search

HenrikBengtsson / future.apply

:rocket: R package: future.apply - Apply Function to Elements in Parallel using Futures
https://future.apply.futureverse.org
208 stars 16 forks source link

Error setallcol(x): verbose must be TRUE or FALSE` #103

Closed tle4336 closed 2 years ago

tle4336 commented 2 years ago

I am currently in need of using the future.chunk.size parameter in future_mapply() function with the goal of speeding up the execution time of a big optimizer function, where there are 3 vectorized arguments with a bunch of other arguments added within the MoreArgs = ... . However, I kept bumping into the error setallcol(x): verbose must be TRUE or FALSE when setting the parameter future.chunk.size = 3 (the lengths of my 3 vectorized arguments are 52). Is this a bug? I used future.apply (version 1.9.0) and future (version 1.25.0).

Also, when I run my code, it seems to me that every time it hits the library(future.apply) row, it starts encountering a very long buffering time where it seems to plan/compute everything before pushing out outputs at a breakneck speed. Is this a normal behavior? The buffering time takes way too long though.

HenrikBengtsson commented 2 years ago

Hi. Yes, a known problem with the latest future 1.25.0, that will be fixed in the next release. Please see https://github.com/HenrikBengtsson/future/discussions/621#discussioncomment-2715653 for details and how to workarounds for now.

HenrikBengtsson commented 2 years ago

Also, when I run my code, it seems to me that every time it hits the library(future.apply) row, it starts encountering a very long buffering time where it seems to plan/compute everything before pushing out outputs at a breakneck speed. Is this a normal behavior? The buffering time takes way too long though.

> system.time(library(future.apply))
Loading required package: future
   user  system elapsed 
  0.681   0.024   0.705

If that's what you see, it's probably something with your setup. Please post to https://github.com/HenrikBengtsson/future/discussions/, since I doubt it's a bug in the future framework.