TeaVM

[shannah]

I'm curious what the rationale was for choosing GWT over TeaVM. Seems like GWT will always face the glass ceiling of not supporting threads or third-party jar dependencies (that haven't been built for GWT), since GWT works on the source code. TeaVM supports threads, works on bytecode, which seems much more flexible.

[salmonb]

hi Steve, I made that choice at the start of the WebFX project back in 2015 or 2016, so the situation was quite different. According to the tests I conducted at that time, I got smaller and faster JS code with GWT than TeaVM, which was my main criteria. TeaVM was faster to compile, but this was not crucial for me, as JS compilation is not required that frequently for final users developing WebFX apps (OpenJFX can be used instead during the main development cycle). Also considering that Google was behind GWT, compared to a single developer behind TeaVM at that time, the balance went in favour of GWT in the end. I don't know if I made the right decision, it was just the prototyping phase, and I had no idea how far the project will go.

But I'm glad that TeaVM grew up since that time, and as you can see in our roadmap, we plan to support other transpilers including TeaVM. This is not our priority however for now, this may happen in a year or so. Is the TeaVM team interested in helping us providing that support?

[reportmill]

I'm a huge fan of TeaVM, but I would suggest CheerpJ as well. CheerpJ doesn't require the additional compilation step - developers can use the same jars for desktop and browser. And I could easily compile to WebFX in the browser with SnapCode.

I think CheerpJ is free for Open-source uses, though it's commercial for commercial uses. And they are limited to Java 8 for the next few months (hopefully they will support Java 17 by year end).