The right idiom to do this (without referencing the reader) is to set the write buffer size to zero and then call drain(). However I think there's a bug -- if in a non-paused state you use set_write_buffer_size() to set the buffer high-water mark below the current buffer content's size, the transport should enter pause state but it doesn't.
So perhaps your wait-for-eof-on-reader strategy is better, unless we manage to fix the above thing. An ideal API might be built on top of all these. Please file bugs to track these issues. On Mon, Feb 17, 2014 at 8:28 AM, Gustavo Carneiro <[email protected]>wrote: > In tulip, if I have a tcp connection to a server > (asyncio.streams.open_connection), it is not trivial to ensure that all > data was written before closing the application. > > If I simply call close() on either the StreamWriter or Transport, it > basically does nothing. Data will continue to be written in the > background, which means that if I exit from the Python script immediately I > risk not having all data written to the socket. > > From what I can tell, I have to close() the StreamWriter, then read from > the StreamReader until I get an empty string (EOF). Am I understanding > this correctly? > > I wish this sort of thing were a bit more intuitive in a future revision > of tulip. A synchronous close() would be helpful, I think. > > -- > Gustavo J. A. M. Carneiro > Gambit Research LLC > "The universe is always one step beyond logic." -- Frank Herbert > -- --Guido van Rossum (python.org/~guido)
