RSDK-654 - refactor message handling for streaming#184
Conversation
|
Ethan (@stuqdog) - I haven't dug into the code, but I have pulled this PR locally and confirmed that in my repro environment, I no longer see the arm wait until all the points have been sent on the stream. So this looks promising! |
Andrew C. Morrow (acmorrow)
left a comment
Andrew C. Morrow (acmorrow)
left a comment
There was a problem hiding this comment.
I took a pass through this, but it has been over a year since I wrote any production rust. You might want Nicolas Menard (@npmenard) to also take a look, especially since he did the BiDi work in the MicroRDK and knows the async rust gRPC interaction quite well.
|
|
||
| /// Signals the end of the client's send stream (the gRPC CloseSend equivalent). | ||
| // note(ethan): the variable that used to carry this signal was named | ||
| // `it_was_all_a_stream` — the best variable name I've ever shipped into production. |
There was a problem hiding this comment.
I love that this is a memorial to a memorial.
There was a problem hiding this comment.
I just couldn't let it go haha
There was a problem hiding this comment.
I understand completely! Long ago, in a closed source system, I managed to slip in the following haiku, which I now have an opportunity to bring back to life:
Exception rising
On the signal handler stack
Portending sorrow
3 participants
Refactors message handling to send messages asynchronously as soon as they come in, defer sending EOS until client messages are demonstrably completed, and only as its own frame.