Automatic virtualisation (was Developing controls based on Canvas?)
Jonathan Giles
jonathan.giles at oracle.com
Sat Aug 10 15:37:15 PDT 2013
I think it would be feasible to develop some form of automatic
virtualisation API, but I don't think it needs to be an API that ships
with JavaFX. A generic implementation would probably need to abstract
out three details:
1) The 'cell factory' (in the lingo of existing controls) to generate
nodes on demand (in other words, a node factory)
2) A means of specifying when a node should be 'removed' from the
scenegraph (and added back to the pile of available nodes that can be
reused)
3) The actual layout of all nodes that are currently part of the
visualisation
I don't think there is any escaping detail #2 above, unless you want to
do a very simplistic 'if (nodeX < 0 || nodeY < 0 || nodeX > w || nodeY >
h)'. Perhaps that would be fine in some simplistic cases, but if you're
working in a zooming case I imagine you might want to prune the
scenegraph in situations where nodes are technically within bounds but
are no longer entirely relevant to the view.
This is essentially all that happens in the VirtualFlow code used in a
number of UI controls. Could this VirtualFlow code be rewritten to be
based on top of a more abstract virtualisation engine? Sure....but it
probably won't be :-)
Hope that helps.
-- Jonathan
On 9/08/2013 9:38 a.m., Felix Bembrick wrote:
> With all this talk of node virtualisation, I am wondering how feasible
> it would to build some kind of "automatic virtualiser" such that you
> pass it your real-world model and then it automagically works out the
> actual nodes and pools required and manages them seamlessly to
> maximise rendering performance and responsiveness? That is, it
> determines how many nodes will be visible at any one time and just
> manages all that stuff for you completely under the hood.
>
> Maybe Jonanthan you have come some way to building such a beast with
> your work on bidirectional virtualisation in FX 8 TableView or other
> controls?
>
> Felix
>
>
> On 9 August 2013 07:31, Felix Bembrick <felix.bembrick at gmail.com
> <mailto:felix.bembrick at gmail.com>> wrote:
>
> T
> hanks Jonathan.
>
> I'll have to check out the virtualisation that you refer to that's
> going on in JavaFX8 with TableView., it sounds very interesting.
>
> I am not saying that controls such as what I am proposing are
> *impossible* to implement using a scenegraph; it just seems
> natural to implement them in a more procedural/programmatic way
> but this could very possibly just be because I do not have a full
> grasp of the capabilities of the scenegraph and what can be
> achieved by manipulating it. To me it seems that the costs of
> repeatedly creating nodes (or hopefully reusing them), potentially
> restructuring the hierarchy of nodes in the scenegraph and
> adjusting each node's properties to enable a truly dynamic control
> would logically be much greater than just cycling through a render
> loop and rendering each graphic element in its correct location
> but, again, you seem very confident that these costs are not
> significant enough to prevent such a control performing well.
>
> In the end if I can get away with a scenegraph representation of
> this control and that massaging said scenegraph is performant
> enough then I am in a happy place. I have already observed that
> some nice stuff can happen just by playing around the scenegraph
> so it is quite possible that I am worrying about performance and
> memory issues that will never actually eventuate in practice.
>
> Felix
>
>
> On 9 August 2013 07:16, Jonathan Giles <jonathan.giles at oracle.com
> <mailto:jonathan.giles at oracle.com>> wrote:
>
> Felix,
>
> As you are restricted in what you can say, that also restricts
> how I can help. However, your example of a spreadsheet not
> being a control that can be implemented in a scenegraph-based
> manner suggests that you might want to re-evaluate your
> assumptions. A spreadsheet would be able to be implemented in
> a scenegraph approach quite nicely - the TableView control
> comes quite close to that already (especially in JavaFX 8.0
> where it can virtualise in both vertical and horizontal
> directions). In fact, I know of at least two spreadsheet
> implementations that already exist in JavaFX based on a
> scenegraph approach.
>
> As I've said elsewhere, the issues you are facing are because
> of merging two approaches (retained and immediate modes) into
> one control. What you are hoping to do is have your cake and
> eat it too, which unfortunately I don't think will end up too
> well for you :-) My advice is to recheck your assumptions -
> the scenegraph approach, in my opinion, should suit your
> requirements far more than an immediate mode approach (given
> that your requirement is to reuse existing scenegraph-based
> controls).
>
> I'm not sure why you think controls must be defined in a
> static, structured way. You could put all nodes of a control
> into a single Pane (or Region, etc) layout and layout all the
> nodes using absolute positioning. You can be as dynamic as you
> want to be.
>
> Unfortunately, without more information I am really unable to
> give any further advice, but I wish you good luck in whichever
> approach you decide to take.
>
> -- Jonathan
>
> On 9/08/2013 9:02 a.m., Felix Bembrick wrote:
>> Hi Jonathan,
>>
>> Thanks for your reply.
>>
>> I am a little restricted in exactly what I can reveal about
>> my plans for this control but I can say that it is one in
>> which its very appearance could change quite significantly in
>> a dynamic way at runtime. The control also needs to support
>> panning and zooming in a very performant way. There will
>> likely be a lot of graphics being draw in a manner which (to
>> me at least) fits the immediate mode of rendering better than
>> a scenegraph because their positions and attributes are most
>> readily defined programmatically rather than declaratively.
>>
>> Though not directly analogous, consider your typical
>> spreadsheet application like Excel where the user is able to
>> pan to the right effectively without limits and that grid
>> lines are constantly being rendered as the panning takes
>> place. Given that screens can be very large these days it is
>> conceivable that a complex spreadsheet will be displaying
>> hundreds of lines to define the cells in a grid at any one
>> time and that it is way more concise to programmatically
>> define how this grid is rendered rather than having a
>> scenegraph containing a node for each line etc. Also, the
>> panning and zooming responses are much simpler to implement
>> programmatically than continually fiddling around with the
>> scenegraph. Then there's the whole issue of virtualisation
>> and keeping the actual number of "active" nodes to a minimum
>> if it were to be done using a scenegraph.
>>
>> I am just not convinced that the costs of memory usage and
>> processing cycles to maintain a retained mode representation
>> of a visual structure like this can be justified or made
>> performant when all you really need is a simple algorithm
>> that draws lines on the screen according to the properties of
>> your cell model. Then consider that the *actual* control
>> will be significantly more complex and graphically rich than
>> this simplistic spreadsheet analogy.
>>
>> To me this is the sort of control that really lends itself to
>> an immediate mode rendering component such as Canvas but
>> there just seem to be so many impediments in the way of
>> actually building a professional control with Canvas at its
>> base. If we continue with the spreadsheet analogy then
>> obviously cells would contain other controls etc. and we have
>> discussed the limitations on "embedding" other controls
>> within a Canvas that seems to suggest it's not practical.
>>
>> I think in the end it gets down to the nature of the control
>> itself and whether its appearance is best defined by data or
>> by an algorithm just like the way certain types of knowledge
>> are best defined by nouns whereas others are best defined
>> procedurally.
>>
>> At this stage it is looking very much as though JavaFX really
>> only supports controls most appropriately defined in a
>> static, structured way like a scenegraph. This in turns
>> limits its applicability to a whole range of software
>> applications.
>>
>> Felix
>>
>>
>> On 6 August 2013 10:10, Jonathan Giles
>> <jonathan.giles at oracle.com
>> <mailto:jonathan.giles at oracle.com>> wrote:
>>
>> I think it would pay to take a step back and understand
>> why you think a 'traditional' scenegraph-based (or
>> retained mode) control is not sufficient for your needs?
>> Unfortunately you've not detailed your use case, so it is
>> hard to give any specific advice. Are you able to give
>> any details about what it is you're trying to build and
>> why you think the normal approach to building controls is
>> not sufficient?
>>
>> We've built some fairly complex controls using this
>> approach, and if implemented wisely, there is very little
>> that a scenegraph-based approach can't do. Specifically,
>> do you think your control will render all of the
>> 'thousands of nodes' at once, or will many of these nodes
>> be off screen or otherwise not visible at any one time?
>> For things like the TableView we only render the nodes
>> that are visible. This means that regardless of whether
>> there are 100 or 1,000,000 rows of data, we only have
>> visual nodes for the 20 visible rows, for example.
>> Keeping your scenegraph as minimal as possible is always
>> a very wise idea, if performance is a concern.
>>
>> As you note, the other problem is that you will run into
>> issues if you want to mix canvas rendering with the
>> scenegraph-based controls like Button. The best you're
>> likely to achieve (having not tried it personally) is to
>> position the control on top of the canvas, rather than
>> attempting to render the control inside the canvas (and
>> having to then deal with event handling, etc). This will
>> likely prove to be finicky, and more cumbersome than
>> simply using an entirely canvas-based or entirely
>> scenegraph-based approach.
>>
>> -- Jonathan
>>
>>
>> On 5/08/2013 10:11 p.m., Felix Bembrick wrote:
>>
>> I am investigating the feasibility of developing a
>> JavaFX 8 control based
>> on Canvas. I have chosen Canvas as the base class as
>> this control is of a
>> very dynamic nature and would not be easy to
>> implement with a retained mode
>> style ancestor (at least as far as I can tell).
>>
>> So is this feasible? While I can readily see how to
>> render the visual
>> aspects of the control, I am not sure how to best
>> "embed" other controls
>> within it should that become necessary (and almost
>> certainly will).
>>
>> For example, how would I go about embedding a Button
>> within my control? It
>> looks to me like I would need to create an actual
>> Button node somewhere
>> else in the scenegraph and then perhaps render it
>> within my control using
>> gc.drawImage() passing in a snapshot of the Button
>> node. That's OK but
>> then I have to somehow handle events and I am not
>> sure how best to do that.
>>
>> Another issue I see is that there seems to be no way
>> to apply effects to
>> individual graphic elements within the Canvas as the
>> applyEffect() method
>> applies to the entire Canvas.
>>
>> Finally, a significant obstacle is this issue:
>>
>> https://javafx-jira.kenai.com/browse/RT-23822
>>
>> This issue relates to the lack of support for LCD
>> font smoothing within
>> Canvas. This may not sound that serious but the
>> difference between LCD
>> font-smoothed text in other controls and the
>> grey-scale text in Canvas is
>> so distinct on my current machine that a control
>> based on Canvas would
>> really stick out like a sore thumb and appear
>> significantly less appealing
>> than a "standard" control.
>>
>> So, am I wasting my time?
>> Are there any other issues I am likely to face?
>> Are there other ways to develop dynamic controls
>> which may involve
>> thousands of nodes (such as lines, curves etc.)?
>>
>> Thanks,
>>
>> Felix
>>
>>
>>
>
>
>
More information about the openjfx-dev
mailing list