SVG canvas size limitations - javascript

Problem
I was planning on using SVG as an infinite canvas on which I can move around using viewBox.
Turns out that when the viewbox reaches a specific sweetspot, the shapes start disappearing.
Example
Watch how the shapes always stay inside of the view box, until the viewbox passes the sweetspot:
Live reproduction
It's as if SVG has a limitation mismatch where the viewbox can keep running freely, but shapes can only render within specific bounds.
Research
Furthermore, the sweetspot is different between browsers:
On Chrome it's 33554399.
On Firefox it's 17895590.
The same happens on the other direction. Apply a minus sign to the above values and the shapes disappear on the other side of the canvas.
My best guess is that the browser implementation of SVG is internally constrained by something parallel to Number.MAX_VALUE (but is significantly lower).
However, I couldn't find any relevant documentation on the subject.
Question
Any way to fix or work around this? I'm hoping to keep my current implementation and not simulate the viewbox on my own.

Related

Canvas HIT testing: dealing with element boundaries and anti-aliasing

The setup
As part of a web vector editing tool written in Javascript, I'm implementing hit testing using a hit canvas strategy similar to that of Concrete.js.
For most of it, it works pretty well: I'm drawing my shapes twice (once on the display canvas, and once on a hit canvas).
When querying the canvas, I check the hovered pixel of the hit canvas and extract interaction information (i.e. which object id is stored there).
The problem
This works well inside shapes, but is painfully flawed at the boundaries where anti-aliasing makes the stored data invalid (it gets mixed with whatever background data was there before).
Are there good strategies for dealing with this data boundary problem?
Without disabling anti-aliasing for canvas methods, then we're bound to have some boundary regions across overlapping regions that will store merged data from multiple regions.
The simple scenario
In a binary scenario (some foreground vs the background), then this can be mitigated as we can assume the background to have no value, and any value becomes some foreground.
The real scenario
However, in a general scenario with multiple shapes overlapping each other on top of the background, is there any reasonable strategy for error detection? (or error correction, but I assume that's harder)
If I can tell that the data is invalid (i.e. it consists of perturbed data due to anti-aliasing), then I can use a different strategy for those few pixels at the boundaries. But I feel that it's impossible to tell whether the data I'm extracting is valid in the general scenario where we can have many overlapping shapes.
Of course, one solution is to NOT use a hit canvas. But I was wondering whether people had found a solution using hit canvases since they seem great for dealing with complex geometries.
Anti-aliasing
The ideal solution would be to disable anti-aliasing, which I don't think is possible for canvas methods [*].
[*] I know we can disable filtering when rendering images (e.g. that question) such as with imageSmoothingEnabled=false or rescaling with image-rendering: pixelated, but those don't solve the problem of anti-aliasing when drawing shapes / paths.

Render objects based on 1 specific object on the canvas

I'm presently implementing a few features into a vector editor based off of fabric.js and have hit a bump in the road with how best to handle this scenario:
In order to resize/scale the canvas smaller and larger via scaling grippies/controls, I've begun the task of implementing a "fake canvas" that stays in the middle of the real canvas (the real canvas is resized to fill it's DOM parent at all times. This presents several issues, the biggest being, dealing with objects on the canvas, ensuring their coordinate are always relative to my "fake canvas" (which is just a Rect with some restrictions applied, e.g. locking movement, rotation, etc).
So I'm really just looking for strategies/suggestions at how to go about doing this. For instance, I know I'd be dealing with newly added objects, moving objects and preventing them to be moved outside of the boundaries of my fake canvas. Among other considerations.
Thanks for the input and suggestions. I'm not looking for code, I'm more just looking for suggestions of how best to handle all the canvas related things, transferred to a fake canvas. I see that the Shutterstock Editor does some of this too, so I guess it doesn't seem like it's out of the realm of possibilities.

SVG zoom with svg-pan-zoom.js performance issues in Mozilla Firefox

I am using javascript svg-pan-zoom.js (https://github.com/ariutta/svg-pan-zoom) libary to zoom and pan svg in web application. Zooming in Firefox is very slow and laggy, while zooming in Chrome and IE11 works very well (tested with 5MB .svg file that presents floor plan - if file is smaller, this issue is not that noticeable). Panning is working fine. I've read many topics on this issue on forum but I haven't find any solution yet. Does anybody know what is causing this problem and how to fix it?
Example: http://jsfiddle.net/coz3fd0L/3/
Check your refreshRate option. Maybe you set a high number.
If not then you may set a low number (ex. 10 which means max 10 frames per second) and if may improve your problem.
Other than that I don't know of any other problems in svg-pan-zoom. At least if pan is smooth then zoom should be the same.
Maybe your SVG has a lot of edges/curves/nodes and Firefox is simply bad at resizing such things. Or it is bad at resizing large SVGs when matrix transform is used (matrix transform is used for zoom/pan in svg-pan-zoom).
Update: From what I see this is purely a Firefox problem (or the way it is). Just opening the SVG from your example http://imgh.us/test_51.svg takes 100% of CPU (for page scroll).
Also I did try to change matrix transform values manually (to test if it is svg-pan-zoom issue) and it is anyway very slow.
The only solution I see is to try to optimize your SVGs (maybe it is possible to make them simpler: less edges, nodes, do not render white elements...).

How much slower is drawing to a transparent canvas and checking if point is transparent than checking if point is in a complex polygon?

This is more of a generic question to be honest, just wondering if anyone has done any sort of research on the subject.
Basically I am adding event support to a small game engine I am creating for my own personal use. I would like pixel perfect hover over 2d object event support and am just thinking of the best way of doing it. Realistically it would be faster for me personally just to invoke a draw of my objects onto a transparent canvas and checking if the mouse x y is over a transparent pixel or not since I dont have to make a set of points defining the outside of an object. This would also allow me to have holes in my object and it would still correctly know if I hovered over or not.
What I am wondering is using methods shown here: How can I determine whether a 2D Point is within a Polygon?
How much slower would my method be to the methods shown there?
Im currently still learning so its not easy for me to implement all of this and just test it myself since it would probably take me ages to get to work correctly and test the speeds.
Side note: I would still have a basic bounding box to save it from redrawing and testing every single time.
Checking if a point is in a polygon will 99.999999% of the time be vastly, vastly faster.
To be slower the polygon would need to be extremely complex.
To do the other method you need to use getImageData, and getting image data on the canvas is very slow.
Point in polygon algorithms do properly account for holes. Make sure you have one that obeys the non-zero winding number rule, because that is what canvas uses (as opposed to the even odd rule) and you may want compatibility with paths constructed in the canvas (either now or later).

Infinite Zoom with Javascript, Jquery or HTML5 Canvas

I've seen this "The Scale of the Universe 2" and I just want to know if this can be done with javascript or jQuery or with HTML5 Canvas.
If you click an item (example the "Human") , an info will pop out beside it.
I searched for 3 days here if someone has a similar question. But I only saw Google Map like behavior where you can zoom in on the map cursor position.
Actually I want to make a "Timeline" like effect or, like the "Time Machine" Recovery on Mac OS X.
Fixed position of zoom. Not like a google map zoom, that you can pan and zoom anywhere.
Can I put (example "The human") images and text on a div?
Are there available articles/tutorials about this?
Options:
Javascript
jQuery
HTML5 Canvas and CSS3 Transform and scrolling it to Z-axis so you can zoom in/out.
Flash/Flex (Well I don't want to use lots of resources on CPU because I need it in a large resolution or in full screen.
It is possible to implement an infinite zoom in HTML canvas, this is the source code of a proof of concept that I implemented and here you can test it live.
It is actually quite tricky to implement, and most probably you'll need to use big decimals.
The approach I followed uses the same coordinate space as d3-zoom. Basically, you have three coordinates, x, y and k. K is the zoom. If k is 2 it means that you have doubled everything.
Now, if you want to do infinite zoom, it is very easy to reach k = 64, but that is already outside of float64 precision and you'll have a lot of artifacts, where translating in the image is not smooth, or you don't zoom in where you want.
One way to avoid those artifacts is to use coordinates that have infinite length, for example BigIntegers. To make my implementation easy and compatible with d3-zoom, I used big decimals instead, but I had to implement my own library of BigDecimals, basically infinite precission on the integer part and 32bits of precision on the decimal part. Of course, you also need to adapt the zooming library to support BigDecimals. Moreover, in the case of d3-zoom, a lot of calculations where done in the initial coordinate space (k=1) but division of floats will always have an error and it is also perceivable once you are deep enough. To avoid that you need to do all computations locally.
It might sound like a lot of hassle to insist on using the d3-zoom library, but zooming UX is actually tricky, specially if you combine that at different k, you'll need to consider scrolling, zooming on the phone, double tapping...
In case you want to use SVG transformations, then you'll need to fake it. You'll introduce nodes when they are barely visible, allow to scale them. However, most probably you'll also need to fake it when they are too big to avoid artifacts there.
There is no infinite zoom. However you can zoom in/out of an SVG image in HTML5 canvas.
SVG supports affine tranformation. You can set the required zoom/pan in the affine transform and show the relavant areas. The behavior/listener can be implemented in Javascript and the SVG can be rendered on HTML5 canvas.
As a starting point you can look at this example: http://www.html5canvastutorials.com/labs/html5-canvas-scaling-a-drawing-with-plus-and-minus-buttons/
This is totally doable in HTML5. Actually, any system able to display and zoom images should be able to. It's not one big image being zoomed, it's a big amount of images being zoomed (for instance the initial human is an image, which is scaled and moved out when you zoom in or out). The idea is splendid, but I don't really see any technical performance in it. As long as you correctly limit the number of images being resized and bitmapped, it should keep a decent FPS rate.

Categories