Screenshot-19

With help from @Frenzie and @paul_irish, the latest not-yet-versioned release of Weppy, my Javascript WebP to WebM conversion library, or something of a polyfill for a format that is yet to be part of any specification (HTML5 seems to specifically reference the image src attribute are examples such as PNG, GIF, JPEG, APNG, PDF, XML, SVG, SMIL, and MNG). The new release brings some awesome new features that really don’t change much and shouldn’t really be used in the real world because most browsers in the world still aren’t Firefox, Chrome or Opera - that is, a large portion of the browser market includes browsers like Safari and IE, either suffering from antiquity (IE6! aah!) or just liking h264 (IE9 + Safari).

The new release supports Opera. I never bothered debugging Opera, I figured it was another huge issue that would demand a rewrite (as supporting Firefox had needed, because the order of the object keys isn’t preserved and breaks the EBML result, or at least for Firefox’s parser which seems to be somewhat stricter than Chrome’s, is that ffmpeg?). And after premature optimization (stripping “unnecessary” EBML tags), my code didn’t work in chrome, so I had to revert to an earlier revision. All my testing code was based on file drag-drop stuff, and Opera doesn’t support that. Until I saw this mozillazine topic, I didn’t care, but it was a lot easier to fix than I feared.

Part of the solution was getting rid of the canvas stage. Admittedly, the canvas stage was pretty useless once the toDataURL() stage was removed before the first public release, but I didn’t feel like deleting code, so it stayed there. Also, I noticed that the global variable that gets introduced was accidentally named “WebM”, which is wrong, it should be “WebP”, but because of the uncreative format naming and similarities, I didn’t notice. Not sure, but it seems to be more stable now.

Chrome probably will add WebP soon, and it needs to be future proof, detecting whether or not a browser supports the WebP format. To do that, it creates an Image, sets the src to a data url of a 4x4 webp image and listens to the onload and onerror events, checking if the size is correct and there were no errors loading it. The routine is expected to error and totally untested as there aren’t any browsers that support the feature yet for me to try.

Another change, is that by default, it will automatically load all the same-origin (because of the limitations of XHR) webp images (from <img> tags), on the DOMContentLoaded event, so the library is practically drop-in now. In any web page, you can pretty much add <script src=”http://antimatter15.github.com/weppy/weppy.js“></script> and on the supported browsers, it should automatically load and replace all WebP images, though not something I would really recommended.

The demo is the same place it always was: http://antimatter15.github.com/weppy/demo.html

There is also this nifty hack that uses <canvas> to add an alpha channel to the WebP image (adapted from the original JPEG one): http://antimatter15.github.com/weppy/alpha/alpha.html

Also, please follow me on twitter.

Screenshot-14

WebP http://code.google.com/speed/webp/

Recently, Google announced a new lossy image compression codec, named WebP, intended to supersede JPEG. It is based on VP8’s intraframe compression algorithm. It’s not natively supported in any browsers yet. Weppy is a compatibility layer that changes WebP files into WebM files that can be loaded on several modern browsers.

How does it work?

WebP is actually a lightweight container for a single VP8 frame (whereas WebM is a container based off Matroska meant for video). WebM support exists already in Chrome, Firefox and Opera, so all that’s needed to render it is to do a little magic to convert the RIFF encoded WebP image into a EBML/Matroska encoded single frame WebM video, loading it in a , using drawImage() on a and replacing the .webp image with the data URL extracted from the using toDataURL().  Issues

• Chrome (on linux anyway) tends to crash a lot.
• Firefox throws a security exception when doing toDataURL() on a canvas after drawImage() of a video loaded from a data url. The hack being used it to replace the image node with the actual canvas instance.
• Opera doesn’t work. I don’t really have the time to investigate. What Browsers?

Chrome 7.0 and Firefox 4.0 were both tested. Opera doesn’t work for reasons that I’m not sure about. I would appreciate it if someone fixes it and submits a patch :)

Demo

http://antimatter15.github.com/weppy/demo.html

DragDropAwesome

I’m making a (really awesome) chrome extension that involves dragging and dropping files. MDC, as usual, has great information on the topic. Gmail does it almost perfectly, with the green file drop target. But all other implementations of this feature suffer from two issues, and Gmail’s code is far too obfuscated for any mere mortal to interpret (no doubt thanks to closure). When implementing the file drop target, two fairly important user experience issues occur. I don’t know what way Gmail does it, and I just used whatever solution worked.

Firstly, is that you only want to trigger the drop target once a file is being dragged onto the web app. You don’t want a file drop target to appear once someone starts dragging text (fairly simple, check that e.dataTransfer.types includes File). Even trickier (even Gmail doesn’t get this right) is dragging links also triggers the file drop target for some reason (it’s tricky because then e.dataTransfer.types also includes the File type for some odd reason). There’s no way to access file data as it’s being dragged (getData always returns undefined). One thing that I still have no idea about is inter-browser-window data transfers.

Secondly, is how to get the dialog to dismiss itself once the dragged image is dragged out of the viewport. It’s pretty tricky because of how the events bubble and don’t always go through <body>. The solution I eventually arrived at was to add a timer on the dragleave and see if within fifty milliseconds (a random amount of time, zero works too, but fifty feels safer), another drag event is fired.

Hopefully, people will find this useful :)