Sep 2013

Greg’s UITextView Work-Arounds

A few days ago Greg Pierce — who I’ve known since my days at UserLand, who writes Drafts — sent me email letting me know he’d published a gist with some UITextView text selection work-arounds.

We’re using that code in the version of Vesper that appeared on the App Store yesterday.

So far so good. It definitely helps. Many thanks to Greg.

Technical Notes on Vesper for iOS 7

(Vesper for iOS 7 is on the App Store.)

TextKit

You remember in Lord of the Rings when Gollum finally gets the ring and he’s holding it up and dancing around? That’s me and TextKit.

So happy together.

Vesper’s detail view — with separate fonts for title and text — was barely possible in iOS 6. Dealing with changes in formatting meant replacing the attributedText and restoring the selection.

But TextKit changed that. You can change the formatting attributes of part or all of the attributedText without replacing the attributedText. Way better.

There’s a bunch more to TextKit — but just that one thing was huge for Vesper.

I didn’t, by the way, end up using TextKit for the notes list. We already had working code that used Core Text, and there was no reason to change that code, so we didn’t.

While I love TextKit, I didn’t love UITextView that much. There are new bugs, and we couldn’t work around all of them. But UITextView in iOS 7 is really a 1.0 release, and so I forgive it. (I need to file some Radars.)

64-bit Support

It took just a few minutes. I bet that’s true for most apps, though games might be an exception. (I don’t know much about games; I could be wrong.)

Since I’d been through this same transition on the Mac, I was already careful to always use NSInteger, CGFloat, and friends.

The compiler caught all the places I needed to make changes. (I use Hosey-level warnings.)

There were a bunch of places I used ceilf and I needed to use ceil for 64-bit processors, since CGFloats are doubles on 64-bit. The other cases were similar (floorf/floor, fabsf/fabs). (I use ceil a bunch in layout code because I want to make sure view frames use integral values, and some things like text measurement return non-integral sizes.)

Toolbars

There are a two places where we use UIToolbar: in the detail view and in the web browser view. I like that the new UIToolbar is translucent, with a live-blurred background.

But I treat UIToolbar like any other view. In the past I’ve found that it’s difficult to precisely control the location and tap target size of buttons inside a UIToolbar, so now I just add buttons as subviews. I don’t particularly like doing that, but it works.

Gesture Recognizers

Vesper doesn’t use a UINavigationController, so we don’t get the automatic pan-to-go-back gesture. Apple provided a new gesture recognizer — UIScreenEdgePanGestureRecognizer — that allows us to handle pan-to-go-back ourselves. It’s a subclass of UIPanGestureRecognizer, as you’d expect. Easy to use.

We don’t, by the way, use UISwipeGestureRecognizer anywhere. This is because a swipe is just a gesture, while pans are direct manipulation.

My favorite new small thing

- (CGFloat)tableView:(UITableView *)tableView estimatedHeightForRowAtIndexPath:(NSIndexPath *)indexPath;

Things I Didn’t Use (or didn’t use much)

Dynamic Text

iOS 7’s dynamic text feature is very cool — but it doesn’t support fonts other than the system font. We added a Typography screen to Vesper, so that you could change the font size and weight in the app (and even use small caps for titles).

The API for dynamic text returns a UIFont. I could have just grabbed the font size and used that with our custom font. But that would have been a hack, and it felt like we were solving a problem that’s really Apple’s problem to solve.

(I can see why dynamic text supports only the system font. But for apps that use custom fonts, it would still be useful to be able to get the user’s preferred font size.)

Dynamics and Motion Effects

It seems like Vesper is exactly the kind of app that should use dynamics and motion effects, and I certainly plan to in the future. But I already had working code to do the things we wanted to do, and there was no reason to rewrite that code.

I did write one parallax effect — see the image behind the sidebar. At first I made the rookie mistake of making the parallax move the reverse of what you’d expect. My designers caught it.

I’m looking forward to using this technology more. So cool.

Autolayout

We use autolayout in one small view in Vesper.

I’m still not sold on it. When I’ve used it, it doesn’t make for less code, and it does make for trickier debugging. I don’t like the way it interferes with animations. (Just animating to new frames doesn’t work — constraints have to be dealt with. Which makes sense, but I don’t like the extra work.)

And it’s not as if layoutSubviews is particularly hard

But still I have this itchy sense that autolayout is the future and I really need to adopt it. So I will, but incrementally.

Xibs and Storyboards

We use xibs only for the Credits screen, and I’ll probably change that to all code as soon as I get a chance.

I don’t recommend this route for every app. Vesper has very few screens and just about everything is custom. Xibs (and storyboards) don’t buy me anything. (But, again, that’s not true for every app. I don’t recommend going xib-less as a general rule.)

Core Data

What prevents me from using Core Data at this point is my concern for scalability and performance. It’s possible I’m just being thick-headed.

Likely, even.

Vesper 1.007

Vesper has been updated for iOS 7. Changes include a design refresh, typography settings, 64-bit-support, and more.

You can get it on the App Store.

Recursive Locks

David Butenhof explains, in 2005 in comp.programming.threads, why POSIX has recursive mutexes: “Because of a dare.”

He also writes:

First, implementation of efficient and reliable threaded code revolves around one simple and basic principle: follow your design. That implies, of course, that you have a design, and that you understand it.

A correct and well understood design does not require recursive mutexes.

Fiery Robot writes, in 2008, Recursive Locks Will Kill You!:

Better to use a real honest-to-goodness lock. This means that if a thread tries to lock a lock that’s already, well, locked, you’ll deadlock. This is good. I like this, just as I like crashes. I know, you think I’m crazy, but if you deadlock on your own thread, then you know you just did something bad.

Google Mac Blog writes, in 2006, in @synchronized swimming:

Shark told us without a doubt that we were paying heavily for the @synchronized block each of the millions of times we were calling fooFerBar… So, I wondered, how exactly does @synchronized work? And is there a cheaper way of getting the same thread-safe result?

(For new Cocoa developers: Shark was the performance profiler we used before performance profiling was added to Instruments.)

Which Locking System I Use

I haven’t used @synchronized in years. Not only is it recursive (which I tend to avoid), it sets up an exception handler. I would way rather have my exceptions go unhandled. (I don’t know if it’s still slow compared to other locking systems, but I strongly suspect it is.)

So there are two that I use:

  • OSSpinLock. It’s great when contention is expected to be low. (Which is surprisingly often.) It’s super easy to use.

  • Pthread mutexes. You can create these as recursive locks, though I try to avoid it.

(There’s one recursive lock in Vesper, and I’m switching it to non-recursive — to OSSpinLock — today.)

Better, though, is to design so as not to need locks. Good use of GCD serial queues, blocks, and the main thread can often make locks unnecessary.

FCModel

After writing my article about SQLite + FMDB for objc.io, I thought about generalizing my database code and releasing it on GitHub.

But now there’s no need: see FCModel by Marco Arment.

(My code and Marco’s code are different, but the approach we take is so similar — probably necessarily so — that a second implementation wouldn’t make sense.)

Concord Open Source

Dave Winer released Concord, a JavaScript outliner, under the GPL license. It’s up on GitHub.

It’s useful anywhere information is structured and organized. Like file systems, mailboxes, chatrooms, databases, documents, presentations, product plans, code, libraries, laws, systems of laws, contracts, rules, server logs, guidelines, principles, docs, manifestos, journals, blogs, podcasts, server, clouds, etc.

SQLite

My article On Using SQLite and FMDB Instead of Core Data appeared in objc.io issue 4 earlier this week.

Archive