-
-
Notifications
You must be signed in to change notification settings - Fork 58
non-mutating munge_data #68
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
We used to have a version like that but the result can't be inferred so it's too costly. |
|
It needs to be done since Zygote cannot infer through the constructor with this mutation. Maybe the easiest fix for now would be to create new dispatches so that Missing does what it currently does and a non-missing dispatch that just does the no-mutating thing but fast. That will make everything with Zygote work well while not impacting Pumas speeds. |
|
It’s fine to rewrite these methods such that they work with Zygote. We just shouldn’t use DataFrames for it because of the inference issue. |
|
I believe the latest commit whould be a possible solution without using dataframes. |
ChrisRackauckas
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yeah, that should be good. I think you can get rid of the collect, but I know Iterators.filter is quite slow so maybe it's better to collect for now anyways. We can go with this and always improve if it's an issue. Let's get an issue open about this minor performance detail and move on.
|
Ok, it's passing. |
No description provided.