If the pull.values were a connection, the error that the connection was dropped should be propagated through the pull pipeline, you are not seeing this happening?

I see what you are saying about not being able to push data twice - it cant return the remainder and return an error (until the next read is requested).

But imagine the case where there is no error, just the flow ends.

pull(
  pullFile('./data'),
  reader
)

function readNextBytes(cb){
  reader.read(1024, cb)
}

Here we attempt to read chunks of 1kb of data at a time, but we lose the remainder when the file ends. Here are some ideas:

non-standard abort + remainder

// when `end` is `true`, any remaining data is passed on in `data`
// if reading from a 1025 byte file, reader would return:
// (null, 1024 byte)
// (true, 1 byte)
reader.read(1024, function(end, data){
  if (end instanceof Error) return handleError(end)
  handleData(data)
  if (end) allDone()
}

cache abort (if not an error)

// when abort signal is received, next read returns remaining data, following read returns abort
// error still takes priority over remainder
// if reading from a 1025 byte file, reader would return:
// (null, 1024 byte)
// (null, 1 byte)
// (true, null)
reader.read(1024, function(end, data){
  if (end instanceof Error) return handleError(end)
  if (end) allDone()
  if (data.length < 1024) return handleError( new Error('App Specific Error - Corrupted save fail') )
  handleData(data)
}

overreading #5

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions