Pros and cons of using XIBs and doing views programmatically - iphone

I want to decide if it is better to use XIBs or to designs my views completely using code.
So far I have read that when you design your views on interface builder they are pre-built, so even if they use more memory the user feels everything is faster.
People say doing everything using code is harder but I find it to be just as easy, so I want to know if anyone has experienced some real speed gains when using nibs.
What have been your experiences, advice, etc?
Thanks!

You should be able to do both -- there are times when building a view programmatically is better/easier, and times when using a .xib is better/easier. Even if you only ever do things one way, you'll run into code that does it the other, and you'll need to be able to deal with that.
If you don't know how to use IB, then building your views in code is certainly easier. That is why you should learn to use IB. Once you understand IB, it's way, way faster to put together most of the view-based UI your app will likely need. IB helps you line things up, center objects, align base lines, connect controls to their targets and actions, etc. I think it's safe to say that everyone who uses IB effectively experiences "real speed gains when using nibs."

You should know how to use both. Performance differences between the two are negligible and should not be the reason that you choose one or the other.
Many people who are new to iOS development have the misconception that nibs (.xib files) are inferior to programmatically creating your UI and that if you use IB you're not a good iOS developer. That view is 100% wrong. IB is created by Apple and in use by Apple's developers to create their own Mac OS X and iOS apps. If IB (as a tool) is good enough to be used by some of the best developers in the world, it's probably good enough for most of us.
In practice I have found that a combination of the two usually fits the bill.
In my own apps I find that .xibs are great for laying out the basics of your views quickly and they allow you to iterate very quickly while giving you a preview of what your view will look like. It's also much easier to use auto layout in a .xib file.
Then when you need to do more advanced things like add fancy animations or move views around that is what IBOutlets are for. Anything that you put into a nib can be referenced through an IBOutlet. This allows you do then programmatically make your view come to life.
Lastly, you should fully understand what a nib (.xib) is doing automagically for you. You should understand what happens when a .xib's objects are unfrozen. There are many resources on the internet to understand .xib files better.
Also, learn how to use .xibs in an encapsulated way. For example, .xibs are crazy useful for things like prototype cells and they allow you to keep your code base modular (much more so than storyboards). Also, you will require less UI code in your view controllers.
Lastly, I always say that people should think of IB/.xibs like jQuery. It's going to save you a lot of time but the best developers still know how to do everything in javascript if they have to.
Good luck and have fun!
TL;DR version
Performance is not a consideration when deciding to use .xibs or not.
Use .xibs because they give you a preview of the view you are creating and they allow you to quickly iterate
In practice most apps will use a combination of both. You will programmatically add animations or move views around but the .xibs will be a starting point
Understand fully what happens when the objects in a .xib are unfrozen
You'll be more productive but be sure you fully understand what is happening behind the scenes.

I would always use XIB files unless there was a reason not to. This allows your views to be maintained easily in the future.
Some reasons for creating the views programmatically might be:
A control needs to be resized,
repositioned or otherwise altered
depending on something else
Controls
need to be added or removed
dynamically
There may be more reasons but there are not too many.
If you programmatically create views when there is no need you make it a lot more difficult for other developers to try to figure out what the view will look like and to change it.

If you build your views programmatically, you have control over the loading of elements. e.g. you could use lazy loading, and load secondary buttons, subviews, etc. a fraction of a second after the more important elements, allowing the key parts of the UI to come up faster. You could even animate some elements into position.
If you use IB, you get guides as to proper element spacings and positioning, but you could always copy the coordinates from IB into code if you aren't changing the design that often.
For simple UI elements, you will end up with more lines of code to maintain if you create them programatically.

IB and NIBs do a lot to optimise loading/unloading of views, but it is largely oriented to minimising memory usage vs. perceived speed for the user. For example, lazy loading if anything might make the app UI slightly slower, but it should make memory usage lower. This in turn could make overall app performance better on a large application, and is very much encouraged, but it's difficult to define "performance" in a narrow way. It's also difficult to say when you should or should not use IB - there will be some times you're much better off doing it in code.
One often overlooked element to the IB or not debate is development speed, especially if you have multiple developers. On a larger team/project you'll probably have some developer(s) who specialise more in the infrastructure, business logic etc. of the app and some developer(s) who specialise more in the UI. In this case, use of IB will make it easier for them to work independently, which should make overall development more efficient.
I view IB as a core part of the development platform for iOS development. It's not the right solution in every situation but not knowing how to use IB will be a real limiting factor.

Related

Modular Architecture for a Large Application With Best Practices and Usage of Storyboards and XIBs

We are a team, creating a very big application for iPad that serves as an eReader for unprivileged children. The app is built halfway and I thought we should take a step backwards and review the whole design of the application. The application we are building should be very, very compliant with the current software development architecture practices for iOS. I have the explanation along with the questions below:
The application right now, has about 50 views (and increasing) and most of the Top-Level views are in the Storyboard (a single storyboard, that is) and the others are in XIBs (for the sub-views, reusable item renderers etc).
Is this approach alright?
Should XIBs be completely omitted for the modern iOS applications?
Should the storyboard be right the way it is or should it be broken down into sub-storyboards? If they should be, how should the exact process of decomposing the storyboard be done? How would the modules be determined?
If you have some tips to manage the application in the right way or some rule of thumbs to assist in a streamlined and modular application, please mention it in your comments.
I'm sorry for the long text up there and I thank you for reading.
This is a bit of a subjective one, but in my opinion it is still valid, and desirable, to use Xibs alongside storyboards. The idea of storyboards is nice, but with the current implementation they are definitely lacking some functionality, the most glaring of which is view reuse. I have worked on a few large projects recently, and storyboards always start out nice, but sooner or later you find yourself copy and pasting table cells, and then controllers, and then you have to change things in multiple places with each update, which is totally unmaintainable. So yes, definitely stick with the Xib files for reusable views.
I am less keen on the idea of multiple storyboards however. While this is easy enough to achieve technically, for me it invalidates the purpose of storyboards entirely. Their only use is as a (relatively) clear overview of the flow of controllers through the application. They add a few small conveniences, but in general I find they often create more code than they save, and create some odd code patterns (eg prepareForSegue:, having to temporarily store ivars when you want to segue to a controller in code that needs properties setting). As you say, how do you split up the storyboards into sub-storyboards? I can only think that whatever partitioning scheme you choose would be somewhat arbitrary, and liable to change at a later date when the app structure changes, which would be no fun at all. Maybe if your application has distinct modes of operation it makes sense to split up the storyboard, but I don't think this would apply to most apps.
So, IMO, single storyboard but use Xibs for reusable views.

Choosing between the two - Interface Builder OR creating via code

When I started with IPhone development I preferred using Interface builder for creating views in my application. With time I considered the option of creating the application via code which I really feel comfortable working with.
Now, when I know both ways to create my user interface, I have doubts which way to follow. I keep thinking where to use IB and where to use code. How can I figure out before starting with my application, which way should I create my UI?
There are a number of factors that can influence your choice. It can come down to personal taste, but there are some advantages to using Interface Builder.
The first thing to recognize is that UI design is inherently a visual task. Interface Builder can allow you to create and modify a UI much faster than you can in code. Rather than endlessly tweaking CGRect values in code, then recompiling, testing, and repeating this process, you can instead get direct visual feedback about your changes. This means you can refine and polish the UI much more efficiently, and even test out radically different layouts without too much time or effort.
Another important point is that if you can create and layout a control in Interface Builder, that means there is less code in your view controller. Less code is always a good thing (less bugs, easier to maintain, ...).
So I believe you should try to define as much of the layout and properties in Interface Builder as you can. This can be hard for people who prefer "full control" over things. Many years ago, visual GUI editors were renowned for doing the wrong thing, and developers often shunned their use over doing everything by hand. But Interface Builder is pretty good at what it does, and you can easily dive into code when you need to.
The situations where you would create controls purely in code are when you need to dynamically create a number of objects, for example thumbnail buttons for a photo gallery, or if you have custom layout requirements.
For example, if you have a label whose height is dynamic at runtime, and you want to layout a bunch of controls underneath it (taking its height into consideration), that can only be done in code (UPDATE: iOS 6's Auto-layout can do this now). You can either create and layout everything in code, or you can still create the controls using Interface Builder and just give them arbitrary positions in the XIB which are modified when you lay them out in code. It's not uncommon to see this in some of my XIBs:
The buttons along the top can be created and layed out fully in IB, but the image views below need their images and positions to be calculated at runtime. So we just give them arbitrary positions in the XIB. Some might find this ugly, but it saves time and effort on writing code to create the image views and set up their properties. Again, less code is a good thing.
Other times you may want to use code is setting certain explicit properties. Let's say you have an image view which needs to receive touch events because it has some gesture recognizers added to it. You need to set the "User Interaction Enabled" property to TRUE. It can be better to do this in code because the requirement is more visible and you can leave appropriate comments as to why you are setting the property. It's also harder to 'lose' the setting like you can in interface builder if you need to delete and recreate a bunch of views.
I think for many applications that use the stock UI, Interface Builder is a great tool to rapidly get things up and linked with the underlying code. Also, it really stresses the paradigm of the View being separated from the Controller as you really can't push code into places where it shouldn't be.
That being said, I use it less and less the more I learn about how to rapidly code interfaces due to the fact that they may need to be more flexible or need a variable number of UI elements based on the Model behind it.
Use interface builder as per your requirement. It's depend on you that how you manage things. See creating a custom cell using interface builder is much easier to update at a later stage while using code you have to do a lot of changes if layout changes. Also you can visualize the view before running the actual app but in code you cannot you just have to assume.There are both prons and cons of both things.

How to achieve reusable code when doing everything programmatically, without using Interface Builder?

I'm the programmatic guy, and I simply don't want to use Interface Builder. I feel out of control, and besides that my GUI is about 90% custom all the time.
Literally every book does everything in Interface Builder and claims that this is the one and only great way to have real MVC going on.
Example: One of those books mentions that programmatically creating an UINavigationController with an Root View Controller and everything else that belongs in there is a big mess and won't be reusable when porting to the iPad, while doing this in XIB is a clever decision. Then the port to iPad using UISplitViewController will be a simple task.
So when I make iPhone apps and want to port those to the iPad too, what strategies work to reuse as much code as possible? I'd like to learn more about how to separate my code and achieve a better overall architectural design without using Interface Builder.
For those who want to tell me I must go with IB: Again, I do a lot of custom UI where IB is often just in the way. And not to mention all the animations. I really have my reasons. For people who make default UI IB is really fine - but please, I don't want to start a fight for IB vs programmatical UI or default UI vs custom UI! It's all about how to achieve great reusable code when doing everything programmatically, and both have their pros and cons.
Although you did not ask for it, I feel compelled to make the case for why people in general (perhaps not you) should consider IB, and then address the issue of custom components.
I use a lot of animations and custom components. And I love to use IB...
The key is to use IB for its strengths, and then decide what to do with the rest from there. What then are the strongest points of IB? Connections, placement, auto-resizing and customizations.
Connections are linking aspects of views and controllers together. It's faster in IB to drag out a few connections to delegates or references, than it is to write the code that forms the connections. And, it's a quick place to review all links to the UI you are building.
Placement IB also does well at. There's a fair amount of code involved in setting up any GGRect correctly. Not only is it easier to enter and review coordinate and size details in IB, but the tool automatically sizes a lot of elements properly for the container and the control, and offers many guides to help things line up properly - that kind of thing can take a lot of repeated testing to get right.
Related is auto-resizing. Although I don't feel that many screens can actually have auto-reiszing rules that rotate the screen and come out the other side looking just right (I almost always do rotated views as a separate XIB file), there still are a lot of shifts that can occur in the course of running your application that make it really useful to have these defined just right. The best example of this is the enlarged status bar while you are on a call.
Lastly comes customization. This again can be a lot of tedious code to write; try setting up all of the properties on a UILabel manually and it'll have you yearning for quick changes in IB.
With all that said, what is a good approach to custom components? I like to use UIViews in IB screens, with the class type set to a custom UIView that then fills out the display at runtime. But at least IB helps me get composition, placement and auto-resizing just right with minimum fuss, and also wire aspects of that custom view into a controller.
The one thing that would really lend IB to use with custom components is if it would simply let me set values for any simple properties the custom view had - then I could adjust parameters like a corner radius or whatever else I had going on.
I urge you to think on IB a little more, as it's a huge productivity boost when used correctly. There should be nothing about IB that gets in the way, it's there to boost your output.
One book I really liked was Erica Sadun's iPhone Cookbook 1st edition. It did everything programmatically.
Unfortunately the second edition is bloated.
If you reuse lots of your custom UI objects, it would make sense to write a code which
reads a plist (or a more general XML file) specifying how the custom UI objects should be placed / animated
and then creates your custom UI objects accordingly.
It's like writing a mini-xib file format tailored to your UI objects; you can also feel that you're in control of everything, as an added bonus.

Good reasons why to not use XIB files?

Are there any good reasons why I should not use XIB / NIB files with an highly customized UI and extensive animations and super low memory footprint needs?
As a beginner I started with XIB. Then I figured out I couldn't do just about everything in them. It started to get really hard to customize things the way I wanted them to be. So at the end, I threw all my XIBs away and did it all programmatically.
So when someone asks me if XIB is good, I generally say: Yeah, if you want to make crappy boring interfaces and don't care too much about performance, go ahead. But what else could be a reason not to use XIB?
Am I the only iPhone developer who prefers doing everything programmatically for this reasons?
I think that Interface Builder is one of the biggest assets of Mac (and by extension, iPhone) software development. GUIs are visual; why not create them using a visual interface? IB is flexible enough that you can lay out an interface using its "generic" components, and then subclass them where necessary. Sure, if you have a unique interface you're going to have to subclass a view class and perform custom drawing, but you can also lay out your interface in IB and then easily use the inspector to switch the class to your custom subclass.
Honestly I think it's a spectrum of convenience. If you are comfortable writing everything in code then go for it. If you design your project well then it should be about the same amount of work creating new windows, etc. But I know that a lot of people aren't as comfortable with the GUI world so nib/xibs work well there.
I honestly find myself using XIBs as a base quite often and editing them with code to get the specific look I want. Personal preference.
For a specific con on that point, views can be difficult to configure after loading them from a xib. When you have conflicting settings between IB and code that can be nasty to troubleshoot.
Here's a question for the list. What is the performance hit to using a xib? I thought they were a plus because they don't get loaded into memory until you need them. That said, that load time is longer which will slow your program down. Thoughts?
One thing I found better about code is for the event connections on controls, when you search for uses of a method (message) you find them if they are coded and you don't find them if they were set in IB.
On the other hand laying out objects on a view is much easier in IB where you can see their size and positions. When you do that in code you have to guess at the size and origin settings and then run it and make adjustments, then run it again to see what it looks like.
When your application has some kind of "standard" views, go with the XIB. If you need real customization, depending on external content (XML...) do it programatically.
I started using XIBs and now it's all code, I find myself more comfortable this way. I had real problems with XIBs, and now writing the interfaces all in code really saves me time.
I save tons of time when dealing with UIControllers (UITabBarControllers, UINavigationControllers etc.) in the start up phase where all the navigation stuff is hooked up.
I just build X viewControllers with a accompanying XIB, throw in the stuff needed in IB, labels, images etc. This means that for almost any sort of app you can have a proof of concept up in a few hours. This is enough to justify spending some time learning the ins and outs of IB. Especially on the iPhone where you can have a ton of good UI ideas, but they all fail when they move from the Simulator to an actual device.
The best thing, in my mind, is to balance it out, if you find yourself using a lot of time doing the "change the frame 3 px -> compile -> ahh.. needs two pixels more -> change 2 px - compile -> ahh.. 1 more px" for something that could be done in IB, you will seriously start to waste time.
I start as above, but afterwards I often throw the XIBs away for custom stuff. The trick is to not spend hours on implementing versions of custom stuff in code over and over again, but figure out how it should be and do the custom stuff once:)
The XML content of a nib file is very complicated. This makes it extremely difficult to review changes or fix merge conflicts with a version control system like Git.
Interface Builder is a nice idea, but Bret Victor, in his talk "Inventing on Principle" and his essay "Learnable Programming," implicitly challenges Apple to build an even better IDE.
One idea, based on Bret Victor's principle: What if I could select a "Move Tool" in the iOS Simulator app that let me move a button in my app and then the frame code changed in the implementation (.m) file? This would be much better.

NIB/XIB files with Cocoa programming - faster development time?

I have been programming with the iPhone SDK for some time now.
I have not been using Interface Builder. This scares me a little. I know that in effect I may be 'fighting the framework' but I do not see it that way.
I find it very easy to just instantiate my UITabBarController in my app delegate, instantiate a UINavigationController, then push and pop view controllers as I go.
Naturally I do not have an extensive knowledge of how to architect an app with XIB files because I have never done so, however I do know the general gist of it, having built some Mac apps in Cocoa using NIBs. So I am not completely ignorant.
My question is whether there is an increase in development time when choosing to lay out UITableViewControllers and UIViewControllers using XIBs rather than programmatically instantiating them and then setting up the ivars.
As I see it, both methods still require you to subclass the view controller for customization which will probably occur for the majority of your views. As well, there are still manual classes required for delegates, and the process of connecting outlets from within the XIB seems comparable to me from setting an ivar.
Or am I missing some other major point?
Thanks!
Code takes much longer to write to configure UIs than IB does.
Plus, you can hand off design to designers and let them tweak the UI.
In the end they both accomplish the same thing. You should use either one depending on the circumstances. Most of the time writing the code to create and position views, and especially maintaining it down the road, will take much longer than using IB. In a simple app for the iPhone though, this might not be true and you'd be just as well off creating everything in code. Basically, you should know how to do both, and pick the path that involves the clearest code and quickest development.
IB shines when you're using it to actually lay out views; even two or three views can be a real hassle to lay out and configure in code. I do tend to use it for tab bar and navigation controllers, and sometimes for subcontrollers (usually only if I think the user is very likely to use it), but that's more just because I'm already there so I find it convenient.
With this new version 3 OS they're announcing next week, I'm hoping Interface Builder gains some of the flexibility it has in Cocoa, where you can add palettes for your own classes and even build up complex non-view data structures (by using custom palettes). We'll have to see, though.
Don't worry too much, IMO Interface Builder is a little over-rated too.
It's definitely useful for getting things up and running quickly, or if you have an app with a lot of screens that are tedious to setup, but you're not missing much.
For the uses you outline just doing things in code is fine, and possibly even a little easier to understand.
Laying out views, or custom cells though... then you get into a ton of font/color/position setting that quickly explodes into a lot of code, hard to maintain and tweak. Much easier to adjust what you want in IB in those cases.