Tuesday, March 30, 2010

Trends in iPhone app design

I spend so much time developing iPhone apps that I rarely get time to
appreciate design changes in other developers app. However due to my
recent brush with RSI I suddenly have some more time. So here is list
of my favourites.


Finanical times
I love the fade transition that they use when you change the device
orientation. I mean really love it. And from a development point of
view it has a couple of major wins. Namely that unlike the standard
rotation animation, you have a nice delay between the switching of the
views. Ideal if you want to load some Additional content or an
additional view. Granted you would have to do this quickly as you
don't want you user staring at a blank screen while you parse so e XML
or execute a network operation.

Also I assume it's just a case if Animating the properties of the ui
to change the super view opacity to get the desired effect.

Yahoo finance
Loving the charts btw. The Apple UIG generally tells devs to stay away
from displaying multiple types of content. However this app does a
nice job of making it work. This design should also scale nicely to
the iPad.

I have a fetish for finance apps not only because of my forays into
the markets, but unlike most apps they need to deliver a lot content/
information to their users.

The other neat thing about this app is that it give the user control
over the size of the widgets that are display for each security. You
can also swap the arranged ment of the items. Now I guessing this is a
hell of a lot of interface code. On top of the code they are using to
generate the graphs. But awesome ui.

Sunday, March 28, 2010

Mocking Core Location

One of my current project is a location aware iPhone App. It requires the device to "unlock" a location when the user/device is close enough (a couple of hundred meters).

One of the problems with iPhone development is that there is no way to simulate moving around with the device, or to feed it pre made coordinate data.

Anyone that has worked with the iPhone SDK and Core Location knows that CLLocationManager is the primary class that you need to interact with in order to get location updates.

What i wanted to do was to create a Mock Core Location Manager, that would allow me to create a text file full of coordinates, set a update delay, and then call my Core Location delegate methods, so i can manipulate my app as i please.

If your like me you would think that this should be as simple as creating a CLLocationManger category, and overloading some methods, namely startUpdatingLocation/Heading & stopUpdatingLocation/Heading. 

Now this works for most applications, however it doesn't work with the Iphone's native MapKit's MapView. For that we need to use a different callback then the publicly documented method.

There is a little app called Class Dump that generates header files from compiled binaries. Now thanks to the big drive by hackers to develop for jail broken iphones, you can find copies of the headers without having to generate them yourself.

MKMapView uses a class called MKLocationManager as it's CoreLocation Delegate. You notice that it adheres to the CLLocationMangerDelegate as you would suspect, but along with the public callback method 

locationManager:didUpdateToLocation:fromLocation: 

is also has a very similar method called

locationManager:didUpdateToLocation:fromLocation:usingSupportInfo: 

This is the winner. MKLocationManger will not respond to the former, but will respond to the later. I don't yet know what the usingSupportInfo parameter does. But my assumption is that it is a NSDictionary, however in the example i've posted i simply pass it nil.

So time for some code ...



This is fresh out of the oven, i plan to add a timer and exec the sendUpdate method after a set delay and read the locations from a text file. I'll post it up on git hub if i ever get it done


Update 29/03/10
Since writing this post i've put up a working version of the code on GitHub, Fork away

Monday, March 22, 2010

SVG Graphics on the iPhone

A new project came across my desk. The client want to import vector graphics, manipulate them and export vector graphics. No problem i thought.

Well the problem is that there is no native support for SVG graphics. Not one iota! So as a developer you have two choices. Building a JS interface and loading it in a UIWebView or implementing your own SVG parser, and loading the loading the objects as native paths using Core Graphics/ Quartz 2D.

Most smart programmers would go for the former method of course, however if you have complex SVG's you might have a performance problem, as illustrated by this video


Yeah, that performance sucks.



My problem is two fold, my client wants to export the data as a vector. Common sense wil tell you, that importing multiple vectors and then trying to export them again as a vector, in raster/bitmap based container isn't going to work.


So it looks like me and the SVG spec are going to get really cosy. The only blessing here is that SVG's that my client wants to use are fairly simple, so i should need to implement the entire spec.

Sunday, March 21, 2010

Non Global Singletons in Obj-C

Everyone seems to hate singletons. Personally i find them really convenient, but then i'm a lazy programmer. On my latest client app i decided to take a swing at making a hybrid singleton.

Generally speaking, Cocoa coventions (at least in all the documentation i've read), recommendation for sharing object instances across multiple controllers is to make it part of the application delegate. Personally i hate this, as it leads to a congested app delegate full of random iVars. Now it could be argued that if you find yourself in this place your doing it wrong, and in all truth you probably are. But considering that apple themselves suggest placing the CoreData ObjectContext in the app delegate, i think we're in good company.

My solution takes advantage of one of objective-c's many unique features, categories. Categories allow you to add methods to a class without modifing or subclassing. To most non cocoa programmers i just blew your mind, just wait to you find out about Swizzling!

Essentially i define a category on the class that i want to use as a singleton. In my scenario i wanted to have a single CLLocationManager In my entire app. This is because i need to access the devices location on a regular basis, and i want a global accuracy configuration ... and i just wanna try out some stuff :).

Code time


What i have done is simulated the typical method that you would expect to see for a singleton instance, but behind the scenes this method calls a property on the app delegate to get the shared instance variable.

In my eyes the positives to this approach are:
No global variable for the instance.
The Share instance is where you would expect it to me, and if required can be serialized on app exit.
Accessing the instance is as easy as [CLLocationManager sharedInstance] vs [[[[UIApplication sharedApplication] delegate] locationManager]


Lastly, if the internets tell me that this is a wacky idea i can refractor this to classic singleton, without making mass changes to my app. Thoughts and opinions.

Friday, March 12, 2010

Reverse Engineering: Youtube & h.264 for the masses

So a client asked me if he could get a youtube video in their new iPhone app. I replied it should be possbile ... Hence this post

Like any sensible mac user i signed up to the youtube beta as soon as i heard about it. My beloved 2007 macbook is starting to show it's age, and while it can handle 1080p videos without melting (it's sweats a fair bit!), flash videos truly bring it to its knees.

So first thing i did was select a random youtube video, this one happened to be a short about the upcoming indian premier league (i personally hate cricket, but anyways...).

I was immediately disheartened, it wasn't going to be as easy as simply reformatting the link ...


This format is new and exclusive to the HTML 5 beta, from my research the h.264 content delivered to the iphone has a different format, and is served from a different domain.

Domains/SubDomains
Also a quick glance shows that a different series of subdomains are used, and while i haven't tested, i'm guessing these are a series of load balance servers, so there is no guarantee that the video will be on the same server.

Signature(s)
Even more bad news the signature param appears to change on every request, suggesting that it's generated for each session. I tested this, my using the same account to watch the same video, but once in safari and once in chrome, checking the signatures each time.

Signature A
25BF3979A2386835A59AE3DA23076950B6323FB1.670A622C94E87A9192D1CD62885B2589D6EC9A1F


Signature B
B2F0121150A88F697711D19D2F1F8B18EBFC3058.9FE89BCC68B319D255A83A613C84A2040CAF3B2B


Beyond the signature nothing else seems to change. My guess is that youtube will have to open this up when they start to provide video embed tags instead or in addition to the standard object/embed mashup.
Retardation, tells me those are 2 SHA1 hashes separated by a dot, the age old question is whats the plain text?

Conclusion
I'm not kevin Mitnick (Holla at me Kevin ;) ). My minimal security knowledge tells me that the plain text includes a variable that is changing. My assumption is that it user session key, but with no know access to it. (Unless the people aka google are dumb uninformed, which is highly unlikely i assume they encrypt their session cookie).

So for now the h.264 videos are safe :(