search

DioxusLabs / taffy

A high performance rust-powered UI layout library
https://docs.rs/taffy
Other
2.04k stars 102 forks source link

WIP: Java bindings #705

Open adjabaev opened 3 weeks ago

adjabaev commented 3 weeks ago

Objective

Make taffy usable in Java programs

Feedback wanted

Feel free to give me advice over what I am doing as it is the first time I am actually writing Rust/ jni bindings

TODO

adjabaev commented 3 weeks ago

Do you think it is a good idea keeping Taffy in name of every class (I am asking the question because I feel like java doesn't have as much concise ways to express certain ideas which leads to some code looking quite long) (not even talking about the fact that in Java the types of variables are usually explicitly written). Also I'd be very curious knowing what do you think so far

nicoburns commented 3 weeks ago

@adjabaev very excited to see this. I'm on vacation (with limited access to computers / internet) atm. But I will definitely review this when I get back.

One thing that I have noticed from writing C and WASM Bindings (neither finished/merged yet), and that I also see here, is that there is a lot of repetitive code related to defining enums, converting enums, fields on Style, methods on TaffyTree, etc that is duplicated in each binding and in the main Rust implementation. And I have been thinking it might be good from a maintenance perspective to use code generation to automatically generate this code from a metadata definition similar to Yoga's enums.py script (this would be pretty helpful when we add new style fields for example).

Regarding naming conventions, I'm not sure. Just Java have namespaces? C doesn't, which is the main reason why it has everything prefixed.

adjabaev commented 3 weeks ago

@adjabaev very excited to see this. I'm on vacation (with limited access to computers / internet) atm. But I will definitely review this when I get back.

Glad you like the initiative! :) I hope you will enjoy the rest of your well deserved vacations (I actually think so, it is crazy the amount of things that have changed since the last time I have seen it!)

One thing that I have noticed from writing C and WASM Bindings (neither finished/merged yet), and that I also see here, is that there is a lot of repetitive code related to defining enums, converting enums, fields on Style, methods on TaffyTree, etc that is duplicated in each binding and in the main Rust implementation. And I have been thinking it might be good from a maintenance perspective to use code generation to automatically generate this code from a metadata definition similar to Yoga's enums.py script (this would be pretty helpful when we add new style fields for example).

It is definitely something I have noticed and I would be more than glad to find a way to automate this stuff, for enum related stuff, I feel like its not gonna be very difficult (thanks to [Java's Enum.ordinal()](https://docs.oracle.com/en/java/javase/21/docs/api/java.base/java/lang/Enum.html#ordinal()). But in the same time I am wondering if we couln't just Serialise/ Deserialise stuff as Json strings. At least I feel like it would lower the implied work by a lot, however I cannot tell what it is going to be performance wise.

Regarding naming conventions, I'm not sure. Just Java have namespaces? C doesn't, which is the main reason why it has everything prefixed.

Yes, Java (and Kotlin) have packages which are basically namespaces, so I removed Taffy from every class (every file still start with "package com.dioxuslabs.taffy;" (except for TaffyTree because it is its original name)

adjabaev commented 2 weeks ago

Hi :) I did some changes to how enums work, these are now generated automatically. This is probably something that could go into a meta PR that would help all bindings, but I thought it would be nice to showcase the idea here before flooding with another PR. What do you guys think?

Also for now, it only generates classic enums without any parameter because in Java there is not such notion as enums that have different parameters (signatures)

Also another question, the MeasureFunction/ NodeContext stuff seems very hard to translate to Java dynamically as NodeContext can by anything, if anyone has an idea, i'm all ears! :)

adjabaev commented 1 week ago

Also clippy doesn't validate but it looks like it isn't due to my code (and it wasn't the case before i wrote it, I tried looking for the reasons but i do not manage to understand it and would like to avoid modifying Taffy code if it isn't necessary)