1.6.1 The "Hello, world" Program

Let's look at a little program that actually does something. Here is your basic "Hello, world" program (use any text editor to type it in):

print ("Hello, world!\n");

This single line is the entire program. The built-in function print starts it off, and in this case has just one argument, a C-like text string. Within this string, the character combination \n stands for a newline character, just as it does in C. The print statement is terminated by a semicolon (;). As in Pascal or C, all simple statements in Perl are terminated by a semicolon.[7]

[7] The semicolon can be omitted when the statement is the last statement of a block, file, or eval.

When you invoke this program, the Perl interpreter parses the entire program and then executes the compiled form. The first and only operation is the execution of the print function, which sends any arguments to the standard output. After the program has completed, the Perl process exits, returning a successful exit code to the parent process.

Soon, you'll see Perl programs in which print and other functions are sometimes called with parentheses, and sometimes called without them. The rule is simple: in Perl, parentheses for built-in functions are never required nor forbidden. Their use can help or hinder clarity, so use your own judgment.

1.6.2 Asking Questions and Remembering the Result

Let's add a bit more sophistication. The Hello, world greeting is a touch cold and inflexible. Let's have the program call you by your name. To do this, we need a place to hold the name, a way to ask for the name, and a way to get a response.

One kind of place to hold values (like a name) is a scalar variable. For this program, we'll use the scalar variable $name to hold your name. In Chapter 2, Scalar Data, we'll go into more detail about what these variables can hold, and what you can do with them. For now, assume that you can hold a single number or string (sequence of characters) in a scalar variable.

The program needs to ask for the name. To do that, we need a way to prompt and a way to accept input. The previous program showed us how to prompt: use the print function. And the way to get a line from the terminal is with the <STDIN> construct, which (as we're using it here) grabs one line of input. We assign this input to the $name variable. This gives us the following program:

print "What is your name? ";
$name = <STDIN>;

The value of $name at this point has a terminating newline (Erik comes in as Erik\n). To get rid of the newline, we use the chomp() function, which takes a scalar variable as its sole argument and removes the trailing newline, if present, from the string:

chomp $name;

Now, all we need to do is say Hello, followed by the value of the $name variable, which we can do by embedding the variable inside the quoted string:

print "Hello, $name!\n";

Putting it all together, we get:

print "What is your name? ";
$name = <STDIN>;
chomp $name;
print "Hello, $name!\n";

1.6.3 Adding Choices

Now, let's say we have a special greeting for Erik, but want an ordinary greeting for anyone else. To do this, we need to compare the name that was entered with the string Erik, and if they are identical, do something special. Let's add a C-like if-then-else branch and a comparison to the program:

print "What is your name? ";
$name = <STDIN>;
chomp $name;
if ($name eq "Erik") {
  print "Hello, Erik! How good of you to be here!\n";
} else {
  print "Hello, $name!\n"; # ordinary greeting
}

The eq operator compares two strings. If they are equal (character for character, and of the same length), the result is true. (No comparable operator[8] exists in C or C++.)

[8] Well, OK, there's a standard C library function. But that's not an operator.

The if statement selects which block of statements (between matching curly braces) is executed - if the expression is true, it's the first block, otherwise it's the second block.

1.6.4 Guessing the Secret Word

Well, now that we have the name, let's have the person running the program guess a secret word. For everyone except Erik, we'll have the program repeatedly ask for guesses until the person guesses properly. First the program, and then an explanation:

$secretword = "gecko"; # the secret word
print "What is your name? ";
$name = <STDIN>;
chomp $name;
if ($name eq "Erik") {
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    print "What is the secret word? ";
    $guess = <STDIN>;
    chomp $guess;
    while ($guess ne $secretword) {
        print "Wrong, try again. What is the secret word? ";
        $guess = <STDIN>;
        chomp $guess;
    }
}

First, we define the secret word by putting it into another scalar variable, $secretword. After the greeting, the (non-Erik) person is asked (with another print) for the guess. The guess is compared with the secret word using the ne operator, which returns true if the strings are not equal (ne is the logical opposite of the eq operator). The result of the comparison controls a while loop, which executes the block as long as the comparison is true.

Of course, this program is not very secure, because anyone who is tired of guessing can merely interrupt the program and get back to the prompt, or even look at the source to determine the word. But, we weren't trying to write a security system, just an example for this book.

1.6.5 More than One Secret Word

Let's see how we can modify this program to allow more than one valid secret word. Using what we've already seen, we could compare the guess repeatedly against a series of good answers stored in separate scalar variables. However, such a list would be hard to modify or read in from a file or compute based on the day of the week.

A better solution is to store all of the possible answers in a data structure called a list, or (preferrably) an array. Each element of the array is a separate scalar variable that can be independently set or accessed. The entire array can also be given a value in one fell swoop. We can assign a value to the entire array named @words so that it contains three possible good passwords:

@words = ("camel","gecko","alpaca");

Array variable names begin with @, so they are distinct from scalar variable names. Another way to write this so that we don't have to put all those quotemarks there is with the qw() syntax, like so:

@words = qw(camel gecko alpaca);

These mean exactly the same thing; the qw makes it as if we had quoted each of three strings.

After the array is assigned, we can access each element by using a subscript reference (subscripts start at zero). So, $words[0] is camel, $words[1] is gecko, and $words[2] is alpaca. The subscript can be an expression as well, so if we set $i to 2, then $words[$i] is alpaca. (Subscript references start with $ rather than @, because they refer to a single element of the array rather than the whole array.) Going back to our previous example:

@words = qw(camel gecko alpaca);
print "What is your name? ";
$name = <STDIN>;
chomp $name;
if ($name eq "Erik") {
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    print "What is the secret word? ";
    $guess = <STDIN>;
    chomp ($guess);
    $i = 0; # try this word first
    $correct = "maybe"; # is the guess correct or not?
    while ($correct eq "maybe") { # keep checking til we know
        if ($words[$i] eq $guess) { # right?
        $correct = "yes"; # yes!
    } elsif ($i < 2) { # more words to look at?
        $i = $i + 1; # look at the next word next time
    } else { # no more words, must be bad
        print "Wrong, try again. What is the secret word?";
        $guess = <STDIN>;
        chomp ($guess);
        $i = 0; # start checking at the first word again
    }
  } # end of while not correct
} # end of "not Erik"

You'll notice we're using the scalar variable $correct to indicate that we are either still looking for a good password, or that we've found one.

This program also shows the elsif block of the if-then-else statement. This exact construct is not present in all programming languages - it's an abbreviation of the else block together with a new if condition, but it does not nest inside yet another pair of curly braces. It's a very Perl-like thing to compare a set of conditions in a cascaded if-elsif-elsif-elsif-else chain. Perl doesn't really have the equivalent of C's switch or Pascal's case statement, although you can build one yourself without too much trouble. See Chapter 2 of Programming Perl or the perlsyn documentation for details.

1.6.6 Giving Each Person a Different Secret Word

In the previous program, any person who comes along could guess any of the three words and be successful. If we want the secret word to be different for each person, we'll need a table that matches people with words. Table 1.1 does just this.

Notice that both Betty and Wilma have the same secret word. This is fine.

The easiest way to store such a table in Perl is with a hash. Each element of the hash holds a separate scalar value (just like the other type of array), but each hash is referenced by a key, which can be any scalar value (any string or number, including noninteger and negative values). To create a hash called %words (notice the use of %, rather than @) with the keys and values given in Table 1.1, we assign a value to %words (much as we did earlier with the array):

%words = qw(
  fred    camel
  barney  gecko
  betty   alpaca
  wilma   alpaca
);

Each pair of values in the list represents one key and its corresponding value in the hash. Note that we broke this assignment over many lines without any sort of line continuation character. We could do so because whitespace is generally insignificant in a Perl program.

To find the secret word for Betty, we need to use Betty as the key in a reference to the hash %words, via some expression such as $words{"betty"}. The value of this reference is alpaca, similar to what we had before with the other array. Also, as before, the key can be any expression, so setting $person to betty and evaluating $words{$person} gives alpaca as well.

Putting all this together, we get a program like this:

%words = qw(
    fred      camel
    barney    gecko
    betty     alpaca
    wilma     alpaca
);
print "What is your name? ";
$name = <STDIN>;
chomp ($name);
if ($name eq "Erik") {
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    $secretword = $words{$name}; # get the secret word
    print "What is the secret word? ";
    $guess = <STDIN>;
    chomp ($guess);
    while ($guess ne $secretword) {
        print "Wrong, try again. What is the secret word? ";
        $guess = <STDIN>;
        chomp ($guess);
    }
}

Note the lookup of the secret word. If the name is not found, the value of $secretword will be an empty string,[9] which we can then check for if we want to define a default secret word for everyone else. Here's how that process looks:

[9] Well, OK, the value is really the undef value, but it looks like an empty string to the eq operator. You'd get a warning about this value if you used -w on the command line, which is why we omitted it here.

[... rest of program deleted ...]
  $secretword = $words{$name}; # get the secret word
  if ($secretword eq "") { # oops, not found
      $secretword = "groucho"; # sure, why a duck?
  }
  print "What is the secret word? ";
[... rest of program deleted ...]

1.6.7 Handling Varying Input Formats

If we enter Erik Olson or erik rather than Erik, we're lumped in with the rest of the users, because the eq comparison requires an exact equality. Let's look at one way to handle that.

Suppose we wanted to look for any string that began with Erik, rather than just a string that was equal to Erik. We could do this with a regular expression: a template that defines a collection of strings that match. The regular expression in Perl that matches any string that begins with Erik is ^Erik. To match this against the string in $name, we use the match operator as follows:

if ($name =~ /^Erik/) {
    ## yes, it matches
} else {
    ## no, it doesn't
}

Note that the regular expression is delimited by slashes. Within the slashes, spaces and other whitespace are significant, just as they are within strings.

This addition almost meets our needs, but it doesn't handle selecting erik or rejecting eriko. To accept erik, we add the ignore-case option, a small i appended after the closing slash. To reject eriko, we add a word boundary special marker in the form of \b in the regular expression. This ensures that the character following the first k in the regular expression is not another letter. The addition also changes the regular expression to be /^erik\b/i, which means "erik at the beginning of the string, no letter or digit following, and OK to be in either case."

When this is added to the rest of the program, the final version looks like this:

%words = qw(
    fred    camel
    barney  gecko
    betty   alpaca
    wilma   alpaca
);
print "What is your name? ";
$name = <STDIN>;
chomp ($name);
if ($name =~ /^erik\b/i) {
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    $secretword = $words{$name}; # get the secret word
    if ($secretword eq "") { # oops, not found
        $secretword = "groucho"; # sure, why a duck?
  }
  print "What is the secret word? ";
  $guess = <STDIN>;
  chomp ($guess);
  while ($guess ne $secretword) {
      print "Wrong, try again. What is the secret word? ";
      $guess = <STDIN>;
      chomp ($guess);
  }
}

As you can see, the program is a far cry from the simple Hello world, but it's still very small and workable, and does quite a bit for being so short. This is The Perl Way.

Perl provides nearly every regular expression feature imaginable. In addition, the way Perl handles string matching is about the fastest on the planet, so you don't lose performance. String matching in Perl often compares favorably to hand-coded C programs written specifically for the same purpose.

1.6.8 Making It Fair for the Rest

So, now we can enter Erik or erik or Erik Olson, but what about everyone else? Barney still has to say exactly barney (not even barney followed by a space).

To be fair to Barney, we need to grab the first word of whatever's entered, and then convert it to lowercase before we look up the name in the table. We do this with two operators: the substitute operator, which finds a regular expression and replaces it with a string, and the translate operator, which puts the string in lowercase.

First, we discuss the substitute operator. We want to take the contents of $name, find the first nonword character, and zap everything from there to the end of the string. /\W.*/ is the regular expression we are looking for - the \W stands for a nonword character (something besides a letter, digit, or underscore), and .* represents any characters from that point to the end of the line. Now, to zap these characters, we need to take whatever part of the string matches this regular expression and replace it with nothing:

$name =~ s/\W.*//;

We're using the same =~ operator that we did before, but now on the right we have a substitute operator: the letter s followed by a slash-delimited regular expression and string. (The string in this example is the empty string between the second and third slashes.) This operator looks and acts very much like the substitution command of various editors.

Now, to get whatever's left into lowercase, we translate the string using the tr operator.[10] This operation takes a list of characters to find, and another list of characters with which to replace them. For our example, to put the contents of $name in lowercase, we use:

[10] This method doesn't work for characters with accent marks, although the uc function would.

$name =~ tr/A-Z/a-z/;

The slashes delimit the searched-for and replacement character lists. The dash between A and Z stands for all the characters in between, so we have two lists that each contain 26 characters. When the tr operator finds a character from the string in the first list, the character is then replaced with the corresponding character in the second list. So, all uppercase A's become lowercase a's, and so on.[11]

[11] Experts will note that we could have also constructed something like s/(\S*).*/\L$1/ to do this processing in one fell swoop, but experts probably won't be reading this section.

Putting everything together results in the following:

%words = qw(
    fred    camel
    barney  gecko
    betty   alpaca
    wilma   alpaca
);
print "What is your name? ";
$name = <STDIN>;
chomp ($name);
$original_name = $name; #save for greeting
$name =~ s/\W.*//; # get rid of everything after first word
$name =~ tr/A-Z/a-z/; # lowercase everything
if ($name eq "erik") { # ok to compare this way now
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $original_name!\n"; # ordinary greeting
    $secretword = $words{$name}; # get the secret word
    if ($secretword eq "") { # oops, not found
        $secretword = "groucho"; # sure, why a duck?
    }
    print "What is the secret word? ";
    $guess = <STDIN>;
    chomp ($guess);
    while ($guess ne $secretword) {
        print "Wrong, try again. What is the secret word? ";
        $guess = <STDIN>;
        chomp ($guess);
    }
}

Notice how the regular expression match for Erik became a simple comparison again. After all, both Erik Olson and Erik become erik after the substitution and translation. And everyone else gets a fair ride, because Fred and Fred Flintstone both become fred, Barney Rubble and Barney, the little guy become barney, and so on.

With just a few statements, we've made the program much more user friendly. You'll find that expressing complicated string manipulation with a few keystrokes is one of Perl's many strong points.

However, hacking away at the name so that we could compare it and look it up in the table destroyed the name that was entered. So, before the program hacks on the name, it saves it in $original_name. (Like C symbols, Perl variable names consist of letters, digits, and underscores and can be of nearly unlimited length.) We can then make references to $original_name later.

Perl has many ways to monitor and mangle strings. You'll find out about most of them in Chapter 7, Regular Expressions, and Chapter 15, Other Data Transformation.

1.6.9 Making It a Bit More Modular

Now that we've added so much to the code, we have to scan through many detailed lines in order to get the overall flow of the program. What we need is to separate the high-level logic (asking for a name, looping based on entered secret words) from the details (comparing a secret word to a known good word). We might do this for clarity, or maybe because one person is writing the high-level part and another is writing (or has already written) the detailed parts.

Perl provides subroutines that have parameters and return values. A subroutine is defined once in a program, and can be invoked repeatedly from within any expression.

For our small-but-rapidly-growing program, let's create a subroutine called good_word that takes a name and a guessed word, and returns true if the word is correct, and false if not. The definition of such a subroutine looks like this:

sub good_word {
    my($somename,$someguess) = @_; # name the parameters
    $somename =~ s/\W.*//; # get rid of everything after first word
    $somename =~ tr/A-Z/a-z/; # lowercase everything
    if ($somename eq "erik") { # should not need to guess
        return 1; # return value is true
  } elsif (($words{$somename} || "groucho") eq $someguess) {
        return 1; # return value is true
  } else {
        return 0; # return value is false
  }
}

First, the definition of a subroutine consists of the reserved word sub, followed by the subroutine name, followed by a block of code (delimited by curly braces). These definitions can go anywhere in the program file, but most people put them at the end.

The first line within this particular definition is an assignment that copies the values of the two parameters of this subroutine into two local variables named $somename and $someguess. (The my() defines the two variables as private to the enclosing block - in this case, the entire subroutine - and the parameters are initially in a special local array called @_.)

The next two lines clean up the name, just like in the previous version of the program.

The if-elsif-else statement decides whether the guessed word ($someguess) is correct for the name ($somename). Erik should not make it into this subroutine, but even if it does, whatever word was guessed is OK.

A return statement can be used to make the subroutine immediately return to its caller with the supplied value. In the absence of an explicit return statement, the last expression evaluated in a subroutine is the return value. We'll see how the return value is used after we finish describing the subroutine definition.

The test for the elsif part looks a little complicated - let's break it apart:

($words{$somename} || "groucho") eq $someguess

The first thing inside the parentheses is our familiar hash lookup, yielding some value from %words based on a key of $somename. The operator between that value and the string groucho is the || (logical-or) operator similar to that used in C. If the lookup from the hash has a value (meaning that the key $somename was in the hash), the value of the expression is that value. If the key could not be found, the string groucho is used instead. This step is a very Perl-like thing to do - specify some expression, and then provide a default value using || in case the expression turns out to be false.

In any case, whether it's a value from the hash, or the default value groucho, we compare it to whatever was guessed. If the comparison is true, we return 1; otherwise, we return 0.

So, expressed as a rule, if the name is erik, or the guess matches the lookup in %words based on the name (with a default of groucho if not found), then the subroutine returns 1; otherwise, it returns 0.

Now, let's integrate all these additions with the rest of the program:

%words = qw(
    fred    camel
    barney  gecko
    betty   alpaca
    wilma   alpaca
);
print "What is your name? ";
$name = <STDIN>;
chomp ($name);
if ($name =~ /^erik\b/i) { # back to the other way :-)
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    print "What is the secret word? ";
    $guess = <STDIN>;
    chomp ($guess);
  while (! good_word($name,$guess)) {
    print "Wrong, try again. What is the secret word? ";
    $guess = <STDIN>;
    chomp $guess;
  }
}
[... insert definition of good_word() here ...]

Notice that we've gone back to the regular expression to check for Erik, because now the main program does not have to pull apart the first name and convert it to lowercase.

The big difference is the while loop containing good_word. Here, we see an invocation of the subroutine passing two parameters, $name and $guess. Within the subroutine, the value of $somename is set from the first parameter, in this case $name. Likewise, $someguess is set from the second parameter, $guess.

The value returned by the subroutine (either 1 or 0, recalling the definition given earlier) is logically inverted with the prefix ! (logical not) operator. As in C, this operator returns true if the expression following is false, and vice versa. The result of this negation controls the while loop. You can read this as "while it's not a good word..." Many well-written Perl programs read very much like English, provided you take a few liberties with either Perl or English. (But you certainly won't win a Pulitzer that way.)

Note that the subroutine assumes that the value of the %words hash is set by the main program.

Such a cavalier approach to global variables doesn't scale very well, of course. Generally speaking, variables not created with my are global to the whole program, while those my creates last only until the block in which they were declared exits. Don't worry; Perl does in fact support a rich variety of other kinds of variables, including those private to a file (or package), as well as variables private to a function that retain their values between invocations (which is what we could really use here). However, at this stage in your Perl education, explaining these variables would only complicate your life. When you're ready for such information, check out what Programming Perl has to say about scoping, subroutines, modules, and objects. Or, see the online documentation in the perlsub , perlmod , perlobj , and perltoot documentation.

1.6.10 Moving the Secret Word List into a Separate File

Suppose we wanted to share the secret word list among three programs. If we store the word list as we have done already, we will need to change all three programs when Betty decides that her secret word should be swine rather than alpaca. This change can get to be a hassle, especially considering how often Betty is likely to change her mind.

So, let's put the word list into a file, and then read the file to get the word list into the program. To do so, we need to create an I/O channel called a filehandle. Your Perl program automatically gets three filehandles called STDIN, STDOUT, and STDERR, corresponding to the three standard I/O channels in many programming environments. We've already been using the STDIN handle to read data from the person running the program. Now, we just have to get another handle attached to a file of our own choice.

Here's a small chunk of code to do that:

sub init_words {
    open (WORDSLIST, "wordslist");
    while (defined ($name = <WORDSLIST>)) {
        chomp ($name);
        $word = <WORDSLIST>;
        chomp ($word);
        $words{$name} = $word;
     }
     close (WORDSLIST);
}

We're putting this code into a subroutine so that we can keep the main part of the program uncluttered. This organization also means that at a later time (hint: after a few more revisions in this stroll), we can change where the word list is stored, or even the format of the list.

The arbitrarily chosen format of the word list is one item per line, with names and words alternating. So, for our current database, we'd have something like this:

fred
camel
barney
gecko
betty
alpaca
wilma
alpaca

The open function creates a filehandle named WORDSLIST by associating it with a file named wordslist in the current directory. Note that the filehandle doesn't have a funny character in front of it as do the three variable types. Also, filehandles are generally uppercase - although they aren't required to be - for reasons detailed later.

The while loop reads lines from the wordslist file (via the WORDSLIST filehandle) one line at a time. Each line is stored into the $name variable. When end-of-file is reached, the value returned by the <WORDSLIST> operation is the empty string,[12] which looks false to the while loop, and terminates it. That's how we get out at the end.

[12] Well, technically the value is undef again, but empty string is close enough for this discussion.

If you were running with the -w option, you would have to check that the return value read in was actually defined. The empty string returned by the <WORDSLIST> operation isn't merely empty - it's undef again. The defined function is how you test for undef when this matters. In the case of reading lines from a file, you'd test as shown:

 while ( defined ($name = <WORDSLIST>) ) { 

But if you were being that careful, you'd probably also have checked to make sure that open returned a true value. You know, that's probably not a bad idea either. The built-in die function is frequently used to exit the program with an error message in case something goes wrong. We'll see an example of this function in the next revision of the program.

On the other hand, the normal case is that we've read a line (including the newline) into $name. First, off comes the newline using the chomp function. Then, we have to read the next line to get the secret word, holding it in the $word variable. This variable also gets the newline hacked off.

The final line of the while loop puts $word into %words with a key of $name, so that the rest of the program can access it later.

After the file has been read, the filehandle can be recycled with the close function. (Filehandles are automatically closed anyway when the program exits, but we're trying to be tidy. If we were really tidy, we'd even check for a true return value from close in case the disk partition which held the file went south, its network filesystem became unreachable, or a similar catastrophe occurred. Yes, these things really do happen. Murphy will always be with us.)

This subroutine definition can go after or before the other one. And, we invoke the subroutine instead of setting %words in the beginning of the program. Therefore, you could wrap up all of this as follows:

init_words();
print "What is your name? ";
$name = <STDIN>;
chomp ($name);
if ($name =~ /^erik\b/i) { # back to the other way :-)
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    print "What is the secret word? ";
    $guess = <STDIN>;
    chomp ($guess);
    while (! good_word($name,$guess)) {
        print "Wrong, try again. What is the secret word? ";
        $guess = <STDIN>;
        chomp ($guess);
    }
}
## subroutines from here down
sub init_words {
    open (WORDSLIST,"wordslist") || 
            die "can't open wordlist:$!";
    while (defined ($name = <WORDSLIST>)) {
        chomp ($name);
        $word = <WORDSLIST>;
        chomp ($word);
        $words{$name} = $word;
    }
    close (WORDSLIST);
}
sub good_word {
    my($somename,$someguess) = @_; # name the parameters
    $somename =~ s/\W.*//; # delete everything after first word
    $somename =~ tr/A-Z/a-z/; # lowercase everything
    if ($somename eq "erik") { # should not need to guess
        return 1; # return value is true
    } elsif (($words{$somename} || "groucho") eq $someguess) {
        return 1; # return value is true
    } else {
        return 0; # return value is false
    }
}

Now our program is starting to look full-grown. Notice the first executable line is an invocation of init_words(). The return value is not used in a further calculation, which is good because we didn't return anything remarkable. In this case, a true value is guaranteed (the value 1, in particular), because if the close had failed, the die would have printed a message to our STDERR error and exited the program. The die function is fully explained in Chapter 10, Filehandles and File Tests, but because the return values of anything that might fail must be checked, we'll get into the habit of using the function right from the start. The $! variable (also explained in Chapter 10) contains the system error message explaining why the system call failed.

The open function is also used to open files for output, or open programs as files (demonstrated shortly). The full scoop on open comes much later in this book, however, in Chapter 10.

1.6.11 Ensuring a Modest Amount of Security

"That secret word list has got to change at least once a week!" cries the Chief Director of Secret Word Lists. Well, we can't force the list to be different, but we can at least issue a warning if the secret word list has not been modified in more than a week.

The best place for handling this warning is the init_words() subroutine - we're already looking at the file there. The Perl operator -M returns the age in days since a file or filehandle has last been modified, so we just need to see whether this value is greater than seven for the WORDSLIST filehandle:

sub init_words {
    open (WORDSLIST,"wordslist") || 
            die "can't open wordlist:$!";
    if (-M WORDSLIST > 7.0) { # comply with bureaucratic policy
        die "Sorry, the wordslist is older than seven days.";
    }
   while (defined ($name = <WORDSLIST>)) {
       chomp ($name);
       $word = <WORDSLIST>;
      chomp ($word);
      $words{$name} = $word;
    }
    close (WORDSLIST) || die "couldn't close wordlist: $!";
}

The value of -M WORDSLIST is compared to seven, and if the value is greater, bingo, we've violated policy.

The rest of the program remains unchanged, so in the interest of saving a few trees, we won't repeat it here.

Besides getting the age of a file, we can also find out its size, access time, and everything else that an operating system maintains about a file. More information about this feature appears in Chapter 10.

1.6.12 Warning Someone When Things Go Astray

We really ought to know when someone guesses incorrectly so that we can watch for break-in attempts. If we were using a UNIX system, we would probably use the mail command to send an email message to someone about the failed attempt. However, on a Windows workstation, no standard mail [13] command exists, so we're going to log failures to a file.[14] We need only do a little work to accomplish this task. We'll add a new subroutine and modify only the good_word() subroutine (thanks to modularity), because we have all the information we need:

[13] Perl for Win32 programmers will encounter this mail comand issue frequently in scripts that they find on the Net. The solution is to use one of a number of readily available command-line mailers, or to use Perl's network interface to talk to an SMTP server directly.

[14] We could also use the Win32::EventLog module to log our warnings to the Windows NT Event Log.

sub good_word {
    my($somename,$someguess) = @_; # name the parameters
    $somename =~ s/\W.*//; # get rid of stuff after first word
    $somename =~ tr/A-Z/a-z/; # lowercase everything
    if ($somename eq "erik") { # should not need to guess
        return 1; # return value is true
    } elsif (($words{$somename}||"groucho") eq $someguess) {
        return 1; # return value is true
    } else {
        log_failure($somename, $someguess);
        return 0; # return value is false
    }
}

sub log_failure {
    my($somename,$someguess) = @_; # name the parameters
    open(LOG, ">>failures.log") || die "failures.log: $!";
    print LOG "bad news: $somename guessed $someguess\n";
    close (LOG)  || die "can't close failures.log: $!";
}

Notice the open, which has a redirection symbol (>>) in the filename. This symbol is a special indication that we are appending to a file. The next statement, a print, shows that a filehandle between the print keyword and the values to be printed selects that filehandle for output, rather than STDOUT.[15] This means that the message will be written to the output file that we've opened. Finally, we close the filehandle.

[15] Well, technically, the currently selected filehandle. That's covered much later, though.

1.6.13 Many Secret Word Files in the Current Directory

Let's change the definition of the secret word filename slightly. Instead of just the file named wordslist, let's look for anything in the current directory that ends in .sec. At the command prompt, we say:

> dir /B *.sec

to get a brief listing of all of these names. As you'll see in a moment, Perl uses a similar wildcard name syntax.

Pulling out the init_words() definition again:

sub init_words {
    while (defined ($filename = glob("*.sec")) ) {
        open (WORDSLIST, $filename) || 
                die "can't open $filename:$!";
        if (-M WORDSLIST <= 7.0) {
            while (defined ($name = <WORDSLIST>)) {
                chomp ($name);
                $word = <WORDSLIST>;
                chomp ($word);
                $words{$name} = $word;
             }
        }
    close (WORDSLIST) || die "couldn't close $filename: $!";
    }
}

First, I've wrapped a new while loop around the bulk of the routine from the previous version. The new thing here is the glob function. This is called a filename glob[16] for historical reasons. The function works much like <STDIN>, in that each time it is accessed, it returns the next value: successive filenames that match the pattern, in this case *.sec. When there are no additional filenames to be returned, the filename glob returns an empty string.[17] In Perl on Windows systems, filename globbing[18] is implemented by means of another program, called PerlGlob.exe, which must be somewhere in your search path (this should usually be the case, because PerlGlob.exe is installed in the same directory as your Perl interpreter by default).

[16] Glob might be a new word to Win32 programmers. We'll talk much more about globbing in Chapter 12, Directory Access.

[17] Yeah, yeah, undef again.

[18] If you're using the ISAPI version of Perl, you'll have better luck if you avoid file globbing altogether and use the following equivalent technique:

opendir(DIR, `.');
@files = grep(/\.sec$/, readdir(DIR));
closedir DIR;

This method leaves you with a list (@files) of all filenames in the current directory that contain the .sec pattern. We'll provide more information on this later.

So, if the current directory contains fred.sec and barney.sec, then $filename is barney.sec on the first pass through the while loop (the names come out in alphabetically sorted order). On the second pass, $filename is fred.sec. And there is no third pass because the glob returns an empty string the third time it is called, perceived by the while loop to be a false, causing an exit from the subroutine.

Within the while loop, we open the file and verify that it is recent enough (less than seven days since the last modification). For the recent-enough files, we scan through as before.

Note that if there are no files that match *.sec and are less than seven days old, the subroutine will exit without having set any secret words into the %words array. In such a case, everyone must use the word groucho. Oh well. (For real code, we would have added some check on the number of entries in %words before returning, and die'd if the check wasn't good. See the keys function when we get to hashes in Chapter 5, Hashes.)

1.6.14 Listing the Secret Words

Well, the Chief Director of Secret Word Lists wants a report of all the secret words currently in use, and how old they are. If we set aside the secret word program for a moment, we'll have time to write a reporting program for the Director.

First, let's get all of the secret words, by stealing some code from the init_words() subroutine:

while ( defined ($filename = glob("*.sec")) ) {
    open (WORDSLIST, $filename) || 
            die "can't open $filename: $!";
    if (-M WORDSLIST <= 7.0) {
        while (defined ($name = <WORDSLIST>) {
            chomp ($name);
            $word = <WORDSLIST>;
            chomp ($word);
            ### new stuff will go here
        }
    }
    close (WORDSLIST) || die "couldn't close $filename: $!";
}

At the point marked "new stuff will go here," we know three things: the name of the file ($filename), someone's name ($name), and that person's secret word ($word). Here's a place to use Perl's report generating tools. We define a format somewhere in the program (usually near the end, like a subroutine):

format STDOUT =
@<<<<<<<<<<<<<<< @<<<<<<<<< @<<<<<<<<<<<
$filename, $name, $word
.

The format definition begins with format STDOUT =, and ends with a single period. The other two lines comprise the format itself. The first line of this format is a field definition line that specifies the number, length, and type of the fields. For this format, we have three fields. The line following a field definition line is always a field value line. The value line gives a list of expressions that will be evaluated when this format is used, and the results of those expressions will be plugged into the fields defined in the previous line.

We invoke this format with the write function, as shown:

while ( defined($filename = glob("*.sec")) ) {
    open (WORDSLIST, $filename) || 
            die "can't open $filename: $!";
    if (-M WORDSLIST <= 7.0) {
        while (defined ($name = <WORDSLIST>) {
            chomp ($name);
            $word = <WORDSLIST>;
            chomp ($word);
            write; # invoke format STDOUT to STDOUT
        }
    }
    close (WORDSLIST) || die "couldn't close $filename: $!";
}
format STDOUT =
@<<<<<<<<<<<<<<< @<<<<<<<<< @<<<<<<<<<<<
$filename, $name, $word
.

When the format is invoked, Perl evaluates the field expressions and generates a line that it sends to the STDOUT filehandle. Because write is invoked once each time through the loop, we'll get a series of lines with text in columns, one line for each secret word entry.

Hmm. We haven't labeled the columns. That's easy enough. We just need to add a top-of-page format, as shown:

format STDOUT_TOP =
Page @<<
$%

Filename         Name       Word
================ ========== ============
.

This format is named STDOUT_TOP, and will be used initially at the first invocation of the STDOUT format, and again every time 60 lines of output to STDOUT have been generated. These column headings line up with the columns from the STDOUT format, so everything comes out tidy.

The first line of this format shows some constant text (Page) along with a three-character field definition. The following line is a field value line, which in this case has one expression. This expression is the $% variable,[19] which holds the number of pages printed - a very useful value in top-of-page formats.

[19] More mnemonic aliases for these predefined scalar variables are available via the English module, which provides English names for Perl's special variables.

The third line of the format is blank. Because this line does not contain any fields, the line following it is not a field value line. This blank line is copied directly to the output, creating a blank line between the page number and the column headers below.

The last two lines of the format also contain no fields, so they are copied as-is, directly to the output. So this format generates four lines, one of which has a part that changes from page to page.

Just tack this definition onto the previous program to get it to work. Perl notices the top-of-page format automatically.

Perl also has fields that are centered or right justified, and supports a filled paragraph area as well. More on these features when we get to formats in Chapter 11, Formats.

1.6.15 Making Those Old Word Lists More Noticeable

As we are scanning through the *.sec files in the current directory, we may find files that are too old. So far, we are simply skipping over those files. Let's go one step more - we'll rename them to *.sec.old so that a directory listing will quickly show us which files are too old, simply by name.

Here's how the init_words() subroutine looks with this modification:

sub init_words {
    while ( defined($filename = glob("*.sec")) ) {
        open (WORDSLIST, $filename) || 
                die "can't open $filename: $!";
        if (-M WORDSLIST <= 7.0) {
            while (defined ($name = <WORDSLIST>)) {
                chomp ($name);
                $word = <WORDSLIST>;
                chomp ($word);
                $words{$name} = $word;
            }
            close (WORDSLIST) ||
                    die "can't close $filename: $!";
        }
        else {
            # must close file before renaming it
            close (WORDSLIST) ||
                    die "can't close $filename.old: $!"
            rename ($filename,"$filename.old") ||
                    die "can't rename $filename: $!";
        }
    }
}

Notice the new else part of the file age check. If the file is older than seven days, it gets renamed with the rename function. This function takes two parameters, renaming the file named by the first parameter to the name given in the second parameter.

Perl has a complete range of file manipulation operators - nearly anything you can do to a file from a C program, you can also do from Perl.

1.6.16 Maintaining a Last-Good-Guess Database

Let's keep track of when the most recent correct guess has been made for each user. One data structure that might seem to work at first glance is a hash. For example, the statement:

$last_good{$name} = time;

assigns the current time in internal format (some large integer above 800 million, incrementing one number per second) to an element of %last_good that has the name for a key. Over time, this method would seem to give us a database indicating the most recent time the secret word was guessed properly for each of the users who had invoked the program.

But, the hash doesn't have an existence between invocations of the program. Each time the program is invoked, a new hash is formed, so at most, we create a one-element hash and then immediately forget it when the program exits.

The dbmopen function[20] maps a hash out into a disk file (actually a pair of disk files) known as a DBM. It's used like this:

[20] On a specific database, use the more low-level tie function, as detailed in Chapters 5 and 7 of Programming Perl, or in the perltie documentation.

dbmopen (%last_good,"lastdb",0666) || 
               die "can't dbmopen lastdb: $!";
$last_good{$name} = time;
dbmclose (%last_good) || die "can't dbmclose lastdb: $!";

The first statement performs the mapping, using the disk filenames of lastdb.dir and lastdb.pag (these names are the normal names for a DBM called lastdb). Showing Perl's UNIX heritage, dbmopen takes an octal file permission mask as the third argument. Although all Windows filesystems support file attributes of some sort, they are largely incompatible with the UNIX filesystem scheme used by Perl. The UNIX file permissions used for these two files, if the files must be created (as they will the first time through), is 0666. This mode means that anyone can read or write the files. This mode is usually the one that you want to use when working with files with read/write attributes.[21]

[21] The Win32::File module provides additional features for setting file attributes. We'll discuss those features in more detail in Chapter 13, File and Directory Manipulation.

The Perl file permission value is composed of a series of bits with read, write, and execute privileges for the user, the user's group, and everyone else. Traditionally, FAT filesystems only keep track of read and write privileges for the user, along with a few other tidbits of information, like whether the file is a hidden or a system file. We'll discuss file permissions and attributes in detail in Chapter 13. For now, just trust us that you want to use 0666 for creating DBM files.

The second statement shows that we use this mapped hash just like a normal hash. However, creating or updating an element of the hash automatically updates the disk files that form the DBM. And, when the hash is later accessed, the values within the hash come directly from the disk image. This gives the hash a life beyond the current invocation of the program - a persistence of its own.

The third statement disconnects the hash from the DBM, much like a file close operation.

You can insert these three statements just ahead of the subroutine definitions.

Although the inserted statements maintain the database adequately (and even create the database initially), we don't have any way of examining the information yet. To do so, we can create a separate little program that looks something like this:

dbmopen (%last_good,"lastdb",0666) || 
                 die "can't dbmopen lastdb: $!";
foreach $name (sort keys %last_good) {
    $when = $last_good{$name};
     $hours = (time - $when) / 3600; # compute hours ago
     write;
}

format STDOUT =
User @<<<<<<<<<<<: last correct guess was @<<< hours ago.
$name, $hours
.

We've got a few new operations here: a foreach loop, sorting a list, and getting the keys of an hash.

First, the keys function takes a hash name as an argument and returns a list of all the keys of that hash in some unspecified order. For the %words hash defined earlier, the result is something like fred, barney, betty, wilma, in some unspecified order. For the %last_good hash, the result will be a list of all users who have guessed their own secret word successfully.

The sort function sorts the list alphabetically (just like passing a text file through the sort command). This function makes sure that the list processed by the foreach statement is always in alphabetical order.

The Perl foreach statement takes a list of values and assigns each one in turn to a scalar variable (here, $name), executing the body of the loop (a block) once for each value. So, for five names in the %last_good list, we get five passes through the loop, with $name being a different value each time.

The body of the foreach loop loads up a couple of variables used within the STDOUT format, and then invokes the format. Note that we figure out the age of the entry by subtracting the stored system time (in the array) from the current time (as returned by time), and then divide that by 3600 (to convert seconds to hours).

Perl also provides easy ways to create and maintain text-oriented databases and fixed-length-record databases. These databases are described in Chapter 17, Database Manipulation.

1.6.17 The Final Programs

Here are the programs from this stroll in their final form so that you can play with them.

First, the "say hello" program:

init_words();
print "What is your name? ";
$name = <STDIN>;
chomp ($name);
if ($name =~ /^erik\b/i) { # back to the other way :-)
    print "Hello, Erik! How good of you to be here!\n";
} else {
    print "Hello, $name!\n"; # ordinary greeting
    print "What is the secret word? ";
    $guess = <STDIN>;
     chomp ($guess);
     while (! good_word($name,$guess)) {
        print "Wrong, try again. What is the secret word? ";
        $guess = <STDIN>;
        chomp ($guess);
     }
}
dbmopen (%last_good,"lastdb",0666) || 
                 die "can't dbmopen lastdb: $!";
$last_good{$name} = time;
dbmclose (%last_good) || die "can't dbmclose lastdb: $!";
sub init_words {
    while ( defined($filename = glob("*.sec")) ) {
        open (WORDSLIST, $filename) || 
                die "can't open $filename: $!";
        if (-M WORDSLIST <= 7.0) {
            while (defined ($name = <WORDSLIST>)) {
                chomp ($name);
                $word = <WORDSLIST>;
                chomp ($word);
                $words{$name} = $word;
            }
            close (WORDSLIST) ||
                    die "can't close $filename: $!";
        }
        else {
            # must close file before renaming it
            close (WORDSLIST) ||
                    die "can't close $filename.old: $!"
            rename ($filename,"$filename.old") ||
                    die "can't rename $filename: $!";
        }
    }
}
sub good_word {
  my($somename,$someguess) = @_; # name the parameters
  $somename =~ s/\W.*//; # delete everything after first word
  $somename =~ tr/A-Z/a-z/; # lowercase everything
  if ($somename eq "erik") { # should not need to guess
    return 1; # return value is true
  } elsif (($words{$somename} || "groucho") eq $someguess) {
    return 1; # return value is true
  } else {
    log_failure($somename,$someguess);
    return 0; # return value is false
  }
}

sub log_failure {
    my($somename,$someguess) = @_; # name the parameters
    open(LOG, ">>failures.log") || die "failures.log: $!";
    print LOG "bad news: $somename guessed $someguess\n";
    close (LOG)  || die "can't close failures.log: $!";
}

Next, we have the secret word lister:

while ( defined($filename = glob("*.sec")) ) {
    open (WORDSLIST, $filename) || 
               die "can't open $filename: $!";
    if (-M WORDSLIST <= 7.0) {
        while (defined ($name = <WORDSLIST>)) {
            chomp ($name);
            $word = <WORDSLIST>;
             chomp ($word);
              write; # invoke format STDOUT to STDOUT
        }
    }
    close (WORDSLIST) || die "can't close $filename: $!";
}

format STDOUT =
@<<<<<<<<<<<<<<< @<<<<<<<<< @<<<<<<<<<<<
$filename, $name, $word
.
format STDOUT_TOP =
Page @<<
$%

Filename         Name       Word
================ ========== ============
.

And finally, the last-time-a-word-was-used display program:

dbmopen (%last_good,"lastdb",0666) || 
                 die "can't dbmopen lastdb: $!";
foreach $name (sort keys %last_good) {
    $when = $last_good{$name};
    $hours = (time - $when) / 3600; # compute hours ago
    write;
}
dbmclose(%last_good) || die "can't dbmclose lastdb: $!";
format STDOUT =
User @<<<<<<<<<<<: last correct guess was @<<< hours ago.
$name, $hours
.

Together with the secret word lists (files named something.sec in the current directory) and the database, lastdb.dir and lastdb.pag, you'll have all you need.