Programming

I'm porting some old C++ code to Java. How old? So old that it doesn't just assume a long is 32 bits, it actually checks to make sure and bails if not. That's 1997 or so, by my guess.

I can already hear the programmers in the audience asking, "Why not just interface with the library using JNI?" We decided to port the code because we must support Unix and 64-bit environments, and the library specifically fails on all our Unix machines. We do have the source, so we could modify it... but then we'd be forced to deliver and maintain it, which carries its own set of headaches. So I'm working on 20,000 SLOC.

(Anybody wondered why the site has been so quiet for so long? It's because I'm porting 20,000 SLOC of OLD C++ to Java.)

Most of the code has been pretty straightforward. There have been the usual problems with C++ unsigned types, which don't have any corresponding Java type, as well as a bit of pointer arithmetic confusion. When I get confused, I just write a quick test program and make sure the C++ does what I expect.

The thing that blew my mind today was the weird use of the "magic number" 0x80000000.

As I said before, I've had some trouble with PHP Facebook development.. This time, I couldn't get a user ID because the user came from a non-Facebook page.

An application's drawing page is called it's "canvas page". It's the starting point for the whole app. Unfortunately, when the user jumps there from a bookmark, or enters the URL into the browser directly, Facebook can't supply the user ID. The app can't do anything, because it doesn't know who the heck it's working for. The user must visit your app's canvas page from a Facebook page.

Since your canvas page is a Facebook page, and the user just visited it directly, you'd think that redirecting to the same page would do the job. And it does! But only if the user clicks on a link.

Then there's the additional problem of users who don't allow you to see their user ID. You have to ask them for permission first.

I put together a skeleton that takes care of all these issues for FBML pages (I don't think it'll work for iFrame pages). Read on for the code.

I'm still looking for the best UML editor. I've just moved to a new project, so it's time to start the cycle again. At least now I've got a reference of everything that failed in the past!

Today I tried Open ModelSphere, another Open Source editor, recommended by one of the commenters in a previous review. It didn't work out very well for me. Let's see what happened.

I know I've been neglecting my blogs lately. And my online forums. And FaceBook. I apologize.

I've been working on a neat little puzzle game. Not a "puzzle game" like Bejeweled or those other match-up toys. Something that actually requires a little bit of thought, like Sudoku or Griddlers, but more logic. I've got a prototype, but I'm not ready to release it just yet; I promise to let you know when it's ready.

What's taken me so long is the mathematical whirlwind I've immersed myself in. It's been delightful. (Yes, the Alpha Geek enjoys math.) For a rather minor aspect of the coding, I needed a way to sum any pair of 36 possible numbers and guarantee that the result was unique for that pair. (In mathematical terms, I needed a set such that a_i + a_j is distinct for all i < j.) I searched online for days before finding what I need: weak Sidon sets, also known as "well-spread sets" or S₂ sets.

It turns out that nobody has published any weak Sidon sequences of any appreciable length, though. I needed to write a program to generate them.

My first try turned up a 36-member set and a 100-member set in short order. But I wanted smaller numbers, so I programmed more exhaustive search methods. One of them ran for 36 hours before Kayla decided to log me out, destroying all my progress.

So I modified the program. It now finds all well-spread sequences of a particular length chosen from the set of integers smaller than a specified maximum. And it saves its progress, so if I'm ever interrupted again, I can simply resume. I can even run on multiple computers by generating new start states (although of course I'll have to quit one of the programs manually).

When I've got my first set of sequences, I'll publish my results and my program. I just thought you should know why you haven't heard much from me.

I'm still looking for a UML editor that fits my criteria, and I'm not stopping until I find one! Today I tried Dia, another Open Source editor recommended by Ubuntu's package manager.

Unfortunately, Dia is really a drawing program with UML objects, not a UML editor. Let's look at the details.