A Primer on Hybrid Apps for iOS | Hacker News Reader

A Primer on Hybrid Apps for iOS

cocoacontrols.com

The problem with people complaining that hybrid iOS apps don't "feel right" is that the defining characteristic of a well-made hybrid app is that you don't notice it's a hybrid app, meaning the only apps people tend to consciously associate with being 'hybrid apps' are the mediocre ones.

For every web-based iOS app that's clunky and slow like Facebook, there's another one like Quora or OKCupid or even the iOS App Store that is almost indistinguishable from a native app, save for a few minor telltale giveaways (one of the most common ones is showing a loading icon while you load in an entire view instead of rendering a view's components piecemeal as they're ready).

The problem is that many developers who go with a hybrid app solution do so because they think it'll be easier or quicker than an all-native app, especially if they need multiplatform support. Doing a hybrid app correctly can in many cases require much more work than simply doing a native app, both technically (there are a lot of very tricky caching issues, since UIWebViews don't properly handle cache manifests) and design-wise in terms of nailing the behavior and feel of a native app. Hybrid apps are neither a complete usability disaster nor a silver bullet that gives you a 'write once, run anywhere' utopia.

I don't know about Quora and OKCupid, but the iOS App Store is a terrible app ! Scrolling sometimes hides the whole table view, the sign out button are not responsive, and don't even try to open past purchases if you have a lot.

I agree with you on the last part though. I'd recommend native app with a good design that makes it flexible via server changes. For big changes, the app store updates is usually just a one week process anyway

For those of us who code in web-land, a delay of one hour would be appalling, let alone an entire week. (And of course even then, it relies on user action, and so will never be 100%.)

One hour is appalling? Are your users that dependent on running the latest version of your app, or is it a concern about fixing show-stopping bugs in reasonable time?

I'm sure he just means that on the web, the turnaround from bug to bugfix to patched instance is near instantaneous.

Step 1, Notice bug

Step 2, Fix bug

Step 3, Deploy bugfix

The steps between 1 and 2 could obviously take a variable amount of time, but Step 3 is 'near instantaneous' on the web. The thought of having to relinquish deployment to somebody else, and that somebody else possibly taking a week to just alert users that there IS a bug fix, much less the time it takes for those users to see and apply the fix, that's a relative eternity.

True, but surely you can retain the ability to alert or otherwise communicate these things to your users without putting any of the main GUI in a web view? Minor bugs require only a little communication and major show-stoppers can be avoided with care.

Maybe I'm a little biased on this, but I hate that a web app I might be relying on can change completely (both in GUI and functionality) without my consent or involvement; at least with app updates I can get some warning and release notes -- a chance to extract my data before taking the plunge... but I guess I'm getting a bit off-topic.

As an 'app owner', it perhaps benefits you to possess the ability to preserve what you've paid for. In some cases, I'm sure this actually is a benefit, as in the times updates are negative, or take away functionality or change the way they do things. In most cases though, it just means that you aren't up to date with what is current, which may or may not be a bad thing.

It is offtopic, but still interesting. One supposes that is the reason why apps cost a set amount of money (usually) while web apps are usually a subscription model. That is of course questionable when the app uses web views to your server, as those customers always cost you money so long as they're using the app vs. standalone apps (like games and the like) where you truly are making a one-time purchase.

Strictly speaking, none of us "own" the majority of the software we pay for. We license it, and we are entirely at the mercy at the vendor's terms, with the only recourse being not buying it in the first place.

It's yet another area where the nature of software and the networked world breaks our object-oriented intuitions in ways we're still trying to resolve in the social contract.

I get the implication, and while that wasn't the point I was making, the purchase of a self-contained app from the app store is essentially the same thing as owning the app.

You've exchanged money for a good, and as the app owner there aren't any readily available means of getting the good back from the owner should you disapprove of how they're using it, just like it would be with a tangible good. Obviously copyright still stays in effect so their ability to copy, resell, yadda yadda are still restricted, but otherwise it's theirs.

That said, in this day and age, fewer and fewer apps are completely self-contained, and/or work without being plugged in to something, so it's effectively a moot point.

Weigh that against a critical bug appearing and having your app be useless for a week while awaiting an already-completed update.

At the same time, I understand your feelings about wanting to be in control of updates. (As a user, I'm even upset that Apple doesn't allow the user to roll back.) But when devs can count on every user being up to date, it enables them to push new features more quickly instead of having to write backwards compatibility, fallbacks, etc.

The question of who owns and controls code is a sticky one, and I'm not sure there's an ideal answer; to the extent that there is, I think it varies highly by the nature of the app.

> or even the iOS App Store that is almost indistinguishable from a native app

Are we using the same app store? It's awful! It's painfully slow and it frequently 'forgets' what I was doing when I press the back button after viewing an app's details (search filters and list positions, to name a few).

I've been trying to get a UIWebView to use a cache manifest with no luck. Is it hopeless or is there a trick/writeup I'm missing?

I haven't used the Quora app, but the OKCupid app is pretty bad - it frequently gets itself into a bad state (especially if the app hasn't been opened in a while) where it will sit at the loading screen forever. It's also painfully slow, navigation is incredibly laggy. It's functional, but hardly great.

It's not the most obvious UIWebView-wrapped app ever, but it's still fairly evident.

I have only ever seen one hybrid app done well - and that's the LinkedIn app. Every single other hybrid app (including Apple's own App Store!) is slow, buggy, and just unpleasant to use.

I can usually spot a hybrid app, but I hadn't even noticed that OKC is one, and my experience with the app has been pretty good. I have seen the stuck loading screen occasionally, but I assumed it was due to connection issues; it's not like that can't happen with native.

Either way, they obviously need to fail a little more gracefully, which should be a solvable problem.

How would you know you can "usually" spot a hybrid app?

Subtle things: controls that don't match exactly, link box outlines, inertial scrolling with a different acceleration, keyboard limitations, or just the general feel. It takes a strong attention to detail to make these things feel native, and the OKC app apparently does this very well.

I think the point the previous chap was making was that you've got no way of knowing if you 'usually' spot hybrid apps as you don't know about the ones you don't spot.

Point taken, classic selection bias.

In addition to everything that's been mentioned about the App Store, it also has the most telling mark of a hybrid app: It occasionally crashes on the iPad 1 on iOS5 because it runs out of memory, especially if there is still something running in the background. Even Safari itself crashed regularly until I realized that Skype had been running in the background for days.

IMHO, hybrid apps would be much nicer if UIWebView could just manage its appetite for RAM a little better.

I'm of the view that in general, it's better to code directly to the iOS APIs rather than trying to implement everything in a UIWebView. Having said that, I've just spent a substantial amount of time developing a word processing application (UX Write), which uses a UIWebView for all document layout, and a large amount of "backend" code in javascript for performing editing operations. All the UI code however is written in Objective C and the Cocoa Touch APIs.

The advantage of using native controls it it provides a much greater integration with the rest of the operating system, and makes it easier to make your app work in the same manner as other apps. For example, there are many features of iOS that are not exposed through javascript APIs, and for that you have no option but to use native code. Nonetheless, I like what Microsoft is doing with providing javascript APIs for metro development, and though I haven't tried it myself, it looks like a good approach.

There are a number of serious problems with the current UIWebView which are also holding it back as development tool:

- You have very limited control over touch events. In my case, I completely disabled all default event handling on the web view, and implemented my own touch event handlers in Objective C, which pass the events through to javascript.

- Selection is broken. You can access the selection, but not modify it. I spent several weeks completely reimplementing the selection & cursor mechanism because of this. One positive outcome though was it gives me much more fine-grained control over user interaction issues.

- contentEditable is also broken. I spent probably 3-4 months reimplementing this functionality myself in javascript to make it usable

- Interfacing between Objective C and Javascript is tricky. Going from ObjC -> JS is fast, but going in the other direction using the technique from the article is painfully slow. There a ways to work around this but it's inelegant and doesn't allow return values from callback functions.

- Lack of Nitro support, as discussed elsewhere. It turns out that for my particular needs performance is quite acceptable, but for more compute-intensive apps like games it's a must. I can understand Apple's desire to prevent W+X permission on memory, but this could be solved by running JS code in a separate process and communicating with the host process via shared memory.

Apple have done some amazing things with WebKit, but the iOS port leaves a lot to be desired. I would really like to see them improve these apps so people can better integrate web content & backend functionality. But I still think you should be using Cocoa Touch for your UI.

I Think the only good reason to do a hybrid app is crossplatform portability, if you don't have the resources to do multiple native apps.

There is also the advantage of instant deployment, and always being up-to-date. I don't think hybrid is the best fit for all apps, or even most, but it can work well and create a good experience for certains types of apps.

But as other commenters have noted, it should never be seen as a "time-saver" for anything beyond trivial functionality; the little quirks of each browser runtime must be paid their due, with the user experience and performance in mind.

This the best article or post on Hybrid apps I have seen in a long while.

My one complaint is that he doesn't seem to acknowledge the possibility of a "minimal Objective-C" app (i.e. mostly HTML-based) that is tailored to iOS and uses in-app storage for most of the UI elements. We need more discussion on methods to make the mostly-HTML approach to work better. The post is a great start.

Whenever hybrid apps comes up, I never hear mention of Instagram. I can't find any specific details on their app stack, but as this screenshot (http://i.imgur.com/p00E2.png) shows, they are definitely using web views for parts of their UI (timeline view).

The timeline/feed is not a WebView from what I can see. The news part/tab is though (which is displayed in the screenshot).

Why do they link to a javascript benchmark that only features differences between ios safari versions? Javascript had terribly performance in ios 4 - if they did basically anything it would be much faster. Android's chrome has a much faster javascript engine.

Hit me up with a link to a better benchmark and I'll happily include it.

Another example of a hybrid app is the iOS (and Mac) App Store. A significant amount of the views are basically HTML.

By far, the biggest improvement one can do is to get rid of the 300ms or so delay on a UIWebView

Incidentally, the 300ms delay is not to prevent accidental touches, it's so the system can detect the difference between a double tap and a single tap. If you don't make a second tap within 300ms, it goes and fires the single tap event. But it can't do that until it knows for certain that it's not simply the first tap in a double-tap sequence.

The solution to this, if you're not using double-taps, is to respond to the event when it enters the initial phase (then the finger is pressed), rather than when it's released. I'm not sure if you can do this through javascript, but I know it's possible in Objective C by either overriding – touchesBegan:withEvent:, or implementing your own gesture recogniser.

What you need to do is use a 'touchUpInside' event, there are a couple wwdc preso's that cover it. Basically you compare coords of a touchend event to the coords of the target. Touchend events fire instantly so you dont have to worry about the 300ms click delay. Dug this out of some old code, wont work out of the box but you get the gist. https://gist.github.com/3067704

Reply on news.ycombinator.com