Slashdot Mirror
Developing Android Apps Visually, In 3 parts
An anonymous reader writes "Dr. Dobb's has a three-part blog (all three parts are up; this is part 1) about using App Inventor. The focus isn't so much on the technology but rather the discussion of 'can visual development let anyone program?' If so, is App Inventor really visual development? And should we be teaching real programmers about visual development. Most of the conclusions are in part 3. As a byproduct, they show you how to put App Inventor output on the Market and there are two games on the market (free) that resulted from the articles." Here's part two, to round out the trilogy.
Coincidentally I just started learning to develop mobile apps last week. I'm using Sencha Touch and PhoneGap, Eclipse, and the Android SDK. The combination works pretty nicely, and lets me build fairly pretty pseudo-native apps, working in JavaScript. Best, they will run on iOS and any future mobile device with WebKit.
My blog
...your spaghetti code will actually look like spaghetti!
Pretty good is actually pretty bad.
...don't use visual tools. They describe the GUI in assembly language, or use torturous frameworks. Of course it is this elitist attitude of making things as difficult as possible that has resulted in 2 decades of user experience that stinks. I don't know how many times I've seen programmers rant that Visual Basic was evil because it was too easy and let anyone program. They somehow think putting together a user form should require 2 weeks and multiple degrees in computer science. On the contrary, it should be ridiculously simple to throw together a user form. There are things you can't simplify like algorithms and complex logic in science and business and THAT is where you NEED to focus and concentrate a developer's attention. Bloated frameworks and non-visual building tools from hell that make things unnecessarily hard are nothing but a hindrance and should be eliminated. There's no shortage of work to go around.
These posts express my own personal views, not those of my employer
Slightly off-topic, but Android development for me has been marred by the steaming pile of dung that is Eclipse. Netbeans is ok but it's android support isn't great.
I finally got around to trying IntelliJ IDEA, and hooray! Android development is now possible on my lowly 2009 PC. It is so much better than Eclipse. You should download it now and forget about Eclipse this instant. Let's see:
Cons compared to Eclipse:
* Not the official android IDE.
* Doesn't have some android tools built in (ddms).
* No GUI editor for the manifest.
* No GUI layout editor (although the Eclipse one is unusable anyway).
* Logcat always autoscrolls. It's slightly annoying.
Pros compared to Eclipse:
* The main UI is way faster and more responsive.
* The 'smart' features (code completion, refactoring etc), are even more clever than in Eclipse -- they practically read my mind.
* No retarded 'workspace' paradigm.
* The code editor is way more responsive.
* The UI is a lot more sane, and much less cluttered, even though it still has a ton of features.
* Built-in git support. Maybe this is in Eclipse, but I'm sure it is way more complicated.
* No retarded 'perspectives'.
* The UI is cleaner IMO, although it is a little win95-ish.
* I have no idea why, but it manages to detect my phone even though adb doesn't. (I know right?)
* It's just way better. There are tons of features that make you think "Wow, they really spent time implementing that (in a good way)?", random example: if you create a new class, edit and press undo, it will ask you if you want to undo creating the class!
In conclusion, fuck you eclipse. You suck.
See, the one fundamental concept programs like this miss is that ANYONE CAN PROGRAM!
I'm sorry guys, I hate to break with the fleet of devoted programmers needing to feel like they have something on the world, here.
Programmers are no better than people in any other skilled trade. And, I'm confident that I could work in any skilled trade I wanted to. If I could learn how to program in twelve languages, who is to say that I wouldn't be a genius with plumbing, or electricity? The difference here is that I want to program applications, so I do it. People who don't want to be programmers don't. That's all there is to it. Anyone can program, and anyone can learn programming.
There's no doubt in my mind that this is development because a program is being created.
And if you're creating a program, you have wanted to create a program.
And that makes you.... a programmer.
Microsoft in the 90's showed us beyond a shadow of a reasonable doubt that no matter who easy you make the programming tools for non programmers, they're not going to use them because non programmers are devoted to the almost religious idea that they can't do it. It's like anything else that way. Tell yourself you can't do something, and you'll be right 100% of the time.
So if you want to create Android apps, create the damn android apps, but like it or not, you're a nerd now.
You're a nerd now.
Now you just need to become an expert at War Craft and Dr. Who, and you'll fit right in with the rest of us.
Later
You may now gaze upon my greatness.
The quality of those apps will speak for itself.
thegodmovie.com - watch it
Should we teach C programmers about Assembly?
Should we teach perl programmers about C?
Should we teach SQL programmers about perl?
Should we teach HTML programmers about SQL?
Should we teach Drupal programmers about HTML?
Should we teach anyone about the insides of the things they use?
Only if we teach Assembly programmers about opcodes, I guess.
I would like to know how App Inventor handles the chaos (read fragmentation) in the Android ecosystem. Chaos stemming from the different screen sizes, types, hardware especially that for processors and graphics and manufacturing quality.
Disclaimer: I am no app developer, but an avid Android fan, currently using the Samsung Galaxy S II and loving it.
...Open Inventor and the image analysis tool called MeVisLab (mevislab.de). Expensive, but best I've ever seen for RAD in this area.
App Inventor seems like labview for android. That means that programmers should hate it and hardware engineers should like it. That being said, I'm a hardware engineer and I'm writing apps in java. I guess I was expecting app inventor to be more intuitive but there is somewhat of a learning curve. I figured if I'm going to have to learn something I might as well learn the fundamentals (java) so I can apply it in other places. So I decided to focus on the whole eclipse android sdk package.
Thats not to say app inventor doesn't have its place. I'm thinking its good for an intro to programming or programming for non programmers in highschool or something. Just to get kids thinking logically. Or interested in something constructive. But serious programming it ain't.
I hate to just say "This is impossible, forget it and go on." I'd love to see it happen, but I don't think it ever will until we have near human level AI that can replace a human coder. Programming is not about dragging pretty blocks around, it's about visualizing and then creating a logical process that accomplishes something. Programmers have the (inborn, I believe) ability to do this; nonprogrammers don't, and no amount of training or simplified tools makes the slightest difference.
I was in college with them, the ones who never got it and repeatedly asked the sort of questions that made it obvious they wouldn't understand the answers. I've had to work with a few, too. It's pretty easy to spot, after a while.
The problem is that we're attacking the wrong problem, and the result so far has been tools that laymen still can't use and programmers hate because the pretty UI's just slow them down. That describes pretty much any system that makes you click and type stuff into little boxes instead of writing it in a text file.
I've done some work in visual languages, like Pure Data/Max mostly, and some things you notice:
It's easy to get started in them but, no matter how easy they make it, eventually you get bogged down in trying to look up the particular name for a block that does X, because any logic that takes more than two lines of real code or relies on tight loops can't be programmed literally in the visual way.
I'd say that visual languages give you a good entre to programming, but really it's just BASIC brain damage all over again -- visual languages use visual cues like lines or sockets to do what in fact are nothing more than GOTOs, you have to do a lot of hard coding, the language makes you do a lot of static decision making, you always are deciding to make (k) objects instead of arbitrary (n) objects; code reuse, structure, or metaprogramming are unheard of.
Don't blame me, I voted for Baltar.
That'll cost you extra, honey.
Eclipse works fine here, too.
I'm a non-programmer (well, only a single first-year university intro course) who tried out App Inventor a while back in beta.
I can say that the visual aspect of the interface is a huge help in explaining how a piece of code works to a non-programmer. However, I quickly found that the whole thing becomes unwieldy and confusing very quickly as the various sockets nest on one another unless you use a lot of functions to keep the code limited to short snippets. More a limitation of the visual metaphor than anything else, but probably also the sign of an unexperienced programmer (i.e. me).
I was, however, able to make a very useful dialer app that appends an access number from Yak Long Distance (using a phone number for the city of the user's choice) to any phone number in the phone's address book. The result is an app that lets me call people for super cheap rates without needing to write down or remember their phone number. The app works well, and I use it almost every day.
However: .apk file is over 2 megs, even though the program is quite simple. These bloated .apks produced by App Inventor consume valuable phone memory since you can't move an App inventor app to SD (unless the phone in question is rooted).
1.) As much as I like my app, I can't share it with the world on the Android market. So my effort felt wasted in the end.
2.) The
3.) App inventor is plagued with a dearth of modules, and the modules that do exist have tons of buggy features that return no values or wrong values.
All in all, App inventor is cool and can be very helpful in teaching basic logic to new programmers. It is not a serious development platform. It also has a long way to go before it will be truly able to live up to its potential.
Disclaimer: After intensively using the software to develop my app, I haven't touched App inventor now in over 6 months. I'm not sure to what extent these problems have been addressed, since Google has and continues to update App Inventor on a regular basis.
- Anonymous Coward - (A.K.A. a lurker who is too lazy to register)
I just start playing with AppInventor this week, and right off the cuff...it's got a lot of potential, but I haven't used it enough yet to know if it's really a serious tool.
:) That's just a minor quibble, however, and while I'm enjoying learning how to create Android apps, I do have a few concerns about the language. First, the language itself is completely obscured. There may be a way to bypass the GUI and see the code AppInventor is generating, but if so, I haven't found it yet. Having spent way more time than I like cleaning up the horrible HTML that both Front Page and Dream Weaver generate when my family members who couldn't (or wouldn't) learn HTML came to me for help -- and at some point, they always came to me to fix their HTML when FP and DW didn't get it right -- I tend to distrust visual coding tools. I would also love to see a comparison between execution times for two identical Android programs, one written in AppInventor and the other coded by hand. I'm curious how AppInventor optimizes the code. Also, I find that the programs get a little hard to follow by the time you get a page full of code blocks on the Block Editor. It may be just another case of the way I think hindering my adoption of the tool, but I seem to have an easier time keeping the code in my head when I type it out by hand rather than when I snap puzzle pieces together on a GUI. Finally, my last concern about AppInventor is that the "command" reference is somewhat lacking. It took me pretty much a full day, and numerous Google searches, to figure out how to use the TinyDB to store persistent data in AppInventor. In the end, the procedure I was using to store data in TinyDB was never running because I was getting an error in the routine that pulls data from the TinyDB because the way to tell if there is any data stored in the database is not exactly intuitive and is completely omitted in the documentation.
The Cons: I tend to be kind of a linear, procedural thinker -- I cut my teeth on BASIC, learned COBOL in high school, learned Pascal and Perl in college, and now use mostly Perl and a little Python -- so AppInventor requires me to approach writing programs a little differently. For example, in Perl, if I want to compare two strings, I think it out the way the line is typed on the console; AppInventor, on the other hand, seems kind of like programming in Reverse Polish Notation
The Pros: I am quite impressed with the ease with which I started using AppInventor. When I first started using Python, it was very easy for me to read someone else's scripts and comprehend what they were doing. Writing Python, on the other hand, was a bigger hurdle. To be fair, a lot of that was because I've been writing Perl for so long, that I try to do things the Perl way (okay...ONE of the many Perl ways ) and then have to search Google to find the way it's supposed to be done in Python. AppInventor, on the other hand, is just a matter of snapping puzzle pieces together. If you try to do something that would be a syntax error in a traditional language, AppInventor immediately pops up an error telling you why you can't do whatever it is you are trying to do -- and the error messages are pretty intuitive. Procedural errors are a whole other story -- see the caveat above about using TinyDB.
Experienced programmers may turn up their nose at tools like AppInventor since it lowers the barrier of entry so much, but IMHO, tools that make it easy for people to learn programming concepts are a Good Thing. Will people churn out crappy code in AppInventor? Yep. Do people already churn out crap code in Perl, Java, C/C++/C#? Yep. Will skilled programmers make well-designed apps in AppInventor? I don't see why not. I imagine the quality of the code will probably depend upon some of the concerns I described above, but the *design* will be a reflection upon the skill and experience of the developer. I don't see any reason why a good developer will suddenly be reduced to creating crappy apps with tools like A
MCSE? No, sir...I don't do Windows. Yes, I am an idealist. What's your point?
1) Yes you can. The DDJ articles have a link to a tool that will package them for market and has the two examples on the market. 2) Granted. 3) Granted.
OMG people are so stupid - this has been done again & again.
The entire first page is someone blithering about themself. There may be some useful information on later pages.
I've played with it, I've mocked up an app using it. App was 4mb by the time i was done. Then I turned around and rewrote it in Java, and it ended up being only 96k compiled so mocking maybe, but it's really not efficient.
I read the first blog post and immediately recalled my experience playing with Scratch http://scratch.mit.edu/
Looks like almost the exact same approach. My short experience with Scratch suggested that interesting apps could indeed be written in the framework, but that the complexity of any 'real' app would soon become burdensome in such a visual programming environment.
RETURN without GOSUB in line 1050
I've been writing software for over 25 years, with the last 20 of them being mostly GUI based.
The visual components of any non-trivial program will compose about 10% of the final product, with the other 90% being the code that does the actual work. AppInventor addresses the least crucial aspect of writing software -- the ability to create a user interface.* The ability to think abstractly, and to implement that abstraction, is far, far more important; and it is the thing that relatively few people can do well.
So no, AppInventor is not going to let just anyone write good software. Without the skills needed for other other 90% of software development, AppInventor will do nothing more than address a trivially insignificant aspect of writing software.
* Do not construe this statement to mean that designing good, clean user interfaces is easy. It is an art form all to itself, but constitutes a relatively small portion of creating software.
i looked it up. cobol does indeed use english works.
Having to work for a living is the root of all evil.
I'm sure someone can create a nice web-based, visual development tool for Android. AppInventor is not it; I found it to be cumbersome and useless.
Real programmers use visual tools when they are the right tool to use. If you are trying to lay out a pixel-perfect preference dialog for a retail app they can't be beat, and real programmers use them. And if anyone ever came up with a visual tool that makes the actual work of programming simpler, real programmers would flock to it; but don't hold your breath.
Programming is a process of progressively deeper understanding of a problem space. Visual tools allow you to easily represent a shallow understanding within the space explicitly supported by the tool designer, but the basic geometry of pictures is hopelessly inefficient (compared with text) at representing anything complex.
The work spent on making visual tools would be better spent writing high-level libraries for a modern language. I like google's go, but there are lots of equally good efforts out there. Given such a high-level library (which implicitly must exist to support the visual tool), the sorts of programs that can actually be expressed with visual block tools - like the one by the article writer - can be written in a short page of code, and not hit a brick wall when you need to step outside the designers problem space.
But the inefficiency of representation of visual tools is not in itself a killer defect. The real problem is that they fool novice programmers into ignoring the genuine complexity of even trivial programs, which leads to the production of bad software.
Hiding the Java code from you is not a problem if you consider App Inventor blocks as a new programming language; it's not a translator therefore it doesn't need to show you an equivalent "generated code" that people claim it's missing.
However, as a new language, it's missing a big feature: googleability. In order to debug and ask questions, you need to be able to easily show a piece of code and let people search it through Google. This means that you need the code to be in text, not diagrams.
Adding a feature like, right click on a block -> export to XML, would be tremendously helpful to make this prime time.