iPhone App Development with Google App Engine

After almost a year of messing around with various iPhone development alternatives such as Phonegap and Titanium, I finally decided to learn Objective-C and do it all properly. I actually think those other frameworks are brilliant as they allow you to use familiar languages like Javascript to quickly create nice apps for both iPhone and Android. But since they rely heavily on the web view element for loading HTML, creating sophisticated apps like Skype would be impossible.

So I set out to create an app for the iPhone with Objective-C. My app is pretty simple. It basically pulls in RSS news, audio podcast and video podcast feeds into a UITableView list, allowing the user to read, listen and watch news stories from the Democracy Now! website.

I managed to put together the app pretty quickly but I ran into a lot of issues when I tried to parse and massage the XML data. For starters, cocoa does not have native support for regular expressions (but there are several external libraries). I wanted to clean up the content I was getting back before displaying it to the user but I soon realised something that would normally take me a few minutes in Python/PHP/Javascript would take a lot longer in Objective-C. Parsing XML using NSXMLParser was an absolute nightmare and extremely slow. I rarely work with XML these days and find JSON a much easier protocol to deal with. I even tested the app with some sample JSON data using the excellent json-framework libarary and it was much easier and faster. Alas, I only had RSS feeds to work with.

The other problem I ran into was slow HTTP requests. It would sometimes take up to 20 seconds just to load the first screen. This was due to a combination of slow connection speeds, long response times from the data provider and a slow XML parser.

The solution I came up with was to do as little as possible in the phone app as far as the data was concerned. I decided to use Google App Engine to fetch the data from the source, parse, rejig, massage and beautify in Python, then serialise and return the results in JSON to the phone app to use.

It may sound like this would increase response times even more since the phone would have to first call GAE, then GAE would need to call the data source and then all the way back to the phone. This is true, however, once the data is with GAE we have the luxury of using memcache and datastore. The RSS and podcast feeds are updated once a day so there’s no reason to request the data from the source every time the user loads the app. Because each time we have to make the HTTP call, parse the data and load it up. This is extremely slow and unnecessary. We can just make one request a day, then parse, cleanup and cache the results for the next user that requests it.

So the app only talks to GAE. GAE first checks memcache to see if we have a cached version. If we don’t, it will make the HTTP call, fetch the data, parse, serialise, cache and return results. If we do have a cached version, there’s nothing else to do but to return the data. A cron job will also run every 24 hours to make sure memcache is up to date.

If you really want a solid and reliable app, you need to think about all the edge cases also. What happens if the cache expires and the data provider’s website is down? At that exact moment a user loads the app only to get an error message saying there’s nothing to show. An unlikely scenario but not impossible. So the way I got around this issue was to store the serialised JSON output in GAE’s datastore as well. We always use the data from memcache but should memcache be empty and the data source down, we can switch over to the datastore and load yesterday’s content instead. Not ideal but better than having a broken app.

This is a bit of an overkill for such a simple app but it’s super fast and efficient and will work well for almost any app that relies on 3rd-party APIs. To be fair, it was my lack of experience with Objective-C that led me to using GAE. I feel much more comfortable in Python than Objective-C and I’m sure an experienced cocoa developer would have no problems parsing and massaging data in the app itself.

Of course there is one other edge case – Google App Engine could go down or worst, the interwebz could break. In which case, a simple error message will suffice.

You can download Democracy Now! app on iTunes