multiprocessing rolling window functions

[randomgambit]

Hi there!

I was thinking: can tsibble be parallelized? There is the furrr package that could be used, but I wonder if this is compatible with tsibble. The resulting syntax might be a bit horrible...

What do you think?

[earowang]

The window functions can be parallelised, since the backend is purrr::map. It may be easier for furrr to support parallelised versions of rolling window functions, since it's already had the functionalities, for example future_map.

[randomgambit]

thanks. yes I get that the backed is map but you are using regular old purrr::map and not furrr:future_map. Are you saying that using future_map that itself calls the windows functions will work in parallel?

[earowang]

i’m suggesting that it would be more of a feature in furrr instead of tsibble.

On Fri, 19 Oct 2018 at 12:51, Olaf ***@***.***> wrote: thanks. yes I get that the backed is map but you are using regular old purrr::map and not furrr:future_map. Are you saying that using future_map that itself calls the windows functions will work in parallel? — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#66 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACV6mKaHV1pjzpWY3Zuv5Joz304anZ4vks5umTAHgaJpZM4XksEr> .

--

…

---------------------------------------------------- Earo Wang B.Comm.(Hons), Monash University http://earo.me

[randomgambit]

@earowang OK I am going to create a request in furrr and cross reference this issue. THanks!

[earowang]

All rolling window functions now have their parallel-processing counterparts, with future_ prefixed.

[randomgambit]

well done bro. i ll try that asap

[earowang]

You may want to close the issue on furrr.