Sunday, January 08, 2012

Pascal's Triangle in Haskell

Summary: I'm continually amazed how concise and powerful Haskell is, compared to mainstream languages. This post describes how to write Pascal's Triangle, and gives some of the advantages of using Haskell.

Often, when programming in Haskell, I feel like I'm cheating. As an example, I recently came across this article by William Shields, which suggests that prospective interview candidates be given simple programming tasks like generating Pascal's Triangle. William gives examples in Python, including some of the answers a typical candidate might give.

Pascal's Triangle in Haskell

William describes Pascal's Triangle as:

"The root element is 1. Every other element is the sum of the one or two above it (diagonally left and diagonally right)."

As a Haskell programmer, the obvious technique to use is induction. The first row of the triangle is [1], and each row can be computed from the previous row by adding the row shifted left, and the row shifted right:

next xs = zipWith (+) ([0] ++ xs) (xs ++ [0])
pascal = iterate next [1]

Here, we define next to take one row and produce the next row. We then use the iterate function to repeatedly apply next starting at the root element. The solution is short, and follows from the definition.

Laziness for Modularity

William originally posed three questions:

  • Print out the triangle to a specific row: print $ take 100 pascal

  • Return a given row of the triangle: pascal !! 50

  • Return a given element (by row and index) of the triangle: pascal !! 10 !! 5

Thanks to laziness, we can concisely answer all these questions in terms of the original pascal definition. In contrast, using a language such as Python, the best solution (Dynamic Programming from the original article) can only perform the first task.

Interview problems

The original article was not about the choice of programming language, but about choosing suitable questions for interviewing programmers. I agree with William that Pascal's Triangle is a suitable problem - it isn't a trick puzzle, it isn't an API knowledge quiz - it's about understanding how to program. Given how much easier the problem is to solve in Haskell, I wonder if using Haskell in a job interview should be considered cheating? ;-)

Hiding Win32 Windows

Summary: This post describes how to hide windows on the Windows operating system, by using the Win32 API from Haskell.

Imagine that whenever your computer restarts, it pops up a message box:

If you interact with this window, your computer will be destroyed. You can't kill the process, or dismiss the window harmlessly. (This scenario isn't hypothetical...) The solution is to hide the window, so it still exists but is out of the way of misplaced clicks. To hide the window, we first find it's OS handle, then we call some Win32 API functions.

Find the Window

To find the handle of the window, we use Spy++. Spy++ comes with Visual Studio, and is bundled in the Express edition (the free version) from 2010 onwards. Start Spy++, got to Search, Find Window, then use the finder tool to select the window in question. Check that the caption of the window matches what Spy++ reports:

The important information is the handle: 0004061E.

Hide the Window

To hide the window you need a programming language capable of making Win32 API calls. In the past I have used Word VBA as the host language, but Haskell is probably easier. Start GHCi, and type:

$ import System.Win32
$ import Graphics.Win32
$ showWindow (castUINTToPtr 0x0004061E) sW_HIDE

Replace 0x0004061E on the final line with 0xyour-handle. The final line should cause the window to be hidden, saving your computer from destruction.

Thanks: Thanks to Roman Leshchinskiy for reminding me that there were better solutions than just trying not to click the window. Thanks to the Win32 Haskell developers - the Win32 binding was a lot of work, which not many people ever see.