Add a heartbeat method to the py.Stream class (more below) #137
Conversation
There are a number of ways to implement this, but this is the most similar to how the rest of the library works. The best approach would be to return messages from the server along with messages generated in the library to give the user as much information as possible.
|
I'm going to refactor this a bit now to try and grab the messages returned from the server. |
|
Just curious from an infrastructure point of view - does holding a connection open take less resources than creating a new connection every few minutes? It seems like we're probably going to be limited by # of connections opened (a hard coded parameter) before we're limited by the cpu or memory usage required to open a connection. I only bring this up because I'm not sure whether we want to encourage heartbeat'ing or not. If every streamer keeps their connection open for eternity, even though they're only send one point every 10 minutes, then it seems like we'll run out of connections. |
|
In any case, this notebook will probably be helpful to you: http://nbviewer.ipython.org/github/chriddyp/chunked_requests/blob/master/stream.plot.ly.ipynb |
|
Its better to keep one open (if needed), and I am just following stream conventions of twitter. As The user will have to write it explicitly in their scripts I don't think we will get many. Also, the point is moot really, we have a few connections open at any one time currently, but I built the infrastructure to scale to hundreds of thousands of connections quite easily. |
|
@chriddyp , @bpostlethwaite , was this resolved? is this alright to merge in? |
|
+1 Sent from my iPhone
|
|
thx! |
Add a heartbeat method to the py.Stream class (more below)
There are a number of ways to implement this, but this is the most
similar to how the rest of the library works. The best approach would
be to return messages from the server along with messages generated
in the library to give the user as much information as possible.