Can JavaFX do CAD?

[John C. Turnbull]

Hi Richard,

I have to stop posting late at night, that one came across as really ANGRY!

It's not anger, it's passion... and frustration.

I am frustrated because I spend much of my day trying to convince my
employer that we should be using JavaFX.  They ask me questions like:

"What happens if Oracle abandons JavaFX just like Sun did with JMF, Java3D,
JOGL etc. ?"

I say:

"This is Oracle, not Sun."

They say:

"Can you show me what JavaFX can do? There must be examples out there
right?"

And I say:

"Sure, here's Ensemble."

They say:

"OK, so it has a nice set of basic controls and can do simple animations but
what about more complex things like Flash?"

...hence the dancing cat reference.

It's not that my employer *needs* dancing cats, it's just that they need to
see that there is more to JavaFX than red circle transitions.  I can't even
prove to them that JavaFX is capable of dancing cats.  They don't have the
resources to fund me to develop something more sophisticated but they tell
me that if JavaFX truly was a "mature" technology (like I tell them) then
where are all the examples?

I am finding it difficult to convince them that JavaFX is production ready
and is not still in "experimental" mode because I am unable to demonstrate
its true capabilities or refer them to many examples of people (and I mean
big companies) actually using it.

The main concerns of my employer and I think many companies in a similar
situation is that JavaFX won't survive long term and that it is only really
suitable for form based applications.  Then of course there is the whole
"HTML5 runs on all platforms" argument but that's another story...

So this is why I think it's imperative that Oracle invests in developing a
true showcase application for JavaFX.  Something that non-technical people
(like managers who make decisions about where the money goes) can look at it
and go "wow!".

I am just not getting my managers to go "wow" at what I can show them with
JavaFX at the moment.

Every comment or apparent criticism I post about JavaFX is from the
perspective that I am trying to deal with real-world problems and people who
require proof (such as demos, reference sites etc.) and not because I myself
think JavaFX is not up to scratch.

It's quite the opposite actually.

I am a very, very strong believer and supporter of JavaFX and have many
reasons both personal and professional as to why I want it to be a massive
success.  As I have said before, there are plenty of people who praise
JavaFX and tend to avoid the very real issues that are restricting its
adoption.  I just think we have to face these issues head on if we are to
compete in what is a very cut-throat industry.

-jct

-----Original Message-----
From: Richard Bair [mailto:richard.bair at oracle.com] 
Sent: Saturday, 27 July 2013 01:40
To: John C. Turnbull
Cc: 'Daniel Zwolenski'; openjfx-dev at openjdk.java.net
Subject: Re: Can JavaFX do CAD?

> For Flash, there are literally millions of examples of 
> fancy/complex/impressive graphics and animations out there that can be 
> really impressive at times.  I have not seen ONE such example in JavaFX!

Point to one?

Have you seen any of the JavaOne examples? The movie wall or movies on a
stack of 3D cubes was pretty good. But I guess you're not interested in the
3D aspect? What is it you are looking for exactly? Different people (on this
list) have had different perceptions on both (a) what's important and (b)
what kind of graphics they're interested in. Most people would deride the
dancing cat as being totally irrelevant to the types of applications they're
trying to build (the basis for much of flash animations is shape morphing,
you can find some code here https://gist.github.com/gontard/5029764).

On the other hand, JavaFX is not a replacement for OpenGL. Drawing 25
million lines is just not something we can do right now, especially in a
resource constrained environment. I've already commented on the memory
overhead (which would continue to be an issue even if the drawing part of
the problem were solved).

I've pushed to graphics repo the StretchyGrid, which is about 300k line
nodes (the actual amount is variable, see the javadoc comments). At 300k
nodes the scene graph overhead is negligible on the FX side, dirty opts is
taking a long time to run, and painting is really slow.

PULSE: 347 [122ms:222ms]
T12 (8 +0ms): CSS Pass
T12 (8 +0ms): Layout Pass
T12 (47 +53ms): Waiting for previous rendering
T12 (100 +1ms): Copy state to render graph
T10 (101 +16ms): Dirty Opts Computed
T10 (117 +105ms): Painted
Counters:
	Nodes rendered: 306565
	Nodes visited during render: 306565

If I were doing this by hand in open GL, I think the drawing would be
essentially free, if I used LINES with GL anti-aliasing, I could send 'em
all down to the card in a single shot (and if I had a modern GL I could do
LINES + FXAA or one of the other per-pixel AA algorithms available and it
would turn out pretty nice). Because our shapes don't implement the non-AA
path, and our AA involves software rasterization and uploading of pixels, I
expect that to be the main source of the 105ms time being spent here.

Also I noticed (by turning on prism.showdirty=true) that the entire grid is
being painted every time, even though visually it looks like only a small
subset actually needs to be changed. But that's really a minor thing, as I
said, drawing this many lines should basically be free if I configure
"smooth" to false in the app. Except that right now it is totally not
implemented (in NGShape):

    public void setAntialiased(boolean aa) {
        // We don't support aliased shapes at this time
    }

The point of stretchy grid is not to say "wow look at this amazing demo".
The point is to say "what happens if I put in 300K nodes. Where does the
system start to fall over?". 

Richard=