Add an easy example so people can write their own schedulers. about libunifex

[ericniebler]

Yes, we need more simple examples, and a simple scheduler example would be very enlightening.

You're not far off here. One mistake you're making is that you're calling the receiver methods more than once. Exactly one receiver method (set_value, set_error, set_done) must be called exactly once. (set_done should only be called when the work has been cancelled, which isn't the case here.)

Better would be to use an algorithm repeat_effect_with_blocking_delay to repeatedly connect the sender returned from schedule multiple times with a sleep between them. You could implement it in terms of repeat_effect_until. (To get an idea, take a look at how repeat_effect is implemented in terms of repeat_effect_until and a forever predicate.)

And if you really want a scheduler with this behavior baked in, you can have your schedule() function return a result of calling repeat_effect_with_blocking_delay.

[tinkerbeast]

Thanks for your response. I am looking into the repeat_effect_until code. In the meantime I had a look at the io_epoll code. What I can gather from the epoll code is that the data is read only when async_read_some is called. This seems to be pulling data from the read_sender, rather that read_sender pushing data (as would happen in normal socket programming / callbacks). Is the generalized concept of unifex pull rather than push? The one receiver method rule you mentioned would also make more sense in the pull context. Or is there a push paradigm also?