If you boil down a UI element's description to a set of lines, colours and gradients, you can easily apply that to other elements at all scale factors, without having to ship piles of images (possibly in four different sizes, if you have iPad-specific assets) and deal with the horrible "stretchable images" system.
That aside, the ease of prototyping makes development that much faster.
Stretchable images are quite elegant: they are relatively easy to use, use less memory and memory bandwidth, perform all drawing on the GPU when paired with UIImageView, animate size changes effortlessly and automatically share their buffer between all image views that use a single resource. You can certainly add all of this to elements that are drawn "in code", but it's considerably more work.
Additionally, paintcode has the flexibility of making say, a button that shrinks/grows based on the shape of another view, say, as is caused when you rotate an ipad, etc.
This product just outputs Obj-C, which must be baked into your source code, right? That's what I think of as hardcoding. I wouldn't call loaded PNG's from outside of the binary "hardcoded".
I can certainly see where having easy programmatic access to the primitives that make up a vector drawing would be useful though.
I'm waiting for someone to tell me what's horribly wrong with this before I click the "Buy" button, but it's taking effort to resist the urge. This looks extremely excellent.
The money is irrelevant. $70 is chump change for any application I care about enough to do custom drawing in. But if there's something unworkable about the approach, I sure wouldn't want to waste my time.
(I have a couple Mac Cocoa apps I'm working on that I kinda sidelined because UI controls were painful).
1) You need to design the selected/deselected states for buttons separately and merge the code.
2) It'll give you drawing code, but you have to maintain those chunks of code directly in your source files. Any modifications you make to the code need to be reflected in the original PaintCode data, and vice-versa. This could be solved with clever comment guards and the ability to rebuild internal structure from the code it generated.
Neither of these really detract from the product. I've only used the demo, but it's honestly pretty amazing.
Core Graphics drawing is actually quite slow. On the Retina iPad, drawing is 4 times as slow and uses 4 times as much memory as the 1st gen iPad, and all the CPU cores in the world are not going to help you.
Having recently ripped out most of my -drawRect: methods, replacing them with custom rendered UIImages, I can attest to the significant performance gains that UIImage gives you. On the Retina iPad, I think it's pretty much mandatory.
With that in mind, I think most people are better off using a tool like this to produce PDF data rather than raw CG calls, and use generic code to render that PDF data to a UIImage. PDF data produced by Quartz is going to render exactly the same as raw CG calls.
It makes for easier project management, and much better performance.
Of course, if you don't need resolution independent graphics, just use png.
Its not that Core Graphics is slow, just that CG doesn't guarantee that drawing will take place on the GPU. It turns out that CG typically prefers the CPU for its its drawing. In comparison, when you load images into a UIImageView, you know that the drawing will be done on the GPU, instead.
Quick question: How does UIImage's -drawInRect: and -drawAtPoint: differ?[1]
With that in mind, I think most people are better off using a tool like this to produce PDF data rather than raw CG calls, and use generic code to render that PDF data to a UIImage
The CGPDFPage* functions that Apple provides are document-oriented (remember: the D in PDF stands for "document"), and don't necessarily work well in all cases, either. A compromise could be to render your CG calls into a context that is then turned into a UIImage that gets cached and stuck into a UIImageView for display. This way, you can do your drawing in CG*, get GPU-backed rendering, and don't have to deal with clumsy PDFs to get images.
So, my advice? Click the "buy" button. Drawing in code is just another tool to have. No reason to avoid doing so, if the situation calls for it -- and the trick to any tool is knowing when to use it.
1. If you guessed that -drawInRect is CPU-backed and drawAtPoint: is GPU-backed, pat yourself on the back. Simply having your resources in a UIImage isn't enough to guarantee GPU-backed drawing. And drawing on the GPU isn't enough to guarantee that it will be fast.
drawAtPoint: and drawInRect: both call through to drawInRect:blendMode:alpha: which calls through to CGContextDrawImage. CGContextDrawImage has some optimizations to handle unscaled blends and further optimizations for unscaled blits, but it doesn't use the GPU on iOS. Even if it did, it would be using the GPU to draw into a temporary buffer that would have to be drawn again when the QuartzCore scene is rendered.
I've used it, and had no problems at all. It's missing a few features before I can whole-heartedly recommended it over Opacity[1], which is what I use for the same purpose right now. Right now I'd need to make separate on/off state, for example, for each button. Opacity's factories & variables make that much easier.
That said, Opacity's development has slowed to a crawl. If the gentleperson(s) behind PaintCode keep it up, they will surely surpass it very soon.
IANAL, but does this violate the part of Apple’s iPhone Developer License Agreement that says “applications must be originally written in Objective-C, C, C++, or JavaScript”? The code is “written” in Objective-C, but it was “originally” not code at all.
I don't think that's part of the developer license agreement any more. Considering how successful Unity, Corona and other non-obj-c development tools have been I'd be surprised if it was.
Don’t know. The last I heard of it was mid-2010, and there doesn’t seem to be new information saying it’s changed. Apple seems bent on killing Flash, and that wording was a perfect way to deal with CS5. Unity, on the other hand, has a good relationship with Apple, so there’s no reason for Apple to invoke Dread Section 3.3.1 against them.
Seriously? That's like saying if you thought of the code, wrote it down on paper or started off as UML diagram that it would be invalid. I hope you were kidding.
I'm interested in pretty much the exact opposite of this product.
I've worked on a few apps where I need to pre-render some high-quality images and include them in the app bundle. In some cases they are relatively difficult to hand-draw in an image editor (at least the sort of image editors I can afford) and I've had to write a Mac app to draw the image. Basically the sort of thing that a CAD program would be good at.
So I'd love to have something that I can script in a relatively high-level language that renders really sharp antialiased graphics at 2x resolution with control over things like end caps.
25 comments
[ 5.7 ms ] story [ 77.3 ms ] threadThat aside, the ease of prototyping makes development that much faster.
So you're hardcoding either way.
Additionally, paintcode has the flexibility of making say, a button that shrinks/grows based on the shape of another view, say, as is caused when you rotate an ipad, etc.
I've been very happy with the product so far.
I can certainly see where having easy programmatic access to the primitives that make up a vector drawing would be useful though.
(I have a couple Mac Cocoa apps I'm working on that I kinda sidelined because UI controls were painful).
1) You need to design the selected/deselected states for buttons separately and merge the code.
2) It'll give you drawing code, but you have to maintain those chunks of code directly in your source files. Any modifications you make to the code need to be reflected in the original PaintCode data, and vice-versa. This could be solved with clever comment guards and the ability to rebuild internal structure from the code it generated.
Neither of these really detract from the product. I've only used the demo, but it's honestly pretty amazing.
Having recently ripped out most of my -drawRect: methods, replacing them with custom rendered UIImages, I can attest to the significant performance gains that UIImage gives you. On the Retina iPad, I think it's pretty much mandatory.
With that in mind, I think most people are better off using a tool like this to produce PDF data rather than raw CG calls, and use generic code to render that PDF data to a UIImage. PDF data produced by Quartz is going to render exactly the same as raw CG calls.
It makes for easier project management, and much better performance.
Of course, if you don't need resolution independent graphics, just use png.
Quick question: How does UIImage's -drawInRect: and -drawAtPoint: differ?[1]
With that in mind, I think most people are better off using a tool like this to produce PDF data rather than raw CG calls, and use generic code to render that PDF data to a UIImage
The CGPDFPage* functions that Apple provides are document-oriented (remember: the D in PDF stands for "document"), and don't necessarily work well in all cases, either. A compromise could be to render your CG calls into a context that is then turned into a UIImage that gets cached and stuck into a UIImageView for display. This way, you can do your drawing in CG*, get GPU-backed rendering, and don't have to deal with clumsy PDFs to get images.
So, my advice? Click the "buy" button. Drawing in code is just another tool to have. No reason to avoid doing so, if the situation calls for it -- and the trick to any tool is knowing when to use it.
1. If you guessed that -drawInRect is CPU-backed and drawAtPoint: is GPU-backed, pat yourself on the back. Simply having your resources in a UIImage isn't enough to guarantee GPU-backed drawing. And drawing on the GPU isn't enough to guarantee that it will be fast.
That said, Opacity's development has slowed to a crawl. If the gentleperson(s) behind PaintCode keep it up, they will surely surpass it very soon.
[1]: http://likethought.com/opacity/
Just kidding: this app is not needed.
I've worked on a few apps where I need to pre-render some high-quality images and include them in the app bundle. In some cases they are relatively difficult to hand-draw in an image editor (at least the sort of image editors I can afford) and I've had to write a Mac app to draw the image. Basically the sort of thing that a CAD program would be good at.
So I'd love to have something that I can script in a relatively high-level language that renders really sharp antialiased graphics at 2x resolution with control over things like end caps.