Monday, January 28, 2008

Debugging with Distel

Until recently, I've done my Erlang debugging either using the built-in debugger or not using a debugger at all and instead relying on the good old io:format() function. The reason I haven't been keen on using the Erlang debugger is that it's not integrated into my editor (I use Emacs). To debug a file, I'd have to open it separately in the debugger UI, where I could set breakpoints and step through the code. For simple bugs, this is more effort than adding a few io:format() statements, reloading the page, and figuring out the bug from the output.

I'm not an IDE fanatic (well, duh, I use Emacs) -- I can do fine without most of the features in advanced IDEs -- but I do expect any decent editor to have at least 3 features: syntax highlighting, auto-indentation, and integrated debugging.

(I'm ambivalent about IDEs with code completion. It adds convenience, but it also it lets you get by without really remembering your APIs. Without code completion, I remember most of the my application's function names and their parameters simply because not remembering them slows me down too much. This is akin to the cellphone effect on remembering numbers: before I had a cellphone, I remembered all my close friends and family members' numbers. Looking them up in an address book was too much work. Now, I remember just a fraction of them. My cellphone made me lazy.)

The Emacs mode for Erlang provides syntax highlighting and auto-indentation, but no integrated debugging. AFAIK, neither Erlide nor ErlyBird have integrated debuggers. I knew Distel was supposed to add debugging support, and I even tried setting it up a while ago but gave up after running into problems I couldn't figure out. Last weekend, though, I finally got it working after following Bill Clemenson's Distel tutorial. I must say that having a debugger integrated into Emacs is a huge improvement. I recommend it to every Erlang hacker.

My only peeve with this debugger is that you must mark each file you want to debug as interpreted (the Emacs shortcut is "C-c C-d i") before you can set breakpoints in it and step into its functions. This seems unnecessary -- Erlang .beam files usually contain metadata including the location of their source files. I wish the debugger used this information to auto-interpret all the files you're trying to debug.

This is a relatively small peeve, though. The Distel debugger is definitely a step up from io:format().

I'm still hoping someone will create a more user-friendly Erlang IDE with a slick UI and an integrated debugger. I like the TextMate interface, but TextMate doesn't do auto-indentation for Erlang or debugging. I'll stick to Emacs+Distel for now.

Wednesday, January 23, 2008

Triggit: Why Didn't I think of That?

Triggit received some great publicity last week. Triggit, a small startup in San Francisco, recently released a product makes it easy to add widgets and ads to any site by dragging and dropping -- no HTML editing required. I think it can really catch on among bloggers who are disinclined to muck around in mounds of PHP code in order to add some widgets to their blogs. I can also see this kind of technology fitting very well in sites like MySpace, where non tech-savvy users could use Triggit to effortlessly soup up their profiles.

Ryan Tecco, Triggit's CTO, told me Triggit uses Erlang for all the heavy lifting in the backend. You probably won't find this shocking, but I think using Erlang is a smart choice. Low latency, scalability and high availability are all essential for this kind of product: it's one thing to have performance/availability issues on *your* site, but it's much worse to let those issues affect other people's sites as well.

Finally, Triggit is an interesting data point on how very small teams can successfully use Erlang to achieve impressive results in a short time frame.

Update: I added this video with Triggit :)

Monday, January 21, 2008

How to Use Concurrency to Improve Response Time in ErlyWeb Facebook Apps

When I was building the Vimagi Facebook app, I came across a common scenario where using concurrency can make your application more responsive for its users.

The typical flow of responding to requests coming from Facebook looks like this:

1) request arrives
2) do some stuff (mostly DB CRUD operations)
3) call Facebook API to send notifications / update newsfeeds and profile FBML
4) send response

When you're building a facebook app with ErlyWeb, you can instead do the following:

1) request arrives
2) do some stuff (mostly DB CRUD operations)

spawn(fun() ->
  3) call Facebook API to send notifications / update newsfeeds and profile FBML
end)

4) send response


The Facebook API calls in step 3 are much more expensive than the typical ErlyWeb controller operations because these calls involve synchronous round trips to the Facebook servers plus XML processing for the responses. By performing the Facebook API calls in a new process we can return the rendered response to the browser immediately and let the Erlang VM schedule the Facebook API calls to happen leisurely in the background.

The only gotcha is that if an error occurs in the spawned process we can't notify the user right away -- but this isn't really problem because we can log the errors and retry later, which is arguably a better approach anyway from a usability standpoint.

It's really that easy! A simple call to spawn() makes our app *feel* much faster. This puts the debate around language performance comparisons in a new light: how do you take into account the observation that some languages make "cheating" so much easier? :)

Sunday, January 20, 2008

Vimagi: The Facebook App

I finally finished the Vimagi Facebook app! Check out the screen shots:

vimagi-fb1.png

vimagi-fb2.png

The app works similar to the vimagi.com website: you can create paintings with titles, tags, and descriptions and share them with anyone. Other Facebook users can comment on the paintings and rate them. Paintings created in Facebook are automatically posted to vimagi.com, where vimagi.com users can also add comments and ratings. All paintings have embed codes you can use to embed them on any site (with playback). In Facebook, you can create paintings for your Facebook friends and those paintings will appear on their (and your) profile. Similar to vimagi.com, the Facebook app has a gallery and a tags page, which contains content from both vimagi.com and the Facebook app.

I created the Facebook app mostly as a learning exercise, but I also believe that the Vimagi features would appeal to more people by being embedded into people's their existing social network rather than in a standalone site.

Unfortunately, I underestimated how tricky it would be to port the vimagi.com features and code into Facebook without breaking the site and while seamlessly blending the paintings and data created on Facebook and on vimagi.com. (One of the pesky issues I encountered is that Facebook applications have root URLs of the form http://apps.facebook.com/[app name], which was incompatible with the vimagi.com relative URLs. The existing URLs all start with a forward slash, assuming they follow immediately after the domain name. My life would have been much better if Facebook used the url scheme "http://[app name].apps.facebook.com/" for Facebook apps as it would have allowed existing URLs to remain unmodified.)

If you like to paint, I hope you enjoy the applications, and if you have friends who like to paint, please send them an invitation. Facebook needs some more Erlang-generated pageviews :)

If you like Vimagi and you're on Facebook, please join the Vimagi fan club. Also, I'll appreciate it if you let me know of any feedback or suggestions you may have.

P.S. Thanks I to Bryan Fink for the excellent erlang2facebook library.

Thursday, January 17, 2008