About a year ago, the lure of developing my very first app and the opportunity to partner with a group of extremely creative people finally drove me to take the plunge into
mobile app development. This blog entry documents the odyssey of a 25 year software development veteran into "uncharted waters".
Native versus Hybrid
Traditionally -- it seems odd to associate that word with something that has been around for little more than 5 years --
all mobile apps were
native. That is, native mobile apps are developed directly for a particular target
device family. Specifically, a native app is developed for iOS devices and written in Objective-C
or developed for Android OS and written in Java. However, there is now another option; a so-called
hybrid mobile app. Hybrid mobile apps are developed without (direct) concern for the target device. Instead, the app is written using standard web technologies -- HTML5, CSS3 and Javascript -- and "wrapped" in a lightweight, native WebView. By using such technologies, we are promised portability.
As you might expect, the primary trade-off to consider when selecting between the 2 approaches is
improved performance (native) versus
greater compatibility (hybrid).
Mostly because I use these standard web technologies on an almost daily basis, and I have not touched any variation of the C language in over 10 years, plus the fact that I did not anticipate any particular need for blazing performance, I chose to travel the hybrid mobile app path.
Tool Selection
The incredible churn in the range, capabilities and even branding of hybrid mobile app tools was intimidating, making my second decision much trickier than the first. I considered many of the tools recommended by a variety of technical websites including:
- app.js
- jQuery Mobile
- Kendo UI
- Sencha Touch
- Ionic
- Intel XDK
- Titanium Mobile
- Knockout
- Montage
For me, my first gate was
cost. If I perceived any possibility of an expense in implementing a tool, it was eliminated from the shortlist. Out went Kendo UI. Secondly, I looked at
server/hosting complexity. I also expected to use a pre-existing domain/host that offered no support for Angular or node.js. Goodbye app.js and Ionic. Thirdly, given my experience with them, I wanted to ensure that the tool offered
tight integration with jQuery and Bootstrap. Finally, I looked at
intangibles. Based upon some negative comments and my own lukewarm experience using it as part of a webapp project, I eliminated jQuery Mobile.
In the end, I think what really sold me on
Intel XDK was the breadth of functionality provided right out-of-the-box and its intuitive workflow. In particular, Intel XDK offered direct support for many UI frameworks, including
Bootstrap, its own
App Framework, and, later,
Framework7. It provided in-the-cloud support for backbone. It included a slick way to test and integrate web services. The tool was tightly integrated with Cordova/PhoneGap and its family of plugins. It offered multiple emulation and testing "modes". It interfaced with each of these and more through an IDE that also offered a WYSIWYG builder. I found the latter to be particularly useful during the first few months, trying to figure out the
art of the possible.
Understanding Intel XDK's Project Options
Now that I had selected Intel XDK, I was then confronted with the selection of project options within the tool itself. Like similar frameworks and IDEs, Intel XDK organizes solutions into projects (i.e., each of my hybrid mobile apps would have a corresponding project). At project creation, Intel XDK presented me with a series of options that boiled down to this:
- Do I want to use Cordova plugins in my app?
- Do I want to build my app using the App Designer WYSIWYG builder?
- Do I want to base my app on an existing sample or demo?
- Which of the supported UI frameworks do I want to use in my app?
After a little experimentation, I arrived at the following answers:
- Yes; unless you are building a truly trivial (or perhaps totally innovative app) there is almost certainly a Cordova plugin that you can leverage
- Yes; in future, however, I might explore building an app without App Designer as its use can make for some tricky debugging and/or unnecessary code
- No; although it was useful to independently review the code behind some of the samples and demos
- App Framework; although I am already exploring the Bootstrap3 and Framework7 options
Apple Developer iOS Certificates, Identifiers and Profiles
There is also considerable set up required simply to allow your app be installed on iOS devices. The dependencies are a little confusing but Intel XDK provides a guide, under the BUILD tab, that helps to clarify. You need an Apple Developer ID with which you create
certificates, define your apps, register devices (on which you want to test), and then generate a
profile (typically, one per app). This is all accomplished through the
Apple Developer website. You download the resulting profile file and install it into the Intel XDK project repository. You then update the Intel XDK PROJECTS tab Build Settings to match the Apple Developer managed values (for App ID, App Name, Provisioning File, etc).
What Went Well
As previously mentioned, Intel XDK's focus on a
soup-to-nuts solution really helped me to stay above water on a project that was my introduction to many terms, concepts and processes. It really seemed intuitive and logical, to me at least, which is particularly impressive given the relative immaturity of
app development.
I was also impressed with Intel XDK's efforts to provide
responsive components out-of-the-box that I could easily test emulating multiple form factors.
I already had an inkling of the App Store Submission/Apple Developer process's labyrinthine nature. Without Intel XDK's related guide, I might still be thrashing around trying to understand those dependencies.
The ability to easily deploy my app builds
on device for "beta testing", before the actual App Store submission process, was a godsend. I don't have any point-of-comparison, but I feel it unlikely that any other tool could make this more straightforward.
Not So Much
Even with Intel XDK's impressive breadth of functionality, I still needed to supplement my project development with:
- Issue tracking (I used Github)
- Authentication and authorization (I used Auth0)
- Backend tools (I used PHP and MySQL, for which I needed separate IDEs)
Lack of built-in source code control was, perhaps, the most disconcerting aspect of working with the Intel XDK. I did see a couple of general references to integration with tools (Github again, I think), but I never explored how practical that is. The need for this was underscored by somewhat frequent crashing of the tool itself, although it is fair to say that the frequency of such crashes diminished as subsequent upgrades to the Intel XDK were made available (currently, I have been through about 5 such upgrades).
There is an almost dizzying array of variations within the Intel XDK, some that seemingly overlap (for example, the options under the BUILD tab). Some of that seems to be the result of Intel's purchase and re-branding of what was once referred to as AppMobi. This general sense of confusion around the plethora of decisions and options is further compounded by the lack of a single unifying and coherent set of documentation.
Next Time
For my next round of mobile app development, I will certainly consider the Intel XDK again. Intel seems to be providing the resources to maintain and update its features. This is absolutely critical given the ever-growing list of competitors and the rapid churn in the associated technologies. I will also take a second look at Ionic and React.js. You can be sure that I will also see what other tools have emerged in mobile app development. I fully expect there to be a bunch more.
