I am a web developer and I want to move my web products to iPhone. One of the products is like Google Maps: show map on the phone screen, you can drag or resize the map and view some information that we add to the map.
Are there other, similar products? What are the differences between them? Which should I choose?
I registered with stackoverflow just for the purpose of commenting on the mostly voted answer on top. The bad thing is stackoverflow does not allow new members to post comments. So I have to make this comment more look like an answer.
So, if your concern for your app is to make it more "native" looking, Titanium is a better choice. If you want to be able to "port" your app to another platform more easily, PhoneGap will be better.
Updated 8/13/2010: Link to a Titanium employee's answer to Mickey's question.
Updated 12/04/2010: I decided to give this post an annual review to keep its information current. Many things have changes in a year that made some of the information in the initial post outdated.
The biggest change came from Titanium. Earlier this year, Appcelerator released Titanium 1.0, which departed drastically from its previous versions from the architectural standpoint. In 1.0, the UIWebView control is no longer in use. Instead, you call Titanium APIs for any UI functions. This change means a couple things:
Your app UI becomes completely native. There is no more web UI in your app since the native Titanium APIs take over control of all your UI needs. Titanium deserves a lot of credit by pioneering on the "Cross-Platform Native UI" frontier. It gives programmers who prefer the look and feel of native UI but dislike the official programming language an alternative.
You won't be able to use HTML or CSS in your app, as the web view is gone. (Note: you can still create web view in Titanium. But there are few Titanium features that you can take advantage of in the web view.)Titanium Q&A: What happened to HTML & CSS?
Move on to PhoneGap. There are not many new things to say about PhoneGap. My perception is that PhoneGap development was not very active until IBM jumped on board later this year. Some people even argued that IBM is contributing more code to PhoneGap than Nitobi is. That being true or not, it is good to know that PhoneGap is being active developed.
Web page rendering moving from relying on CPU to using GPU acceleration. Graphic intensive tasks such as page transition and 3D animation become a lot smoother with the help of hardware acceleration. GPU Accelerated Compositing in Chrome
Such improvements that are originated from desktop browsers are being delivered to mobile browsers quickly. In fact, since iOS 3.2 and Android 2.0, the mobile web view control has become much more performing and HTML5 friendly. The future of mobile web is so promising that it has attracted a big kid to town: JQuery has recently announced its mobile web framework. With JQuery Mobile providing UI gadgets, and PhoneGap providing phone features, they two combined creates a perfect mobile web platform in my opinion.
I should also mention Sencha Touch as another mobile web UI gadget framework. Sencha Touch version 1.0 was recently released under a dual licensing model that includes GPLv3. Sencha Touch works well with PhoneGap just as JQuery Mobile does.
If you are a GWT programmer(like me), you may want to check out GWT Mobile, an open source project for creating mobile web apps with GWT. It includes a PhoneGap GWT wrapper that enables the use of PhoneGap in GWT.
From what I've gathered, here are some differences between the two:
Titanium source gets compiled down to native bits. That is, your html/js/etc. aren't simply attached to a project and then hosted inside a web browser control - they're turned into native apps. That means, for example, that your app's interface will be composed of native UI components. There are ways of getting native look-and-feel without having a native app, but... well... what a nightmare that usually turns out to be.
Titanium apps become native apps - they're just developed using web dev tech.
What does this actually mean?
A Titanium app will look like a "real" app because, ultimately, it is a "real" app.
A PhoneGap app will look like a web app being hosted in a browser control because, ultimately, it is a web app being hosted in a browser control.
Which is right for you?
If you want to write native apps using web dev skills, Titanium is your best bet.
You might be asking: Why would I want to write a PhoneGapp (I've decided to use the name) rather than a web app that's hosted on the web? Can't I still access some native device features that way, but also have the convenience of true web deployment rather than forcing the user to download my "native" app and install it?
The answer is: Because you can submit your PhoneGapp to the App Store and charge for it. You also get that launcher icon, which makes it harder for the user to forget about your app (I'm far more likely to forget about a bookmark than an app icon).
You could certainly charge for access to your web-hosted web app, but how many people are really going to go through the process to do that? With the App Store, I pick an app, tap the "Buy" button, enter a password, and I'm done. It installs. Seconds later, I'm using it. If I had to use someone else's one-off mobile web transaction interface, which likely means having to tap out my name, address, phone number, CC number, and other things I don't want to tap out, I almost certainly wouldn't go through with it. Also, I trust Apple - I'm confident Steve Jobs isn't going to log my info and then charge a bunch of naughty magazine subscriptions to my CC for kicks.
Anyway, except for the fact that web dev tech is involved, PhoneGap and Titanium are very different - to the point of being only superficially comparable.
I hate web apps, by the by, and if you read iTunes App Store reviews, users are pretty good at spotting them. I won't name any names, but I have a couple "apps" on my phone that look and run like garbage, and it's because they're web apps that are hosted inside UIWebView instances. If I wanted to use a web app, I'd open Safari and, you know, navigate to one. I bought an iPhone because I want things that are iPhone-y. I have no problem using, say, a snazzy Google web app inside Safari, but I'd feel cheated if Google just snuck a bookmark onto Springboard by presenting a web app as a native one.
Have to go now. My girlfriend has that could-you-please-stop-using-that-computer-for-three-seconds look on her face.
I'm taking a course in Android/iPhone development and we spent 8 weeks with Titanium (not full time) (Version was Titanium 1.4.2 and time was around November 2010). Here is my experience.
iPhone Android dual targetting
Even though the API guides claim that the functionality is available for both the Android and iPhone, this is not the case. Much of the stuff simply don't work on one of the platforms. Some things works differently.
A lot of the people in the class has done iPhone applications, and they can not make them work on Android without major rewrites. I developed a simple childrens app called Animap (see android market / Appstore in Sweden) and started developing under Windows. Once the Android target was working I opened the project on OS X. It does not show any build stuff for iPhone, just for Android. You need to start a dual target project under OS X. (Ok, I copied the relevant files to a new project). Next problem - the animations does not work on iPhone (they work on Android). The scrolling events does not work the same on the iPhone. (i.e on Android you get the untouch event when user stops scrolling and releases their finger from the screen, this does not happen on the iPhone).
Since this is not mentioned somewhere you basicly need to do trial and error programming on first one platform, then on the other platform. By trial and error I mean it will take about two days to get such a simple App as Animap working on the other platform. You will also need to have if (android) then... or if(iphone)... all over your code...
Download and setup
You must follow the instructions to the letter. Do not try to use java 64 bit. It will not compile the KitchenSink 1.4.0 demo application. (1.3 works OK!) You must put files directly on the C drive as long pathnames will make the external program not receiving all command line parameters if they get to long. (Fine for small programs though) 1/3 of the times, the toolchain simply stops and you must press 'launch' again. Then it will probably work... very unreliable. The simulator will not be found on startup and then you must simply kill of adb.exe with Ctrl+Alt+Delete and retry.
On a wifi-network you sometimes looses the live connection and Titanium crashes on you (the compile/deploy interface) If you do not have a working internet connection it will not start as it can not log you in to their servers.
CSS, HTML and jQuery is a breeze compared to this. Titanium resembles any other old GUI API, and you need to set some properties for every single button/field/etc. Getting a field wrong is just to easy, remembering all the properties that needs to be set? Did you spell it with capital letters at the right place? (as this is not caught by the compiler, but will be seen as a runtime error if you are lucky to test that part)
In Titanium things simply break when you add another view on top of a control or click somewhere else in the GUI.
Several API pages carry the Android symbol, but will only return a null when you try to create the control. They are not simply available on the Android platform despite the symbols. Sometimes Android is mention to not support a particular method, but then the whole API is missing.
The demo application. Did I mention it does not compile if you put it in your Eclipse project folder because the path gets too long? Must be put on your C drive in the root folder. I currently use a symbolik link (mklink /J ...)
You must propably use things as label.setText('Hello World') to change a label reliable but this is not documented at all.
Titanium.API.info('Printouts are the only way to debug');
The APIs are not available in any good format so you can not get ordinary code-completion with help etc. in Eclipse. Aptana please help out!
It seems that the compiler/tools are not multithreaded so a fast computer with a fast harddrive is a must, as you must do a lot of trial & error. Did I mention the poor documentation? You must try out everything there as you can't trust it!
Some positive things
- Open Source
From previous projects I have promised myself never ever to use closed source again as you can't simply fix things just by throwing hours and manpower at it. Important when you are late in the project and need to deliver for a hard deadline. This is open source and I have been able to see why the tool chain breaks and actually fix it as well.
It's also open. You can simply see that your not alone and do a workaround instead of another 4 hours spent on trial&error.
- Seems to be active on their forums.
- Titanium 1.4 is not threadsafe. That means if you make use of threads (use the url: property in a createWindow call) and program like the threads are working and send events with data back and forth you run into a lot of very, very strange stuff - lost handlers, lost windows, too many events, too few events, etc. etc. This is all dependent on the timing, putting the rows of code in different order might crash or heal your application. Adding a window in another file.js breaks your app.js execution... This also trashes internal datastructures in Titanium, as they sometimes can update internal datastructures in paralell, overwriting a just changed value with something else.
Much of the problems I have had with Titanium comes from my background on realtime systems like OSE who support hundreds of threads, events and message passing. This is supposed to work in Titanium 1.4 but it simply doesn't do it reliably.
Then we have more simple bugs in Titanium, like some parameters not working in the functions (which is quite common on the Android platform at least).
Trial and Error debug cycle speed Having run Titnium Developer on several computers, I noticed that the bottleneck is the harddrive. An SSD drive on a laptop makes the build cycle about 3-5 times faster than on a 4200 rpm drive. On a desktop, having dual drives in RAID 1 (striping mode) makes the build about 25 percent faster than on a single drive with a somewhat faster CPU and it also beats the SSD drive laptop.
- From the comments in this thread there seems to be a fight for the number of platforms a tool like this can deliver app's for. The number of API seems to be the key selling-point.
This shines through very much when you start using it. If you look at the open bugtracker you see that the number of bugs keeps increasing faster than the number of fixed bugs. This is usually a sign that the developers keep adding more functionality, rather than concentrating on getting the number of bugs down.
As a consultant trying to deliver rather simple apps to multiplatforms for a customer - I'm not sure this is actually faster than doing native app development on two platforms. This is due to the fact that when you are up to speed you are fast with Titanium, but then suddenly you look down and find yourself in a hole so deep you don't know how many hours must be spent for a workaround. You can simply NOT promise a certain functionality for a certain deadline/time/cost.
[edit - added part with bugs and not being thread safe] [Edit - now having worked with it for a month+, mostly on PC but some on OS X as well. Added iPhone and Android dual targetting. Added Trial and Error debug cycle speed.]
The Corona SDK (Ansca Mobile) uses Lua as its coding language. See lua.org for more about Lua.
I have been working with Titanium for over a week now and feel like I have a good feel about its weakness.
1) If you hoping you use the same code on multiple platforms good luck! You'll see something like backgroundGradient and be amazed until you find out android version doesn't support it. Then have to revert to using a gradient image, might as well use it for both versions to make the code easier right?
2) A lot of weird behaviors, on the Titanium android sdk you need to understand what a "heavy" window is just to get the back button to work, or even better orientation event tracking. This isn't how the android platform really is, its just how Titanium tries to make their API work.
3) Your thrown in the dark, Things will crash and you have to start to comment code and then when you find it, never use it. There are certain obvious bugs, like orientation and percents on android that have been a problem for over six months.
4) Bugs .... there are a lot of bugs and they will be reported, sit around for months, get fixed in a few days. I am surprised they even are planning to release a black berry mobile sdk when there are so many other problems with android.
If you stay away from a lot of the native UI parts, i.e instead use setInterval to detect orientation changes, sticking with gradient images, forget about the back button, build your own animations, forget window header, toolbars, and dashboard. You really can make an api that works on both that doesn't require of lot of rewriting. But at that points its just as sluggish as a webapp.
So is it worth it? After all the pain, its worth every minute. You can abstract the logic and just build different UI for each rather then if elseing everywhere. Titanium lets you make fluid applications, that feel fast. You lose the powerful layout abilities of each platform but if you think simple, things can get done under a single language.
Why not a web app? On entry level market android phones its horribly slow to generate a webview and consumes a lot of memory you could be using to do more complex logic.
Here's a more recent and in depth analysis of Appcelerator and PhoneGap: http://savagelook.com/blog/portfolio/a-deeper-look-at-appcelerator-and-phonegap
native mapkit is supported in Titanium
Making HTML5 widgets tha look like iphone widgets is one thing, but making them perform equally well is another matter altogether. Performance of html5 animations (even plain View Transitions), scrolling long lists, responsiveness to gestures feel sticky and jerky. An iPhone user will notice the difference.
There's also some differences in the kinds of gestures that are supported by different devices which results in platform specific code and usability issues as well.
I'll stay with native apps for now I guess.
Rhomobile Rhodes (http://rhomobile.com/products/rhodes) is very similar in approach to PhoneGap, but is the only framework with:
- a Model View Controller pattern (as most web frameworks provide)
- an Object Relational Manager
- support for all popular smartphones (including Windows Phone 7)
- a hosted development service (not just hosted build): http://rhohub.com
- a full debugger and SDK-less emulator in the RhoStudio IDE
- support for synchronized offline data
For anybody interested in Titanium i must say that they don't have a very good documentation some classes, properties, methods are missing. But a lot is "documented" in their sample app the KitchenSink so it is not THAT bad.
UPDATE: Titanium added Maps API in version 0.8 of their framework.
Of the solutions you mentioned, none of them appear to give you direct access to the MapKit framework introduced in OS 3.0.
As the Google Maps HTML widgets aren't nearly as good as MapKit (see Google Latitude for an example), you are probably best off developing a native Cocoa touch application, or choosing a solution you can extend to add MapKit integration. PhoneGap is extensible in this manner (it's open-source so it is by default), and some of the other solutions might be as well.
edit: Titanium now has support for MapKit
I've tried corona. It was good until I discovered it doesn't support streaming mp3 audio. So, I stopped right there. I think if I really want to be an iphone app developer I should learn obj c. All I wanted to make an app which has a list of radio stations and you click on them it start playing.