Re: [Jmol-users] Jmol/SwingJS and JSmol update

[Robert H. <ha...@st...>]

I've did a review of CheerpJ and TVM for a grant proposal. I don't know
about JWebAssembly. That sounds plausible. I realize that CheerpJ has a
nice website and *appears* to have a following, but I have done over 500
applets now, and I am 100% certain that they could not have been
successfully done well with CheerpJ. For one thing, it is a bit-image
creator. That is, it uses exactly the graphics that Java does, turning
everything into a set of pixels. (Yes, I know that that is what a browser
does, but a browser does that particularly well, and to simply emulate the
Java canvas in JavaScript has all sorts of downsides.) But, also, some
aspects of Java cannot be directly converted to JavaScript -- 64 bit
integers, for Example. I guess that could be done in WASM, but that would
be a huge performance hit unless there is a sustained process that would
need those. I wonder how they do 32-bit floating-point math. JavaScript is
64-bit.

<quote>

JSweet[8] uses the Java language as a naive proxy for TypeScript. It has no
place in the expansion of
established Java projects as concurrently developed JavaScript equivalents.

TeaVM[9] is a prototype compiler for turning specialized Java bytecode into
JavaScript. It is extremely
limited, does not work with OpenJDK classes, and has no practical use in
relation to this project.

CheerpJ[10] is a commercial product that interprets Java bytecode and
emulates a subset of Java
programs from a Java jar file delivered to a web page. There was an
unsuccessful attempt in 2016 to use
this technology for relatively simple PhET[11] applets, but the technology
proved fundamentally
unscalable, unoptimizable, and generally unusable.

</quote>

If you can find a good description of the limitations of CheerpJ, I would
love to see that. There must be one somewhere.

Bob