Skip to content


Browse files Browse the repository at this point in the history
Remove our useless atEnd handler, since it was interim to begin with.…
… Protip: "this is not correct" in a comment means it is very likely to contain a bug :)
  • Loading branch information
aredridel committed Oct 25, 2015
1 parent 700a169 commit 2cadcf1
Showing 1 changed file with 4 additions and 18 deletions.
22 changes: 4 additions & 18 deletions index.js
Expand Up @@ -68,6 +68,10 @@ module.exports = {
// If we forget to call next(), our stream will just never end
// and things may hang.
// The second function is optional, used if we need to do anything at
// the end of the stream. We don't, since we just pipe any remainder
// on blindly to our caller.

// For simplicity, we're going to _completely_ ignore that binary
// data like images are a thing that could ever exist.
Expand Down Expand Up @@ -151,24 +155,6 @@ module.exports = {

next(); // Ready for next chunk.
}, function atEnd(next) {
// through2 takes 2 functions: forEachChunk and atEnd -- this is the second one,
// it gets called when the data incoming to through2 ends. Any clean-up or final
// actions we need to take before we stop getting called happen here.

// there's still a stream 'end' event that has to happen, after this stream will
// do no more. We don't want a race between our cleanup and the caller's last read
// or the end event. They may assume they got it all if end was emitted before this,
// but we had one more thing to do, so we have to tell through2 we're not just done,
// but done-done.
// If we had no atEnd, it would just send 'end' when we're done. But we had one thing
// pending.

// So if we had leftovers, let's put them in headers. This isn't correct, but
// vanishing data in a closure variable is no good either.
// Whatever comes out of our header-splitting stream parser must be the
// body. Because that's what we designed, right?
Expand Down

0 comments on commit 2cadcf1

Please sign in to comment.