Work the Shell

Last month, we finally wrangled the issue of Aces being worth either one or 11 by recognizing the value of what I call lazy coding—trying to solve the specific situation rather than creating an elegant and sophisticated general solution. I've since talked with some programmers and found that many don't agree with my philosophy. Indeed, one chap basically said I was completely wrong, especially if the code's to be published—as this column obviously is—and that I should always be putting in maximal effort to present exemplary, elegant solutions to any problems encountered.

But what if, I countered, that would make a program two times longer? Or four times longer? He looked at me blankly and asked, “Since when does the length of a program matter when you're trying to solve a problem or model a particular behavior?” It's a fair question, but I have to think that's where all of our code bloat comes from—where no programs can be simple, efficient and streamlined, and everything does everything else, including word processors that can send e-mail, e-mail programs that can do rudimentary page layout, text editors that can convert ASCII to HTML, Web browsers that, well, seem to include just about every feature up to and including KitchenSink.app.

And, don't get me started on the Emacs versus vi philosophy either, okay? Let's just say that an editor that can replace your OS kernel might just be a wee bit over-implemented for the vast majority of users.

The real cost of all this complexity, however, isn't the length of the code. After all, gigs of disk space are dirt cheap nowadays, and even RAM seems to be not much more than a decent meal. The cost is in the complexity itself—in the fact that large, complex systems are harder to work with, harder to debug and harder to make bulletproof than simple systems. You need but look at the challenges Microsoft is facing as it tries to ship its new Windows OS, Vista, before spring 2010.

So, what does this all have to do with shell script programming? I suggest that this discussion is pivotal to all programming tasks, actually, and that if you can't figure out a simple and manageable solution within the world of shell scripting, it might well be time to move to a more sophisticated development environment. Early on in this column, I joked about eventually rewriting the popular game Doom as a shell script, but of course, it's the complexity that would kill us, even if we might well be able to hammer our square peg into the round hole of shell scripting.

Let's get back to our Blackjack game and see where we are vis-�vis complexity, length and functionality. My suspicion is that we're going to be able to wrap up this project by the end of the next column, because it's really all the mechanics that we can create as a script; the game itself is always doomed to be a primitive command-line interface, because, well, it's a shell script. Could we layer an interface with Tk or some other graphical toolkit? Sure, but then, would you really want the underlying code base to be a shell script?

Our code is now getting into the category of seriously long shell scripts, weighing in at 177 lines. That's actually beyond my usual cutoff of 150 lines maximum for a script, so we definitely need to wrap this up before it gets too much longer anyway.

The missing piece we'll deal with in this column is to test for a dealer blackjack and player blackjack situation. If you are dealt an Ace and a ten-point card (a ten or a face card) you have a blackjack. If the dealer has a blackjack, the game ends and the player loses. If the player gets a blackjack on the deal, the game ends and the dealer loses (without being able to get additional cards, because a 21-point hand that involves more than two cards does not tie a player's blackjack). If you have the extraordinary situation of both the player and dealer having a blackjack, it's a “push” or tie, just as it is in any other situation where both player and dealer have the same point value in their hands.

The key spot where we need to modify the code to test for blackjacks is where the hands are dealt, and we might as well deal both hands before adding the additional test. Here's the hand-dealing code:

player[1]=${newdeck[1]}
player[2]=${newdeck[3]}
nextplayercard=3        # player starts with two cards

dealer[1]=${newdeck[2]}
dealer[2]=${newdeck[4]}
nextdealercard=3        # dealer also has two cards

To see whether either hand is a blackjack quickly, we'll use the handValue function:

handValue ${newdeck[1]} ${newdeck[3]}

Recall that because you can't return values in functions, handValue simply sets the global variable $handvalue to the numeric value of the hand. This means the test is straightforward:

# test for dealer or player blackjack

handValue ${player[1]} ${player[2]}
playerhandvalue=$handvalue
handValue ${dealer[1]} ${dealer[2]}
dealerhandvalue=$handvalue

echo ""

if [ $playerhandvalue -eq 21 -a $dealerhandvalue -eq 21 ] ; then
  echo "Extraordinary! Both the dealer and player were dealt a blackjack!"
  echo "Unfortunately, this means you didn't win: it's a push (or tie)."
  echo ""
  exit 0
fi

if [ $dealerhandvalue -eq 21 ] ; then
  echo "Darn it! Dealer pulled a blackjack out of his hat. You lose."
  echo ""
  exit 0
elif [ $playerhandvalue -eq 21 ] ; then
  echo "NICE! You got a blackjack and won the game. Payout would be 3:2!"
  echo ""
  exit 0
fi

You can see that I simply create playerhandvalue and dealerhandvalue for these numeric tests, and then check whether both are 21, the dealer is 21 and the player is 21. That's all there is to it.

To test the new code, simply slip in either or both of the following lines right before the handValue call above:

player[1]=1 ; player[2]=11
dealer[1]=1 ; dealer[2]=12

Then you can, um, stack the deck and create the specific test situations you desire.

We've run out of space for this column, but all that's left is to implement the overall play logic. The game as it stands now shuffles the deck, deals cards for the player and dealer, lets the player add new cards (“hit” in Blackjack parlance) until he or she is satisfied or exceeds 21 points, and shows the dealer hand. Next column, we'll write the loop that lets the dealer play out its hand and determine who won the hand. No doubt we'll sneak in some interface features too, of course, but one way or the other, next column will be a wrap for this particular scripting project.

This means I'll be ready to tackle something new! If you have a particular scripting project you think could be an interesting study, please drop me an e-mail, and I'll consider it for a future column series. Otherwise, I've been eyeing the game Yahtzee pretty closely.