What is the best way to clone shared types in pycrdt?

[armanckeser]

In the JS library there is the .clone() utility method that returns a fresh copy of the sharedtype that can be included in a new ydoc. I couldn't find an equivalent in pycrdt, any tips on cloning sharedtypes from one ydoc to another?

Thanks!

[davidbrochart]

I'm not sure Yrs has this functionality, @Horusiath ? Otherwise, we could implement it in Python. I guess it would have to support nested types.

[Horusiath]

yrs doesn't have such thing. Depending on the type your could call toJson (for arrays/maps) or get formatted delta chunks (for text types).

[armanckeser]

I see, good to know I wasn't missing anything. Not sure what you mean by delta chunks though, also does that apply to XML fragments? Thanks!

[Horusiath]

Actually I think that AsPrelim is what you're looking for - it basically can take any yrs object and change it into equivalent input object that can be integrated into another document. Ofc. this means that it's stripped off the metadata required for conflict resolution, so it's only for things like copying.

[armanckeser]

That's exactly our use case! What would it take to implement the bindings for this on the Python side?

[davidbrochart]

Probably not much, do you want to try and open a PR?

[armanckeser]

@davidbrochart before I open the PR, as far as I can tell, the Python implementation does not expose Prelim types and types them as Any and hides them in the sharedtype's class. In the implementation, should I return something like Array::from_prelim(prelim) from the as_prelim binding as opposed to returning the prelim itself as Any?

[davidbrochart]

Hmm right... Actually, isn't the .to_py() method what you are looking for?

[armanckeser]

We can't use .to_py() on Xml types, which is our main sharedtype

[armanckeser]

Any thoughts?

[davidbrochart]

Maybe we should implement .to_py() on Xml types?

[armanckeser]

Maybe, but Xml representations in Python are in a weird spot. The built-in element tree is slow and doesn't have many features, the de-facto standard is lxml library, but both xml and lxml will get flagged by Bandit for being insecure. Even though it shouldn't be a huge problem for this use case, since we are not parsing strings but converting nodes, the code might require some #nosemgreps if this CI is using Bandit. Apart from that, we would also need to change the __init__ for XmlFragment's to accept and transform our pythonic representation back into pycrdt.

Maybe we can return the Prelim type from to_py of XmlFragments? No added dependencies?