Make Timer(f, ...) tasks match the stickiness of the parent task. Add spawn kwarg.
#56745
Conversation
|
Why is this a new option? Isn't spawning strictly better? |
|
In case there are use cases where code relies on the behavior. Maybe not needed? |
Sadly needed. |
Co-authored-by: Valentin Churavy <vchuravy@users.noreply.github.com>
IanButterworth
changed the title
why is it needed? |
|
We should also look for uses of Timer and change them. I also think that this should change the default. Having spooky action at a distance be why someones code is working is not correct. The fact that we sticky is not documented anywhere. |
|
What about spawn if the parent task isn't sticky already? |
|
I'm proposing merging this as opt-in, but once the discussion above is settled, maybe it can still be made opt-out. |
IanButterworth
added
the
merge me
I like this idea. |
IanButterworth
removed
the
merge me
IanButterworth
changed the title
Timer(f, ...) tasks match the stickiness of the parent task. Add spawn kwarg.
Now updated to this. |
Currently
Timer(f, 1)will create a sticky task and have the same effect as@asyncin making the parent task also sticky.I'm not sure if this should be treated as a bugfix (i.e. backport) or a minor behavior change (i.e. introduce in 1.12).
Or just an opt-in new feature, so
spawn=falseby default.Now matches parent task stickiness by default.