg;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAIAAACQkWg2AAAAMUlEQVR4nGJRFC5jwAaUQo9iFWfCKooHjGogBrCsOqyCVaLnxEvq2DCqgRgACAAA//+FNgXZnyy3gAAAAABJRU5ErkJggg=='

function handler(req, res) {
  const base64Data = IMAGE_BASE64.split(',')[1]
  const buffer = Buffer.from(base64Data, 'base64')
  res.writeHead(200, {
    'Content-Type': 'image/jpeg',
  })
  res.write(buffer)
  res.end()
}

export default handler

While reaching on browser the route, you should get an error that saying that the navigator can't display the image since it contains errors.

Second problem, which is more specific, is that we use on some routes pre-gzipped content. For example, our sitemap.xml is a pre-gzipped API route. In this case, the header Content-Encoding is already present and the response should take in account the header and be base64encoded for binary content.

I also have simplified the check for isBinaryContentType(headers["content-type"]) since parseHeaders returns a Record<string, string>.

Thanks

[conico974]

LGTM! Thanks