On David Pogue’s Review of App Inventor

David Pouge gave a poor review to App Inventor in his New York Times article, D.I.Y. Tool for Android Needs Work . Because of pre-beta glitches, he and his son were unable to complete the tutorials. I think his frustration led to him making some off-base conclusions about App Inventor. The conclusions he came up with after a day certainly don’t mesh at all with the experience I’ve had teaching the tool for the past year.

He says:

“But make no mistake — this is programming. You can’t do it unless you’re a programmer, or have one nearby.”

He is correct in that App Inventor app development is indeed programming. But he is wrong in that you– someone who has never programmed before– can do it! Learning to write apps in App Inventor is significantly easier to learn than a traditional text language (more on this later). It lowers the bar dramatically. There are limitations, and some will be able to do more than others. But all of my students, with no prior programming experienced, got to the point that they could create apps without aid.  David’s sentence is misleading, and I hate to see it dampen people’s enthusiasm for this great tool.

App Inventor programming vs. textual programming

David asks:

“is there really such a huge difference between having to type out ‘AccelerometorSensor1.XAccel,’ as in real programming — and dragging a block bearing that name?”

This is a great and fair question, and to look at it clearly I think it really helps to be a programmer and to have taught beginners. From that vantage point, the answer is emphatically YES! In practice, there is a huge difference between typing commands and choosing from pre-defined blocks. Many of a beginner’s frustrations come from syntax errors– trying to instruct the computer with a textual language and being barked at with indecipherable messages when there’s something wrong. Of course the complex logic they face later is another barrier, but its the “syntax” frustration that turns many, many beginners away from even getting to the more fun logical problem solving. If App Inventor just gets people past that first hump, it is a great tool.

Another key is that App Inventor is an event-handling language, meaning events are first-class objects. You can directly drag out a button.click event block and then place the response to the event inside. Seems logical, and you’d think after all these years it wouldn’t be such a big deal. Unfortunately it is– you can’t even imagine how much conceptual understanding you must have to do the same in Java or Python.

David should ask a programmer who has never programmed the Android to write a Java program that prints out the readings from the Accelerometer. Hell, he could even ask one with some Android experience. See how long it takes the programmer. He’ll find that that single Accelerometer block will be replaced not with a single line of code, but an import statement, an object creation statement, manifest files, etc. The App Inventor team has done a  lot of work to package a bunch of ugly stuff inside those blocks. You have to think of it as a great code library, specially designed for beginners, and transformed into visual blocks language.

For more thoughts on why App Inventor works, check out my previous post.

No Fundamentals Mentioned

David didn’t mention any fundamental issues to back his negative conclusions about App Inventor. In a reply to a comment concerning this, he mentioned only one thing:

“I explicitly let Google off the hook for the aspects that are related to the early stage of the software, and critiqued ONLY the fundamentals. For example, this business of needing two different apps to do your development, switching back and forth. That’s not going to change when the software is final.”

In my experience, the switching back and forth between the component designer and the blocks editor is barely bothersome to most when they begin,  and completely forgotten by all soon thereafter. There are plenty of fundamental issues with learning to program, but this is not one of them.


David’s article is interesting and provocative, but my guess is that the start-up glitches and his anti-hype feelings had an undue influence on his opinions. Start-up glitches often do this– we all know the importance of first impressions. Google probably should have made sure more phones “worked” with App Inventor out of the box. Its scary to think that if David had been using a Nexus One on a Mac, his critique would have been completely different.

In any case, I’m inviting David and his son to San Francisco– I think given a day I can change his mind!

Why App Inventor Works

Sharon Michaels wrote a great review of App Inventor and like many lauds its visual, drag-and-drop interface.  In teaching App Inventor for a year, I’ve seen that it does work for non-programmers and I’ve thought a great deal about why. Here are the key features of the language:

No syntax — The blocks language eliminates the need to remember and type code. When I teach beginners textual languages, their biggest pain is syntax, the blank page, and the cryptic error messages that Python/Java/etc. provides for them. Most beginners quit because of the frustrating experience of syntax errors.

Events at the top level — Traditional programming languages  were designed when the best metaphor for a program was “a program is like a recipe– a set of instructions”.

Since the advent of the graphical user interface, programs have not been recipes, but sets of event-handlers. “When this happens, the app does this. ” is the correct conceptual model.

With App Inventor, an app is a set of event-handlers. I know this is amazing, but you can say, “when a user clicks this button, ” by dragging out a block. Compare this with Java, which requires you to understand classes, objects, and listeners in order to express an event-handler. Even if App Inventor were a textual language, its 1st order events would dramatically ease the development task.

Choose from a set of options — The components and functions are organized into drawers. You program by finding a block and dragging it into the program. You don’t have to remember how to enter an instruction or refer to a programming manual.

Only some blocks plug-in — Instead of chastising the programmer with cryptic error messages, App Inventor’s blocks language restricts the programmer from making mistakes in the first place. For instance, if a function block expects a certain type of parameter, you aren’t allowed to plug in a different type. This doesn’t eliminate all errors, but it sure helps.

Concreteness — You program components, not abstractions. When you drag out a component, an object is created and function blocks created for it. This is a problem in terms of code reuse and program size, but the concreteness is a boon for beginners and is pretty nice for experts as well.

High-level components — The app inventor team has built a great library with simplicity the main goal. There are months of programming expertise embedded in the components.

App Inventor: the electronic napkin for designing apps

What excites me is the prototyping capabilities of App Inventor. Suddenly you have an electronic napkin for sketching out app ideas. Having a semi-working, interactive app is huge in terms of formulating ideas and getting them across to others. Like new terminology and abstractions, it gives people a way to talk and riff on ideas.

In a broader economic sense, it enables a bunch of smart, creative people to get involved in something that they’ve pretty much had to leave to the programmers. Consider two  twenty-somethings sitting around in their kitchen drinking a beer. They start talking and come up with a great idea for an app. One dude says, “We should really do it!”. The other says, “but who would we hire to program it.” The first dude takes a chug of beer, “Oh yeah, forget it.”

Now envision a world with App Inventor. The idea comes up and now the second dude (actually a dudette) says “this is cool. Let me get my laptop and we can build a prototype for it TONIGHT.”  They stay up building the prototype and in the process refining their ideas and getting EXCITED. Their wild idea has manifested itself into something tangible (well, virtually tangible). Not a complete, perfect app, but something that will spur them on, something they can show others and get feedback on, something they can show to angel funders!

They live happily ever after. Dudette even asks Dude to marry him, if he’ll stop working on the computer so much.

On App Inventor and Lowering the Bar

Steve Lohr of the New York Times calls it Do-It-Yourself App Creation Software. Salon’s Dan Gillmor says it is the Hypercard for mobile development. Others have put it in a more recent context by calling it Scratch for App Development. Mike Loukides of O’Reilly wrote:

This is revolutionary; they’re not trying to lower the bar, they’re throwing it away entirely.

I’ve taught App Inventor to non-programmers for a year as part of Google’s pilot program and I can say from experience that these guys are right: App Inventor will change both how apps are created, and who creates them.

Will the bar be thrown away? Not exactly, but a lot more people can get under the bar– and past the syntax barrier that has forever kept them out– to enter the app building world. Once in there are still difficult logical and design mountains to climb, but by then your hooked. Its like the barriers to learning simple conversational French have been eliminated, so that you can immediately chat with that French woman you meet at the bar. To get the girl you have to be smart and speak some poetry at some point, but now you’re motivated, a hell of a lot more motivated than if you had to go back to Cleveland and spend a year learning how to say, “Souhaitez-vous un verre de vin”

I see App Inventor creation at three levels:

Level 0, Anyone: some simple but very useful apps. These apps make use of the high-level components provided by App Inventor (Texting, LocationSensor, TextToSpeech) but require no repeat loops or complex logic. An example is this screencast of No Text While Driving, an app designed by Daniel Finnegan, a Politics major in my class at USF.

Level 1: Anyone who likes French Accents: Apps with some sophisticated logic. This level requires some hard thinking and problem solving. The blocks language has gotten you past the syntax, but now you need to piece together some repeat loops and index variables, and maybe a nested if-else. Examples of apps at this level include the Quiz and Create a Quiz tutorials I wrote for the App Inventor site. Who can do this? Just about anyone, but here you start losing some people that just don’t want to think that hard.

Level 2.0:  Apps that talk to APIs: You can get past the limitations of the visual environment by talking to back-end web services. Here you need to program in another language like Python, or know a programmer real well. App Inventor has provided a bridge with its TinyWebDB component. Basically, a programmer has to write a web service proxy that talks to the service you want and follows the App Inventor protocol. Then App Inventor programmers can create client apps that use TinyWebDB to communicate. I’ve written a few App Inventor web services using Python and App Engine, including proxies for Amazon, Yahoo Finance, and San Francisco’s NextMuniBus API (my students used this service to build their DroidMuni app with App Inventor).

The web service programming is quite simple,  for a programmer. You can download one of my samples and build an App-Inventor compliant service within minutes. But the barrier is back up, big time: most people will never write web services. But I do envision the following:

  • Some App Inventor programmers graduating to Python (or another language) because they are so motivated.
  • Organizations providing App-Inventor-compliant APIs that App Inventor programmers can then mash.
  • Collaborations, on-line or off, of App Inventor client programmers and App Inventor API builders. For instance, one of my CS 1 students worked with some CS 0 (App Inventor) students in this way to build an app using the Eventful API.
  • Yahoo-Pipes-like functionality for App Inventor

Level 2.0, the bridge to “back-end code”,is the key to App Inventor’s success. Without it, App Inventor would have been a nice educational tool. With it, App Inventor will change the world of app development.

Not sure? I’m not either, but they say you should bet on the toolmaker, along with the tool. Dan Gillmor wrote this about Hal Abelson, App Inventor co-director:

a brilliant computer scientist who also understands how app development needs to get into wider distribution, not just the coder community

This statement resonates with my experience working with Hal and the team during the university pilot and this summer. They are money, and they don’t even know it. This is why I’ve thrown my hat in with App Inventor.