perlop - Perl operators and precedence

Operator precedence and associativity work in Perl more or less like they do in mathematics.

Operator precedence means some operators are evaluated before others. For example, in CW2 + 4 * 5, the multiplication has higher precedence so CW4 * 5 is evaluated first yielding CW2 + 20 == 22 and not CW6 * 5 == 30.

Operator associativity defines what happens if a sequence of the same operators is used one after another: whether the evaluator will evaluate the left operations first or the right. For example, in CW8 - 4 - 2, subtraction is left associative so Perl evaluates the expression left to right. CW8 - 4 is evaluated first making the expression CW4 - 2 == 2 and not CW8 - 2 == 6.

Perl operators have the following associativity and precedence, listed from highest precedence to lowest. Operators borrowed from C keep the same precedence relationship with each other, even where C's precedence is slightly screwy. (This makes learning Perl easier for C folks.) With very few exceptions, these all operate on scalar values only, not array values.

    left        terms and list operators (leftward)
    left        ->
    nonassoc    ++ --
    right       **
    right       ! ~ \ and unary + and -
    left        =~ !~
    left        * / % x
    left        + - .
    left        << >>
    nonassoc    named unary operators
    nonassoc    < > <= >= lt gt le ge
    nonassoc    == != <=> eq ne cmp
    left        &
    left        | ^
    left        &&
    left        ||
    nonassoc    ..  ...
    right       ?:
    right       = += -= *= etc.
    left        , =>
    nonassoc    list operators (rightward)
    right       not
    left        and
    left        or xor

In the following sections, these operators are covered in precedence order.

Many operators can be overloaded for objects. See overload.

A TERM has the highest precedence in Perl. They include variables, quote and quote-like operators, any expression in parentheses, and any function whose arguments are parenthesized. Actually, there aren't really functions in this sense, just list operators and unary operators behaving as functions because you put parentheses around the arguments. These are all documented in perlfunc.

If any list operator (print(), etc.) or any unary operator (chdir(), etc.) is followed by a left parenthesis as the next token, the operator and arguments within parentheses are taken to be of highest precedence, just like a normal function call.

In the absence of parentheses, the precedence of list operators such as CWprint, CWsort, or CWchmod is either very high or very low depending on whether you are looking at the left side or the right side of the operator. For example, in

    @ary = (1, 3, sort 4, 2);
    print @ary;         # prints 1324

the commas on the right of the sort are evaluated before the sort, but the commas on the left are evaluated after. In other words, list operators tend to gobble up all arguments that follow, and then act like a simple TERM with regard to the preceding expression. Be careful with parentheses:

    # These evaluate exit before doing the print:
    print($foo, exit);  # Obviously not what you want.
    print $foo, exit;   # Nor is this.

    # These do the print before evaluating exit:
    (print $foo), exit; # This is what you want.
    print($foo), exit;  # Or this.
    print ($foo), exit; # Or even this.

Also note that

    print ($foo & 255) + 1, "\n";

probably doesn't do what you expect at first glance. The parentheses enclose the argument list for CWprint which is evaluated (printing the result of CW$foo & 255). Then one is added to the return value of CWprint (usually 1). The result is something like this:

    1 + 1, "\n";    # Obviously not what you meant.

To do what you meant properly, you must write:

    print(($foo & 255) + 1, "\n");

See Named Unary Operators for more discussion of this.

Also parsed as terms are the CWdo {} and CWeval {} constructs, as well as subroutine and method calls, and the anonymous constructors CW[] and CW{}.

See also Quote and Quote-like Operators toward the end of this section, as well as I/O Operators.

"CW->" is an infix dereference operator, just as it is in C and . If the right side is either a CW[...], CW{...}, or a CW(...) subscript, then the left side must be either a hard or symbolic reference to an array, a hash, or a subroutine respectively. (Or technically speaking, a location capable of holding a hard reference, if it's an array or hash reference being used for assignment.) See perlreftut and perlref.

Otherwise, the right side is a method name or a simple scalar variable containing either the method name or a subroutine reference, and the left side must be either an object (a blessed reference) or a class name (that is, a package name). See perlobj.

++ and -- work as in C. That is, if placed before a variable, they increment or decrement the variable by one before returning the value, and if placed after, increment or decrement after returning the value.

    $i = 0;  $j = 0;
    print $i++;  # prints 0
    print ++$j;  # prints 1

Note that just as in C, Perl doesn't define when the variable is incremented or decremented. You just know it will be done sometime before or after the value is returned. This also means that modifying a variable twice in the same statement will lead to undefined behaviour. Avoid statements like:

    $i = $i ++;
    print ++ $i + $i ++;

Perl will not guarantee what the result of the above statements is.

The auto-increment operator has a little extra builtin magic to it. If you increment a variable that is numeric, or that has ever been used in a numeric context, you get a normal increment. If, however, the variable has been used in only string contexts since it was set, and has a value that is not the empty string and matches the pattern CW/^[a-zA-Z]*[0-9]*\z/, the increment is done as a string, preserving each character within its range, with carry:

    print ++($foo = '99');      # prints '100'
    print ++($foo = 'a0');      # prints 'a1'
    print ++($foo = 'Az');      # prints 'Ba'
    print ++($foo = 'zz');      # prints 'aaa'

CWundef is always treated as numeric, and in particular is changed to CW0 before incrementing (so that a post-increment of an undef value will return CW0 rather than CWundef).

The auto-decrement operator is not magical.

Binary ** is the exponentiation operator. It binds even more tightly than unary minus, so -2**4 is -(2**4), not (-2)**4. (This is implemented using C's pow(3) function, which actually works on doubles internally.)

Unary ! performs logical negation, i.e., not. See also CWnot for a lower precedence version of this.

Unary - performs arithmetic negation if the operand is numeric. If the operand is an identifier, a string consisting of a minus sign concatenated with the identifier is returned. Otherwise, if the string starts with a plus or minus, a string starting with the opposite sign is returned. One effect of these rules is that -bareword is equivalent to the string -bareword.

Unary ~ performs bitwise negation, i.e., 1's complement. For example, CW0666 & ~027 is 0640. (See also Integer Arithmetic and Bitwise String Operators.) Note that the width of the result is platform-dependent: ~0 is 32 bits wide on a 32-bit platform, but 64 bits wide on a 64-bit platform, so if you are expecting a certain bit width, remember to use the & operator to mask off the excess bits.

Unary + has no effect whatsoever, even on strings. It is useful syntactically for separating a function name from a parenthesized expression that would otherwise be interpreted as the complete list of function arguments. (See examples above under Terms and List Operators (Leftward).)

Unary \ creates a reference to whatever follows it. See perlreftut and perlref. Do not confuse this behavior with the behavior of backslash within a string, although both forms do convey the notion of protecting the next thing from interpolation.

Binary =~ binds a scalar expression to a pattern match. Certain operations search or modify the string CW$_ by default. This operator makes that kind of operation work on some other string. The right argument is a search pattern, substitution, or transliteration. The left argument is what is supposed to be searched, substituted, or transliterated instead of the default CW$_. When used in scalar context, the return value generally indicates the success of the operation. Behavior in list context depends on the particular operator. See Regexp Quote-Like Operators for details and perlretut for examples using these operators.

If the right argument is an expression rather than a search pattern, substitution, or transliteration, it is interpreted as a search pattern at run time.

Binary !~ is just like =~ except the return value is negated in the logical sense.

Binary * multiplies two numbers.

Binary / divides two numbers.

Binary % computes the modulus of two numbers. Given integer operands CW$a and CW$b: If CW$b is positive, then CW$a % $b is CW$a minus the largest multiple of CW$b that is not greater than CW$a. If CW$b is negative, then CW$a % $b is CW$a minus the smallest multiple of CW$b that is not less than CW$a (i.e. the result will be less than or equal to zero). Note that when CWuse integer is in scope, % gives you direct access to the modulus operator as implemented by your C compiler. This operator is not as well defined for negative operands, but it will execute faster.

Binary x is the repetition operator. In scalar context or if the left operand is not enclosed in parentheses, it returns a string consisting of the left operand repeated the number of times specified by the right operand. In list context, if the left operand is enclosed in parentheses, it repeats the list. If the right operand is zero or negative, it returns an empty string or an empty list, depending on the context.

    print '-' x 80;             # print row of dashes

    print "\t" x ($tab/8), ' ' x ($tab%8);      # tab over

    @ones = (1) x 80;           # a list of 80 1's
    @ones = (5) x @ones;        # set all elements to 5

Binary + returns the sum of two numbers.

Binary - returns the difference of two numbers.

Binary . concatenates two strings.

Binary << returns the value of its left argument shifted left by the number of bits specified by the right argument. Arguments should be integers. (See also Integer Arithmetic.)

Binary >> returns the value of its left argument shifted right by the number of bits specified by the right argument. Arguments should be integers. (See also Integer Arithmetic.)

Note that both << and >> in Perl are implemented directly using << and >> in C. If CWuse integer (see Integer Arithmetic) is in force then signed C integers are used, else unsigned C integers are used. Either way, the implementation isn't going to generate results larger than the size of the integer type Perl was built with (32 bits or 64 bits).

The result of overflowing the range of the integers is undefined because it is undefined also in C. In other words, using 32-bit integers, CW1 << 32 is undefined. Shifting by a negative number of bits is also undefined.

The various named unary operators are treated as functions with one argument, with optional parentheses.

If any list operator (print(), etc.) or any unary operator (chdir(), etc.) is followed by a left parenthesis as the next token, the operator and arguments within parentheses are taken to be of highest precedence, just like a normal function call. For example, because named unary operators are higher precedence than ||:

    chdir $foo    || die;       # (chdir $foo) || die
    chdir($foo)   || die;       # (chdir $foo) || die
    chdir ($foo)  || die;       # (chdir $foo) || die
    chdir +($foo) || die;       # (chdir $foo) || die

but, because * is higher precedence than named operators:

    chdir $foo * 20;    # chdir ($foo * 20)
    chdir($foo) * 20;   # (chdir $foo) * 20
    chdir ($foo) * 20;  # (chdir $foo) * 20
    chdir +($foo) * 20; # chdir ($foo * 20)

    rand 10 * 20;       # rand (10 * 20)
    rand(1) * 20;      # (rand 10) * 20
    rand (10) * 20;     # (rand 10) * 20
    rand +(1) * 20;    # rand (10 * 20)

Regarding precedence, the filetest operators, like CW-f, CW-M, etc. are treated like named unary operators, but they don't follow this functional parenthesis rule. That means, for example, that CW-f($file).".bak" is equivalent to CW-f "$file.bak".

See also Terms and List Operators (Leftward).

Binary < returns true if the left argument is numerically less than the right argument.

Binary > returns true if the left argument is numerically greater than the right argument.

Binary <= returns true if the left argument is numerically less than or equal to the right argument.

Binary >= returns true if the left argument is numerically greater than or equal to the right argument.

Binary lt returns true if the left argument is stringwise less than the right argument.

Binary gt returns true if the left argument is stringwise greater than the right argument.

Binary le returns true if the left argument is stringwise less than or equal to the right argument.

Binary ge returns true if the left argument is stringwise greater than or equal to the right argument.

Binary == returns true if the left argument is numerically equal to the right argument.

Binary != returns true if the left argument is numerically not equal to the right argument.

Binary <=> returns -1, 0, or 1 depending on whether the left argument is numerically less than, equal to, or greater than the right argument. If your platform supports NaNs (not-a-numbers) as numeric values, using them with <=> returns undef. NaN is not <, ==, >, <= or >= anything (even NaN), so those 5 return false. NaN != NaN returns true, as does NaN != anything else. If your platform doesn't support NaNs then NaN is just a string with numeric value 0.

    perl -le '$a = NaN; print "No NaN support here" if $a == $a'
    perl -le '$a = NaN; print "NaN support here" if $a != $a'

Binary eq returns true if the left argument is stringwise equal to the right argument.

Binary ne returns true if the left argument is stringwise not equal to the right argument.

Binary cmp returns -1, 0, or 1 depending on whether the left argument is stringwise less than, equal to, or greater than the right argument.

lt, le, ge, gt and cmp use the collation (sort) order specified by the current locale if CWuse locale is in effect. See perllocale.

Binary & returns its operands ANDed together bit by bit. (See also Integer Arithmetic and Bitwise String Operators.)

Note that & has lower priority than relational operators, so for example the brackets are essential in a test like

        print "Even\n" if ($x & 1) == 0;

Binary | returns its operands ORed together bit by bit. (See also Integer Arithmetic and Bitwise String Operators.)

Binary ^ returns its operands XORed together bit by bit. (See also Integer Arithmetic and Bitwise String Operators.)

Note that | and ^ have lower priority than relational operators, so for example the brackets are essential in a test like

        print "false\n" if (8 | 2) != 10;

Binary && performs a short-circuit logical AND operation. That is, if the left operand is false, the right operand is not even evaluated. Scalar or list context propagates down to the right operand if it is evaluated.

Binary || performs a short-circuit logical OR operation. That is, if the left operand is true, the right operand is not even evaluated. Scalar or list context propagates down to the right operand if it is evaluated.

The CW|| and CW&& operators return the last value evaluated (unlike C's CW|| and CW&&, which return 0 or 1). Thus, a reasonably portable way to find out the home directory might be:

    $home = $ENV{'HOME'} || $ENV{'LOGDIR'} ||
        (getpwuid($<))[7] || die "You're homeless!\n";

In particular, this means that you shouldn't use this for selecting between two aggregates for assignment:

    @a = @b || @c;              # this is wrong
    @a = scalar(@b) || @c;      # really meant this
    @a = @b ? @b : @c;          # this works fine, though

As more readable alternatives to CW&& and CW|| when used for control flow, Perl provides CWand and CWor operators (see below). The short-circuit behavior is identical. The precedence of and and or is much lower, however, so that you can safely use them after a list operator without the need for parentheses:

    unlink "alpha", "beta", "gamma"
            or gripe(), next LINE;

With the C-style operators that would have been written like this:

    unlink("alpha", "beta", "gamma")
            || (gripe(), next LINE);

Using or for assignment is unlikely to do what you want; see below.

Binary .. is the range operator, which is really two different operators depending on the context. In list context, it returns a list of values counting (up by ones) from the left value to the right value. If the left value is greater than the right value then it returns the empty list. The range operator is useful for writing CWforeach (1..10) loops and for doing slice operations on arrays. In the current implementation, no temporary array is created when the range operator is used as the expression in CWforeach loops, but older versions of Perl might burn a lot of memory when you write something like this:

    for (1 .. 1_000_000) {
        # code
    }

The range operator also works on strings, using the magical auto-increment, see below.

In scalar context, .. returns a boolean value. The operator is bistable, like a flip-flop, and emulates the line-range (comma) operator of sed, awk, and various editors. Each .. operator maintains its own boolean state. It is false as long as its left operand is false. Once the left operand is true, the range operator stays true until the right operand is true, AFTER which the range operator becomes false again. It doesn't become false till the next time the range operator is evaluated. It can test the right operand and become false on the same evaluation it became true (as in awk), but it still returns true once. If you don't want it to test the right operand till the next evaluation, as in sed, just use three dots (...) instead of two. In all other regards, ... behaves just like .. does.

The right operand is not evaluated while the operator is in the false state, and the left operand is not evaluated while the operator is in the true state. The precedence is a little lower than || and &&. The value returned is either the empty string for false, or a sequence number (beginning with 1) for true. The sequence number is reset for each range encountered. The final sequence number in a range has the string E0 appended to it, which doesn't affect its numeric value, but gives you something to search for if you want to exclude the endpoint. You can exclude the beginning point by waiting for the sequence number to be greater than 1.

If either operand of scalar .. is a constant expression, that operand is considered true if it is equal (CW==) to the current input line number (the CW$. variable).

To be pedantic, the comparison is actually CWint(EXPR) == int(EXPR), but that is only an issue if you use a floating point expression; when implicitly using CW$. as described in the previous paragraph, the comparison is CWint(EXPR) == int($.) which is only an issue when CW$. is set to a floating point value and you are not reading from a file. Furthermore, CW"span" .. "spat" or CW2.18 .. 3.14 will not do what you want in scalar context because each of the operands are evaluated using their integer representation.

Examples:

As a scalar operator:

    if (101 .. 200) { print; } # print 2nd hundred lines, short for
                               #   if ($. == 101 .. $. == 200) ...
    next line if (1 .. /^$/);  # skip header lines, short for
                               #   ... if ($. == 1 .. /^$/);
    s/^/> / if (/^$/ .. eof()); # quote body

    # parse mail messages
    while (<>) {
        $in_header =   1  .. /^$/;
        $in_body   = /^$/ .. eof;
        if ($in_header) {
            # ...
        } else { # in body
            # ...
        }
    } continue {
        close ARGV if eof;             # reset $. each file
    }

Here's a simple example to illustrate the difference between the two range operators:

    @lines = ("   - Foo",
              "01 - Bar",
              "1  - Baz",
              "   - Quux");

    foreach(@lines)
    {
        if (/0/ .. /1/)
        {
            print "$_\n";
        }
    }

This program will print only the line containing Bar. If the range operator is changed to CW..., it will also print the Baz line.

And now some examples as a list operator:

    for (101 .. 200) { print; } # print $_ 100 times
    @foo = @foo[0 .. $#foo];    # an expensive no-op
    @foo = @foo[$#foo-4 .. $#foo];      # slice last 5 items

The range operator (in list context) makes use of the magical auto-increment algorithm if the operands are strings. You can say

    @alphabet = ('A' .. 'Z');

to get all normal letters of the English alphabet, or

    $hexdigit = (0 .. 9, 'a' .. 'f')[$num & 15];

to get a hexadecimal digit, or

    @z2 = ('01' .. '31');  print $z2[$mday];

to get dates with leading zeros. If the final value specified is not in the sequence that the magical increment would produce, the sequence goes until the next value would be longer than the final value specified.

Because each operand is evaluated in integer form, CW2.18 .. 3.14 will return two elements in list context.

    @list = (2.18 .. 3.14); # same as @list = (2 .. 3);

Ternary ?: is the conditional operator, just as in C. It works much like an if-then-else. If the argument before the ? is true, the argument before the : is returned, otherwise the argument after the : is returned. For example:

    printf "I have %d dog%s.\n", $n,
            ($n == 1) ? '' : "s";

Scalar or list context propagates downward into the 2nd or 3rd argument, whichever is selected.

    $a = $ok ? $b : $c;  # get a scalar
    @a = $ok ? @b : @c;  # get an array
    $a = $ok ? @b : @c;  # oops, that's just a count!

The operator may be assigned to if both the 2nd and 3rd arguments are legal lvalues (meaning that you can assign to them):

    ($a_or_b ? $a : $b) = $c;

Because this operator produces an assignable result, using assignments without parentheses will get you in trouble. For example, this:

    $a % 2 ? $a += 10 : $a += 2

Really means this:

    (($a % 2) ? ($a += 10) : $a) += 2

Rather than this:

    ($a % 2) ? ($a += 10) : ($a += 2)

That should probably be written more simply as:

    $a += ($a % 2) ? 10 : 2;

= is the ordinary assignment operator.

Assignment operators work as in C. That is,

    $a += 2;

is equivalent to

    $a = $a + 2;

although without duplicating any side effects that dereferencing the lvalue might trigger, such as from tie(). Other assignment operators work similarly. The following are recognized:

    **=    +=    *=    &=    <<=    &&=
           -=    /=    |=    >>=    ||=
           .=    %=    ^=
                 x=

Although these are grouped by family, they all have the precedence of assignment.

Unlike in C, the scalar assignment operator produces a valid lvalue. Modifying an assignment is equivalent to doing the assignment and then modifying the variable that was assigned to. This is useful for modifying a copy of something, like this:

    ($tmp = $global) =~ tr [A-Z] [a-z];

Likewise,

    ($a += 2) *= 3;

is equivalent to

    $a += 2;
    $a *= 3;

Similarly, a list assignment in list context produces the list of lvalues assigned to, and a list assignment in scalar context returns the number of elements produced by the expression on the right hand side of the assignment.

Binary , is the comma operator. In scalar context it evaluates its left argument, throws that value away, then evaluates its right argument and returns that value. This is just like C's comma operator.

In list context, it's just the list argument separator, and inserts both its arguments into the list.

The CW=> operator is a synonym for the comma, but forces any word (consisting entirely of word characters) to its left to be interpreted as a string (as of 5.001). If the argument on the left is not a word, it is first interpreted as an expression, and then the string value of that is used.

The CW=> operator is helpful in documenting the correspondence between keys and values in hashes, and other paired elements in lists.

On the right side of a list operator, it has very low precedence, such that it controls all comma-separated expressions found there. The only operators with lower precedence are the logical operators and, or, and not, which may be used to evaluate calls to list operators without the need for extra parentheses:

    open HANDLE, "filename"
        or die "Can't open: $!\n";

See also discussion of list operators in Terms and List Operators (Leftward).

Unary not returns the logical negation of the expression to its right. It's the equivalent of ! except for the very low precedence.

Binary and returns the logical conjunction of the two surrounding expressions. It's equivalent to && except for the very low precedence. This means that it short-circuits: i.e., the right expression is evaluated only if the left expression is true.

Binary or returns the logical disjunction of the two surrounding expressions. It's equivalent to || except for the very low precedence. This makes it useful for control flow

    print FH $data              or die "Can't write to FH: $!";

This means that it short-circuits: i.e., the right expression is evaluated only if the left expression is false. Due to its precedence, you should probably avoid using this for assignment, only for control flow.

    $a = $b or $c;              # bug: this is wrong
    ($a = $b) or $c;            # really means this
    $a = $b || $c;              # better written this way

However, when it's a list-context assignment and you're trying to use || for control flow, you probably need or so that the assignment takes higher precedence.

    @info = stat($file) || die;     # oops, scalar sense of stat!
    @info = stat($file) or die;     # better, now @info gets its due

Then again, you could always use parentheses.

Binary xor returns the exclusive-OR of the two surrounding expressions. It cannot short circuit, of course.

Here is what C has that Perl doesn't:

unary &

Address-of operator. (But see the \ operator for taking a reference.)

unary *

Dereference-address operator. (Perl's prefix dereferencing operators are typed: $, @, %, and &.)

(TYPE)

Type-casting operator.

While we usually think of quotes as literal values, in Perl they function as operators, providing various kinds of interpolating and pattern matching capabilities. Perl provides customary quote characters for these behaviors, but also provides a way for you to choose your quote character for any of them. In the following table, a CW{} represents any pair of delimiters you choose.

    Customary  Generic        Meaning        Interpolates
        ''       q{}          Literal             no
        ""      qq{}          Literal             yes
        ``      qx{}          Command             yes*
                qw{}         Word list            no
        //       m{}       Pattern match          yes*
                qr{}          Pattern             yes*
                 s{}{}      Substitution          yes*
                tr{}{}    Transliteration         no (but see below)
        <<EOF                 here-doc            yes*

        * unless the delimiter is ''.

Non-bracketing delimiters use the same character fore and aft, but the four sorts of brackets (round, angle, square, curly) will all nest, which means that

        q{foo{bar}baz}

is the same as

        'foo{bar}baz'

Note, however, that this does not always work for quoting Perl code:

        $s = q{ if($a eq "}") ... }; # WRONG

is a syntax error. The CWText::Balanced module (from CPAN, and starting from Perl 5.8 part of the standard distribution) is able to do this properly.

There can be whitespace between the operator and the quoting characters, except when CW# is being used as the quoting character. CWq#foo# is parsed as the string CWfoo, while CWq #foo# is the operator CWq followed by a comment. Its argument will be taken from the next line. This allows you to write:

    s {foo}  # Replace foo
      {bar}  # with bar.

The following escape sequences are available in constructs that interpolate and in transliterations.

    \t          tab             (HT, TAB)
    \n          newline         (NL)
    \r          return          (CR)
    \f          form feed       (FF)
    \b          backspace       (BS)
    \a          alarm (bell)    (BEL)
    \e          escape          (ESC)
    \033        octal char      (ESC)
    \x1b        hex char        (ESC)
    \x{263a}    wide hex char   (SMILEY)
    \c[         control char    (ESC)
    \N{name}    named Unicode character

NOTE: Unlike C and other languages, Perl has no \v escape sequence for the vertical tab (VT - ASCII 11).

The following escape sequences are available in constructs that interpolate but not in transliterations.

    \l          lowercase next char
    \u          uppercase next char
    \L          lowercase till \E
    \U          uppercase till \E
    \E          end case modification
    \Q          quote non-word characters till \E

If CWuse locale is in effect, the case map used by CW\l, CW\L, CW\u and CW\U is taken from the current locale. See perllocale. If Unicode (for example, CW\N{} or wide hex characters of 0x100 or beyond) is being used, the case map used by CW\l, CW\L, CW\u and CW\U is as defined by Unicode. For documentation of CW\N{name}, see charnames.

All systems use the virtual CW"\n" to represent a line terminator, called a newline. There is no such thing as an unvarying, physical newline character. It is only an illusion that the operating system, device drivers, C libraries, and Perl all conspire to preserve. Not all systems read CW"\r" as ASCII CR and CW"\n" as ASCII LF. For example, on a Mac, these are reversed, and on systems without line terminator, printing CW"\n" may emit no actual data. In general, use CW"\n" when you mean a newline for your system, but use the literal ASCII when you need an exact character. For example, most networking protocols expect and prefer a CR+LF (CW"\015\012" or CW"\cM\cJ") for line terminators, and although they often accept just CW"\012", they seldom tolerate just CW"\015". If you get in the habit of using CW"\n" for networking, you may be burned some day.

For constructs that do interpolate, variables beginning with "CW$ or CW@" are interpolated. Subscripted variables such as CW$a[3] or CW$href->{key}[0] are also interpolated, as are array and hash slices. But method calls such as CW$obj->meth are not.

Interpolating an array or slice interpolates the elements in order, separated by the value of CW$", so is equivalent to interpolating CWjoin $", @array. Punctuation arrays such as CW@+ are only interpolated if the name is enclosed in braces CW@{+}.

You cannot include a literal CW$ or CW@ within a CW\Q sequence. An unescaped CW$ or CW@ interpolates the corresponding variable, while escaping will cause the literal string CW\$ to be inserted. You'll need to write something like CWm/\Quser\E\@\Qhost/.

Patterns are subject to an additional level of interpretation as a regular expression. This is done as a second pass, after variables are interpolated, so that regular expressions may be incorporated into the pattern from the variables. If this is not what you want, use CW\Q to interpolate a variable literally.

Apart from the behavior described above, Perl does not expand multiple levels of interpolation. In particular, contrary to the expectations of shell programmers, back-quotes do NOT interpolate within double quotes, nor do single quotes impede evaluation of variables when used within double quotes.

Here are the quote-like operators that apply to pattern matching and related activities.

?PATTERN?

This is just like the CW/pattern/ search, except that it matches only once between calls to the reset() operator. This is a useful optimization when you want to see only the first occurrence of something in each file of a set of files, for instance. Only CW?? patterns local to the current package are reset.

    while (<>) {
        if (?^$?) {
                            # blank line between header and body
        }
    } continue {
        reset if eof;       # clear ?? status for next file
    }

This usage is vaguely deprecated, which means it just might possibly be removed in some distant future version of Perl, perhaps somewhere around the year 2168.

m/PATTERN/cgimosx

/PATTERN/cgimosx

Searches a string for a pattern match, and in scalar context returns true if it succeeds, false if it fails. If no string is specified via the CW=~ or CW!~ operator, the CW$_ string is searched. (The string specified with CW=~ need not be an lvalueit may be the result of an expression evaluation, but remember the CW=~ binds rather tightly.) See also perlre. See perllocale for discussion of additional considerations that apply when CWuse locale is in effect. Options are:

    c   Do not reset search position on a failed match when /g is in effect.
    g   Match globally, i.e., find all occurrences.
    i   Do case-insensitive pattern matching.
    m   Treat string as multiple lines.
    o   Compile pattern only once.
    s   Treat string as single line.
    x   Use extended regular expressions.

If / is the delimiter then the initial CWm is optional. With the CWm you can use any pair of non-alphanumeric, non-whitespace characters as delimiters. This is particularly useful for matching path names that contain /, to avoid LTS (leaning toothpick syndrome). If ? is the delimiter, then the match-only-once rule of CW?PATTERN? applies. If ' is the delimiter, no interpolation is performed on the PATTERN. PATTERN may contain variables, which will be interpolated (and the pattern recompiled) every time the pattern search is evaluated, except for when the delimiter is a single quote. (Note that CW$(, CW$), and CW$| are not interpolated because they look like end-of-string tests.) If you want such a pattern to be compiled only once, add a CW/o after the trailing delimiter. This avoids expensive run-time recompilations, and is useful when the value you are interpolating won't change over the life of the script. However, mentioning CW/o constitutes a promise that you won't change the variables in the pattern. If you change them, Perl won't even notice. See also qr/STRING/imosx. If the PATTERN evaluates to the empty string, the last successfully matched regular expression is used instead. In this case, only the CWg and CWc flags on the empty pattern is honoured - the other flags are taken from the original pattern. If no match has previously succeeded, this will (silently) act instead as a genuine empty pattern (which will always match). If the CW/g option is not used, CWm// in list context returns a list consisting of the subexpressions matched by the parentheses in the pattern, i.e., (CW$1, CW$2, CW$3...). (Note that here CW$1 etc. are also set, and that this differs from Perl 4's behavior.) When there are no parentheses in the pattern, the return value is the list CW(1) for success. With or without parentheses, an empty list is returned upon failure. Examples:

    open(TTY, '/dev/tty');
    <TTY> =~ /^y/i && foo();    # do foo if desired

    if (/Version: *([0-9.]*)/) { $version = $1; }

    next if m#^/usr/spool/uucp#;

    # poor man's grep
    $arg = shift;
    while (<>) {
        print if /$arg/o;       # compile only once
    }

    if (($F1, $F2, $Etc) = ($foo =~ /^(\S+)\s+(\S+)\s*(.*)/))

This last example splits CW$foo into the first two words and the remainder of the line, and assigns those three fields to CW$F1, CW$F2, and CW$Etc. The conditional is true if any variables were assigned, i.e., if the pattern matched. The CW/g modifier specifies global pattern matchingthat is, matching as many times as possible within the string. How it behaves depends on the context. In list context, it returns a list of the substrings matched by any capturing parentheses in the regular expression. If there are no parentheses, it returns a list of all the matched strings, as if there were parentheses around the whole pattern. In scalar context, each execution of CWm//g finds the next match, returning true if it matches, and false if there is no further match. The position after the last match can be read or set using the pos() function; see pos in perlfunc. A failed match normally resets the search position to the beginning of the string, but you can avoid that by adding the CW/c modifier (e.g. CWm//gc). Modifying the target string also resets the search position. You can intermix CWm//g matches with CWm/\G.../g, where CW\G is a zero-width assertion that matches the exact position where the previous CWm//g, if any, left off. Without the CW/g modifier, the CW\G assertion still anchors at pos(), but the match is of course only attempted once. Using CW\G without CW/g on a target string that has not previously had a CW/g match applied to it is the same as using the CW\A assertion to match the beginning of the string. Note also that, currently, CW\G is only properly supported when anchored at the very beginning of the pattern. Examples:

    # list context
    ($one,$five,$fifteen) = (`uptime` =~ /(\d+\.\d+)/g);

    # scalar context
    $/ = "";
    while (defined($paragraph = <>)) {
        while ($paragraph =~ /[a-z]['")]*[.!?]+['")]*\s/g) {
            $sentences++;
        }
    }
    print "$sentences\n";

    # using m//gc with \G
    $_ = "ppooqppqq";
    while ($i++ < 2) {
        print "1: '";
        print $1 while /(o)/gc; print "', pos=", pos, "\n";
        print "2: '";
        print $1 if /\G(q)/gc;  print "', pos=", pos, "\n";
        print "3: '";
        print $1 while /(p)/gc; print "', pos=", pos, "\n";
    }
    print "Final: '$1', pos=",pos,"\n" if /\G(.)/;

The last example should print:

    1: 'oo', pos=4
    2: 'q', pos=5
    3: 'pp', pos=7
    1: '', pos=7
    2: 'q', pos=8
    3: '', pos=8
    Final: 'q', pos=8

Notice that the final match matched CWq instead of CWp, which a match without the CW\G anchor would have done. Also note that the final match did not update CWpos CWpos is only updated on a CW/g match. If the final match did indeed match CWp, it's a good bet that you're running an older (pre-5.6.0) Perl. A useful idiom for CWlex-like scanners is CW/\G.../gc. You can combine several regexps like this to process a string part-by-part, doing different actions depending on which regexp matched. Each regexp tries to match where the previous one leaves off.

 $_ = <<'EOL';
      $url = new URI::URL "http://www/";   die if $url eq "xXx";
 EOL
 LOOP:
    {
      print(" digits"),         redo LOOP if /\G\d+\b[,.;]?\s*/gc;
      print(" lowercase"),      redo LOOP if /\G[a-z]+\b[,.;]?\s*/gc;
      print(" UPPERCASE"),      redo LOOP if /\G[A-Z]+\b[,.;]?\s*/gc;
      print(" Capitalized"),    redo LOOP if /\G[A-Z][a-z]+\b[,.;]?\s*/gc;
      print(" MiXeD"),          redo LOOP if /\G[A-Za-z]+\b[,.;]?\s*/gc;
      print(" alphanumeric"),   redo LOOP if /\G[A-Za-z0-9]+\b[,.;]?\s*/gc;
      print(" line-noise"),     redo LOOP if /\G[^A-Za-z0-9]+/gc;
      print ". That's all!\n";
    }

Here is the output (split into several lines):

 line-noise lowercase line-noise lowercase UPPERCASE line-noise
 UPPERCASE line-noise lowercase line-noise lowercase line-noise
 lowercase lowercase line-noise lowercase lowercase line-noise
 MiXeD line-noise. That's all!

q/STRING/

A single-quoted, literal string. A backslash represents a backslash unless followed by the delimiter or another backslash, in which case the delimiter or backslash is interpolated.

    $foo = q!I said, "You said, 'She said it.'"!;
    $bar = q('This is it.');
    $baz = '\n';                # a two-character string

qq/STRING/

A double-quoted, interpolated string.

    $_ .= qq
     (*** The previous line contains the naughty word "$1".\n)
                if /\b(tcl|java|python)\b/i;      # :-)
    $baz = "\n";                # a one-character string

qr/STRING/imosx

This operator quotes (and possibly compiles) its STRING as a regular expression. STRING is interpolated the same way as PATTERN in CWm/PATTERN/. If ' is used as the delimiter, no interpolation is done. Returns a Perl value which may be used instead of the corresponding CW/STRING/imosx expression. For example,

    $rex = qr/my.STRING/is;
    s/$rex/foo/;

is equivalent to

    s/my.STRING/foo/is;

The result may be used as a subpattern in a match:

    $re = qr/$pattern/;
    $string =~ /foo${re}bar/;   # can be interpolated in other patterns
    $string =~ $re;             # or used standalone
    $string =~ /$re/;           # or this way

Since Perl may compile the pattern at the moment of execution of qr() operator, using qr() may have speed advantages in some situations, notably if the result of qr() is used standalone:

    sub match {
        my $patterns = shift;
        my @compiled = map qr/$_/i, @$patterns;
        grep {
            my $success = 0;
            foreach my $pat (@compiled) {
                $success = 1, last if /$pat/;
            }
            $success;
        } @_;
    }

Precompilation of the pattern into an internal representation at the moment of qr() avoids a need to recompile the pattern every time a match CW/$pat/ is attempted. (Perl has many other internal optimizations, but none would be triggered in the above example if we did not use qr() operator.) Options are:

    i   Do case-insensitive pattern matching.
    m   Treat string as multiple lines.
    o   Compile pattern only once.
    s   Treat string as single line.
    x   Use extended regular expressions.

See perlre for additional information on valid syntax for STRING, and for a detailed look at the semantics of regular expressions.

qx/STRING/

`STRING`

A string which is (possibly) interpolated and then executed as a system command with CW/bin/sh or its equivalent. Shell wildcards, pipes, and redirections will be honored. The collected standard output of the command is returned; standard error is unaffected. In scalar context, it comes back as a single (potentially multi-line) string, or undef if the command failed. In list context, returns a list of lines (however you've defined lines with $/ or CW$INPUT_RECORD_SEPARATOR), or an empty list if the command failed. Because backticks do not affect standard error, use shell file descriptor syntax (assuming the shell supports this) if you care to address this. To capture a command's STDERR and STDOUT together:

    $output = `cmd 2>&1`;

To capture a command's STDOUT but discard its STDERR:

    $output = `cmd 2>/dev/null`;

To capture a command's STDERR but discard its STDOUT (ordering is important here):

    $output = `cmd 2>&1 1>/dev/null`;

To exchange a command's STDOUT and STDERR in order to capture the STDERR but leave its STDOUT to come out the old STDERR:

    $output = `cmd 3>&1 1>&2 2>&3 3>&-`;

To read both a command's STDOUT and its STDERR separately, it's easiest to redirect them separately to files, and then read from those files when the program is done:

    system("program args 1>program.stdout 2>program.stderr");

Using single-quote as a delimiter protects the command from Perl's double-quote interpolation, passing it on to the shell instead:

    $perl_info  = qx(ps $$);            # that's Perl's $$
    $shell_info = qx'ps $$';            # that's the new shell's $$

How that string gets evaluated is entirely subject to the command interpreter on your system. On most platforms, you will have to protect shell metacharacters if you want them treated literally. This is in practice difficult to do, as it's unclear how to escape which characters. See perlsec for a clean and safe example of a manual fork() and exec() to emulate backticks safely. On some platforms (notably DOS-like ones), the shell may not be capable of dealing with multiline commands, so putting newlines in the string may not get you what you want. You may be able to evaluate multiple commands in a single line by separating them with the command separator character, if your shell supports that (e.g. CW; on many Unix shells; CW& on the Windows NT CWcmd shell). Beginning with v5.6.0, Perl will attempt to flush all files opened for output before starting the child process, but this may not be supported on some platforms (see perlport). To be safe, you may need to set CW$| ($AUTOFLUSH in English) or call the CWautoflush() method of CWIO::Handle on any open handles. Beware that some command shells may place restrictions on the length of the command line. You must ensure your strings don't exceed this limit after any necessary interpolations. See the platform-specific release notes for more details about your particular environment. Using this operator can lead to programs that are difficult to port, because the shell commands called vary between systems, and may in fact not be present at all. As one example, the CWtype command under the POSIX shell is very different from the CWtype command under DOS. That doesn't mean you should go out of your way to avoid backticks when they're the right way to get something done. Perl was made to be a glue language, and one of the things it glues together is commands. Just understand what you're getting yourself into. See I/O Operators for more discussion.

qw/STRING/

Evaluates to a list of the words extracted out of STRING, using embedded whitespace as the word delimiters. It can be understood as being roughly equivalent to:

    split(' ', q/STRING/);

the differences being that it generates a real list at compile time, and in scalar context it returns the last element in the list. So this expression:

    qw(foo bar baz)

is semantically equivalent to the list:

    'foo', 'bar', 'baz'

Some frequently seen examples:

    use POSIX qw( setlocale localeconv )
    @EXPORT = qw( foo bar baz );

A common mistake is to try to separate the words with comma or to put comments into a multi-line CWqw-string. For this reason, the CWuse warnings pragma and the -w switch (that is, the CW$^W variable) produces warnings if the STRING contains the , or the # character.

s/PATTERN/REPLACEMENT/egimosx

Searches a string for a pattern, and if found, replaces that pattern with the replacement text and returns the number of substitutions made. Otherwise it returns false (specifically, the empty string). If no string is specified via the CW=~ or CW!~ operator, the CW$_ variable is searched and modified. (The string specified with CW=~ must be scalar variable, an array element, a hash element, or an assignment to one of those, i.e., an lvalue.) If the delimiter chosen is a single quote, no interpolation is done on either the PATTERN or the REPLACEMENT. Otherwise, if the PATTERN contains a $ that looks like a variable rather than an end-of-string test, the variable will be interpolated into the pattern at run-time. If you want the pattern compiled only once the first time the variable is interpolated, use the CW/o option. If the pattern evaluates to the empty string, the last successfully executed regular expression is used instead. See perlre for further explanation on these. See perllocale for discussion of additional considerations that apply when CWuse locale is in effect. Options are:

    e   Evaluate the right side as an expression.
    g   Replace globally, i.e., all occurrences.
    i   Do case-insensitive pattern matching.
    m   Treat string as multiple lines.
    o   Compile pattern only once.
    s   Treat string as single line.
    x   Use extended regular expressions.

Any non-alphanumeric, non-whitespace delimiter may replace the slashes. If single quotes are used, no interpretation is done on the replacement string (the CW/e modifier overrides this, however). Unlike Perl 4, Perl 5 treats backticks as normal delimiters; the replacement text is not evaluated as a command. If the PATTERN is delimited by bracketing quotes, the REPLACEMENT has its own pair of quotes, which may or may not be bracketing quotes, e.g., CWs(foo)(bar) or CWs<foo>/bar/. A CW/e will cause the replacement portion to be treated as a full-fledged Perl expression and evaluated right then and there. It is, however, syntax checked at compile-time. A second CWe modifier will cause the replacement portion to be CWevaled before being run as a Perl expression. Examples:

    s/\bgreen\b/mauve/g;                # don't change wintergreen

    $path =~ s|/usr/bin|/usr/local/bin|;

    s/Login: $foo/Login: $bar/; # run-time pattern

    ($foo = $bar) =~ s/this/that/;      # copy first, then change

    $count = ($paragraph =~ s/Mister\b/Mr./g);  # get change-count

    $_ = 'abc123xyz';
    s/\d+/$&*2/e;               # yields 'abc246xyz'
    s/\d+/sprintf("%5d",$&)/e;  # yields 'abc  246xyz'
    s/\w/$& x 2/eg;             # yields 'aabbcc  224466xxyyzz'

    s/%(.)/$percent{$1}/g;      # change percent escapes; no /e
    s/%(.)/$percent{$1} || $&/ge;       # expr now, so /e
    s/^=(\w+)/&pod($1)/ge;      # use function call

    # expand variables in $_, but dynamics only, using
    # symbolic dereferencing
    s/\$(\w+)/${$1}/g;

    # Add one to the value of any numbers in the string
    s/(\d+)/1 + $1/eg;

    # This will expand any embedded scalar variable
    # (including lexicals) in $_ : First $1 is interpolated
    # to the variable name, and then evaluated
    s/(\$\w+)/$1/eeg;

    # Delete (most) C comments.
    $program =~ s {
        /\*     # Match the opening delimiter.
        .*?     # Match a minimal number of characters.
        \*/     # Match the closing delimiter.
    } []gsx;

    s/^\s*(.*?)\s*$/$1/;        # trim whitespace in $_, expensively

    for ($variable) {           # trim whitespace in $variable, cheap
        s/^\s+//;
        s/\s+$//;
    }

    s/([^ ]*) *([^ ]*)/$2 $1/;  # reverse 1st two fields

Note the use of $ instead of \ in the last example. Unlike sed, we use the \<digit> form in only the left hand side. Anywhere else it's $<digit>. Occasionally, you can't use just a CW/g to get all the changes to occur that you might want. Here are two common cases:

    # put commas in the right places in an integer
    1 while s/(\d)(\d\d\d)(?!\d)/$1,$2/g;

    # expand tabs to 8-column spacing
    1 while s/\t+/' ' x (length($&)*8 - length($`)%8)/e;

tr/SEARCHLIST/REPLACEMENTLIST/cds

y/SEARCHLIST/REPLACEMENTLIST/cds

Transliterates all occurrences of the characters found in the search list with the corresponding character in the replacement list. It returns the number of characters replaced or deleted. If no string is specified via the =~ or !~ operator, the CW$_ string is transliterated. (The string specified with =~ must be a scalar variable, an array element, a hash element, or an assignment to one of those, i.e., an lvalue.) A character range may be specified with a hyphen, so CWtr/A-J/0-9/ does the same replacement as CWtr/ACEGIBDFHJ/0246813579/. For sed devotees, CWy is provided as a synonym for CWtr. If the SEARCHLIST is delimited by bracketing quotes, the REPLACEMENTLIST has its own pair of quotes, which may or may not be bracketing quotes, e.g., CWtr[A-Z][a-z] or CWtr(+\-*/)/ABCD/. Note that CWtr does not do regular expression character classes such as CW\d or CW[:lower:]. The <tr> operator is not equivalent to the tr(1) utility. If you want to map strings between lower/upper cases, see lc in perlfunc and uc in perlfunc, and in general consider using the CWs operator if you need regular expressions. Note also that the whole range idea is rather unportable between character setsand even within character sets they may cause results you probably didn't expect. A sound principle is to use only ranges that begin from and end at either alphabets of equal case (a-e, A-E), or digits (0-4). Anything else is unsafe. If in doubt, spell out the character sets in full. Options:

    c   Complement the SEARCHLIST.
    d   Delete found but unreplaced characters.
    s   Squash duplicate replaced characters.

If the CW/c modifier is specified, the SEARCHLIST character set is complemented. If the CW/d modifier is specified, any characters specified by SEARCHLIST not found in REPLACEMENTLIST are deleted. (Note that this is slightly more flexible than the behavior of some tr programs, which delete anything they find in the SEARCHLIST, period.) If the CW/s modifier is specified, sequences of characters that were transliterated to the same character are squashed down to a single instance of the character. If the CW/d modifier is used, the REPLACEMENTLIST is always interpreted exactly as specified. Otherwise, if the REPLACEMENTLIST is shorter than the SEARCHLIST, the final character is replicated till it is long enough. If the REPLACEMENTLIST is empty, the SEARCHLIST is replicated. This latter is useful for counting characters in a class or for squashing character sequences in a class. Examples:

    $ARGV[1] =~ tr/A-Z/a-z/;    # canonicalize to lower case

    $cnt = tr/*/*/;             # count the stars in $_

    $cnt = $sky =~ tr/*/*/;     # count the stars in $sky

    $cnt = tr/0-9//;            # count the digits in $_

    tr/a-zA-Z//s;               # bookkeeper -> bokeper

    ($HOST = $host) =~ tr/a-z/A-Z/;

    tr/a-zA-Z/ /cs;             # change non-alphas to single space

    tr [\200-\377]
       [\000-\177];             # delete 8th bit

If multiple transliterations are given for a character, only the first one is used:

    tr/AAA/XYZ/

will transliterate any A to X. Because the transliteration table is built at compile time, neither the SEARCHLIST nor the REPLACEMENTLIST are subjected to double quote interpolation. That means that if you want to use variables, you must use an eval():

    eval "tr/$oldlist/$newlist/";
    die $@ if $@;

    eval "tr/$oldlist/$newlist/, 1" or die $@;

<<EOF

A line-oriented form of quoting is based on the shell here-document syntax. Following a CW<< you specify a string to terminate the quoted material, and all lines following the current line down to the terminating string are the value of the item. The terminating string may be either an identifier (a word), or some quoted text. If quoted, the type of quotes you use determines the treatment of the text, just as in regular quoting. An unquoted identifier works like double quotes. There must be no space between the CW<< and the identifier, unless the identifier is quoted. (If you put a space it will be treated as a null identifier, which is valid, and matches the first empty line.) The terminating string must appear by itself (unquoted and with no surrounding whitespace) on the terminating line.

       print <<EOF;
    The price is $Price.
    EOF

       print << "EOF"; # same as above
    The price is $Price.
    EOF

       print << `EOC`; # execute commands
    echo hi there
    echo lo there
    EOC

       print <<"foo", <<"bar"; # you can stack them
    I said foo.
    foo
    I said bar.
    bar

       myfunc(<< "THIS", 23, <<'THAT');
    Here's a line
    or two.
    THIS
    and here's another.
    THAT

Just don't forget that you have to put a semicolon on the end to finish the statement, as Perl doesn't know you're not going to try to do this:

       print <<ABC
    179231
    ABC
       + 20;

If you want your here-docs to be indented with the rest of the code, you'll need to remove leading whitespace from each line manually:

    ($quote = <<'FINIS') =~ s/^\s+//gm;
       The Road goes ever on and on, 
       down from the door where it began.
    FINIS

If you use a here-doc within a delimited construct, such as in CWs///eg, the quoted material must come on the lines following the final delimiter. So instead of

    s/this/<<E . 'that'
    the other
    E
     . 'more '/eg;

you have to write

    s/this/<<E . 'that' 
     . 'more '/eg; 
    the other 
    E

If the terminating identifier is on the last line of the program, you must be sure there is a newline after it; otherwise, Perl will give the warning Can't find string terminator END anywhere before EOF.... Additionally, the quoting rules for the identifier are not related to Perl's quoting rules CWq(), CWqq(), and the like are not supported in place of CW'' and CW"", and the only interpolation is for backslashing the quoting character:

    print << "abc\"def";
    testing...
    abc"def

Finally, quoted strings cannot span multiple lines. The general rule is that the identifier must be a string literal. Stick with that, and you should be safe.

When presented with something that might have several different interpretations, Perl uses the DWIM (that's Do What I Mean) principle to pick the most probable interpretation. This strategy is so successful that Perl programmers often do not suspect the ambivalence of what they write. But from time to time, Perl's notions differ substantially from what the author honestly meant.

This section hopes to clarify how Perl handles quoted constructs. Although the most common reason to learn this is to unravel labyrinthine regular expressions, because the initial steps of parsing are the same for all quoting operators, they are all discussed together.

The most important Perl parsing rule is the first one discussed below: when processing a quoted construct, Perl first finds the end of that construct, then interprets its contents. If you understand this rule, you may skip the rest of this section on the first reading. The other rules are likely to contradict the user's expectations much less frequently than this first one.

Some passes discussed below are performed concurrently, but because their results are the same, we consider them individually. For different quoting constructs, Perl performs different numbers of passes, from one to five, but these passes are always performed in the same order.

Finding the end

The first pass is finding the end of the quoted construct, whether it be a multicharacter delimiter CW"\nEOF\n" in the CW<<EOF construct, a CW/ that terminates a CWqq// construct, a CW] which terminates CWqq[] construct, or a CW> which terminates a fileglob started with CW<. When searching for single-character non-pairing delimiters, such as CW/, combinations of CW\\ and CW\/ are skipped. However, when searching for single-character pairing delimiter like CW[, combinations of CW\\, CW\], and CW\[ are all skipped, and nested CW[, CW] are skipped as well. When searching for multicharacter delimiters, nothing is skipped. For constructs with three-part delimiters (CWs///, CWy///, and CWtr///), the search is repeated once more. During this search no attention is paid to the semantics of the construct. Thus:

    "$hash{"$foo/$bar"}"

or:

    m/ 
      bar       # NOT a comment, this slash / terminated m//!
     /x

do not form legal quoted expressions. The quoted part ends on the first CW" and CW/, and the rest happens to be a syntax error. Because the slash that terminated CWm// was followed by a CWSPACE, the example above is not CWm//x, but rather CWm// with no CW/x modifier. So the embedded CW# is interpreted as a literal CW#.

Removal of backslashes before delimiters

During the second pass, text between the starting and ending delimiters is copied to a safe location, and the CW\ is removed from combinations consisting of CW\ and delimiteror delimiters, meaning both starting and ending delimiters will should these differ. This removal does not happen for multi-character delimiters. Note that the combination CW\\ is left intact, just as it was. Starting from this step no information about the delimiters is used in parsing.

Interpolation

The next step is interpolation in the text obtained, which is now delimiter-independent. There are four different cases. No interpolation is performed. The only interpolation is removal of CW\ from pairs CW\\. CW\Q, CW\U, CW\u, CW\L, CW\l (possibly paired with CW\E) are converted to corresponding Perl constructs. Thus, CW"$foo\Qbaz$bar" is converted to CW$foo . (quotemeta("baz" . $bar)) internally. The other combinations are replaced with appropriate expansions. Let it be stressed that whatever falls between CI\Q and CI\E is interpolated in the usual way. Something like CW"\Q\\E" has no CW\E inside. instead, it has CW\Q, CW\\, and CWE, so the result is the same as for CW"\\\\E". As a general rule, backslashes between CW\Q and CW\E may lead to counterintuitive results. So, CW"\Q\t\E" is converted to CWquotemeta("\t"), which is the same as CW"\\\t" (since TAB is not alphanumeric). Note also that:

  $str = '\t';
  return "\Q$str";

may be closer to the conjectural intention of the writer of CW"\Q\t\E". Interpolated scalars and arrays are converted internally to the CWjoin and CW. catenation operations. Thus, CW"$foo XXX '@arr'" becomes:

  $foo . " XXX '" . (join $", @arr) . "'";

All operations above are performed simultaneously, left to right. Because the result of CW"\Q STRING \E" has all metacharacters quoted, there is no way to insert a literal CW$ or CW@ inside a CW\Q\E pair. If protected by CW\, CW$ will be quoted to became CW"\\\$"; if not, it is interpreted as the start of an interpolated scalar. Note also that the interpolation code needs to make a decision on where the interpolated scalar ends. For instance, whether CW"a $b -> {c}" really means:

  "a " . $b . " -> {c}";

or:

  "a " . $b -> {c};

Most of the time, the longest possible text that does not include spaces between components and which contains matching braces or brackets. because the outcome may be determined by voting based on heuristic estimators, the result is not strictly predictable. Fortunately, it's usually correct for ambiguous cases. Processing of CW\Q, CW\U, CW\u, CW\L, CW\l, and interpolation happens (almost) as with CWqq// constructs, but the substitution of CW\ followed by RE-special chars (including CW\) is not performed. Moreover, inside CW(?{BLOCK}), CW(?# comment ), and a CW#-comment in a CW//x-regular expression, no processing is performed whatsoever. This is the first step at which the presence of the CW//x modifier is relevant. Interpolation has several quirks: CW$|, CW$(, and CW$) are not interpolated, and constructs CW$var[SOMETHING] are voted (by several different estimators) to be either an array element or CW$var followed by an RE alternative. This is where the notation CW${arr[$bar]} comes handy: CW/${arr[0-9]}/ is interpreted as array element CW-9, not as a regular expression from the variable CW$arr followed by a digit, which would be the interpretation of CW/$arr[0-9]/. Since voting among different estimators may occur, the result is not predictable. It is at this step that CW\1 is begrudgingly converted to CW$1 in the replacement text of CWs/// to correct the incorrigible sed hackers who haven't picked up the saner idiom yet. A warning is emitted if the CWuse warnings pragma or the -w command-line flag (that is, the CW$^W variable) was set. The lack of processing of CW\\ creates specific restrictions on the post-processed text. If the delimiter is CW/, one cannot get the combination CW\/ into the result of this step. CW/ will finish the regular expression, CW\/ will be stripped to CW/ on the previous step, and CW\\/ will be left as is. Because CW/ is equivalent to CW\/ inside a regular expression, this does not matter unless the delimiter happens to be character special to the RE engine, such as in CWs*foo*bar*, CWm[foo], or CW?foo?; or an alphanumeric char, as in:

  m m ^ a \s* b mmx;

In the RE above, which is intentionally obfuscated for illustration, the delimiter is CWm, the modifier is CWmx, and after backslash-removal the RE is the same as for CWm/ ^ a \s* b /mx. There's more than one reason you're encouraged to restrict your delimiters to non-alphanumeric, non-whitespace choices. This step is the last one for all constructs except regular expressions, which are processed further.

Interpolation of regular expressions

Previous steps were performed during the compilation of Perl code, but this one happens at run timealthough it may be optimized to be calculated at compile time if appropriate. After preprocessing described above, and possibly after evaluation if catenation, joining, casing translation, or metaquoting are involved, the resulting string is passed to the RE engine for compilation. Whatever happens in the RE engine might be better discussed in perlre, but for the sake of continuity, we shall do so here. This is another step where the presence of the CW//x modifier is relevant. The RE engine scans the string from left to right and converts it to a finite automaton. Backslashed characters are either replaced with corresponding literal strings (as with CW\{), or else they generate special nodes in the finite automaton (as with CW\b). Characters special to the RE engine (such as CW|) generate corresponding nodes or groups of nodes. CW(?#...) comments are ignored. All the rest is either converted to literal strings to match, or else is ignored (as is whitespace and CW#-style comments if CW//x is present). Parsing of the bracketed character class construct, CW[...], is rather different than the rule used for the rest of the pattern. The terminator of this construct is found using the same rules as for finding the terminator of a CW{}-delimited construct, the only exception being that CW] immediately following CW[ is treated as though preceded by a backslash. Similarly, the terminator of CW(?{...}) is found using the same rules as for finding the terminator of a CW{}-delimited construct. It is possible to inspect both the string given to RE engine and the resulting finite automaton. See the arguments CWdebug/CWdebugcolor in the CWuse re pragma, as well as Perl's -Dr command-line switch documented in Command Switches in perlrun.

Optimization of regular expressions

This step is listed for completeness only. Since it does not change semantics, details of this step are not documented and are subject to change without notice. This step is performed over the finite automaton that was generated during the previous pass. It is at this stage that CWsplit() silently optimizes CW/^/ to mean CW/^/m.

There are several I/O operators you should know about.

A string enclosed by backticks (grave accents) first undergoes double-quote interpolation. It is then interpreted as an external command, and the output of that command is the value of the backtick string, like in a shell. In scalar context, a single string consisting of all output is returned. In list context, a list of values is returned, one per line of output. (You can set CW$/ to use a different line terminator.) The command is executed each time the pseudo-literal is evaluated. The status value of the command is returned in CW$? (see perlvar for the interpretation of CW$?). Unlike in csh, no translation is done on the return datanewlines remain newlines. Unlike in any of the shells, single quotes do not hide variable names in the command from interpretation. To pass a literal dollar-sign through to the shell you need to hide it with a backslash. The generalized form of backticks is CWqx//. (Because backticks always undergo shell expansion as well, see perlsec for security concerns.)

In scalar context, evaluating a filehandle in angle brackets yields the next line from that file (the newline, if any, included), or CWundef at end-of-file or on error. When CW$/ is set to CWundef (sometimes known as file-slurp mode) and the file is empty, it returns CW'' the first time, followed by CWundef subsequently.

Ordinarily you must assign the returned value to a variable, but there is one situation where an automatic assignment happens. If and only if the input symbol is the only thing inside the conditional of a CWwhile statement (even if disguised as a CWfor(;;) loop), the value is automatically assigned to the global variable CW$_, destroying whatever was there previously. (This may seem like an odd thing to you, but you'll use the construct in almost every Perl script you write.) The CW$_ variable is not implicitly localized. You'll have to put a CWlocal $_; before the loop if you want that to happen.

The following lines are equivalent:

    while (defined($_ = <STDIN>)) { print; }
    while ($_ = <STDIN>) { print; }
    while (<STDIN>) { print; }
    for (;<STDIN>;) { print; }
    print while defined($_ = <STDIN>);
    print while ($_ = <STDIN>);
    print while <STDIN>;

This also behaves similarly, but avoids CW$_ :

    while (my $line = <STDIN>) { print $line }

In these loop constructs, the assigned value (whether assignment is automatic or explicit) is then tested to see whether it is defined. The defined test avoids problems where line has a string value that would be treated as false by Perl, for example a " or a 0" with no trailing newline. If you really mean for such values to terminate the loop, they should be tested for explicitly:

    while (($_ = <STDIN>) ne '0') { ... }
    while (<STDIN>) { last unless $_; ... }

In other boolean contexts, CW<CIfilehandleCW> without an explicit CWdefined test or comparison elicit a warning if the CWuse warnings pragma or the -w command-line switch (the CW$^W variable) is in effect.

The filehandles STDIN, STDOUT, and STDERR are predefined. (The filehandles CWstdin, CWstdout, and CWstderr will also work except in packages, where they would be interpreted as local identifiers rather than global.) Additional filehandles may be created with the open() function, amongst others. See perlopentut and open in perlfunc for details on this.

If a <FILEHANDLE> is used in a context that is looking for a list, a list comprising all input lines is returned, one line per list element. It's easy to grow to a rather large data space this way, so use with care.

<FILEHANDLE> may also be spelled CWreadline(*FILEHANDLE). See readline in perlfunc.

The null filehandle <> is special: it can be used to emulate the behavior of sed and awk. Input from <> comes either from standard input, or from each file listed on the command line. Here's how it works: the first time <> is evaluated, the CW@ARGV array is checked, and if it is empty, CW$ARGV[0] is set to -, which when opened gives you standard input. The CW@ARGV array is then processed as a list of filenames. The loop

    while (<>) {
        ...                     # code for each line
    }

is equivalent to the following Perl-like pseudo code:

    unshift(@ARGV, '-') unless @ARGV;
    while ($ARGV = shift) {
        open(ARGV, $ARGV);
        while (<ARGV>) {
            ...         # code for each line
        }
    }

except that it isn't so cumbersome to say, and will actually work. It really does shift the CW@ARGV array and put the current filename into the CW$ARGV variable. It also uses filehandle ARGV internally--<> is just a synonym for <ARGV>, which is magical. (The pseudo code above doesn't work because it treats <ARGV> as non-magical.)

You can modify CW@ARGV before the first <> as long as the array ends up containing the list of filenames you really want. Line numbers (CW$.) continue as though the input were one big happy file. See the example in eof in perlfunc for how to reset line numbers on each file.

If you want to set CW@ARGV to your own list of files, go right ahead. This sets CW@ARGV to all plain text files if no CW@ARGV was given:

    @ARGV = grep { -f && -T } glob('*') unless @ARGV;

You can even set them to pipe commands. For example, this automatically filters compressed arguments through gzip:

    @ARGV = map { /\.(gz|Z)$/ ? "gzip -dc < $_ |" : $_ } @ARGV;

If you want to pass switches into your script, you can use one of the Getopts modules or put a loop on the front like this:

    while ($_ = $ARGV[0], /^-/) {
        shift;
        last if /^--$/;
        if (/^-D(.*)/) { $debug = $1 }
        if (/^-v/)     { $verbose++  }
        # ...           # other switches
    }

    while (<>) {
        # ...           # code for each line
    }

The <> symbol will return CWundef for end-of-file only once. If you call it again after this, it will assume you are processing another CW@ARGV list, and if you haven't set CW@ARGV, will read input from STDIN.

If what the angle brackets contain is a simple scalar variable (e.g., <$foo>), then that variable contains the name of the filehandle to input from, or its typeglob, or a reference to the same. For example:

    $fh = \*STDIN;
    $line = <$fh>;

If what's within the angle brackets is neither a filehandle nor a simple scalar variable containing a filehandle name, typeglob, or typeglob reference, it is interpreted as a filename pattern to be globbed, and either a list of filenames or the next filename in the list is returned, depending on context. This distinction is determined on syntactic grounds alone. That means CW<$x> is always a readline() from an indirect handle, but CW<$hash{key}> is always a glob(). That's because CW$x is a simple scalar variable, but CW$hash{key} is notit's a hash element.

One level of double-quote interpretation is done first, but you can't say CW<$foo> because that's an indirect filehandle as explained in the previous paragraph. (In older versions of Perl, programmers would insert curly brackets to force interpretation as a filename glob: CW<${foo}>. These days, it's considered cleaner to call the internal function directly as CWglob($foo), which is probably the right way to have done it in the first place.) For example:

    while (<*.c>) {
        chmod 0644, $_;
    }

is roughly equivalent to:

    open(FOO, "echo *.c | tr -s ' \t\r\f' '\\012\\012\\012\\012'|");
    while (<FOO>) {
        chomp;
        chmod 0644, $_;
    }

except that the globbing is actually done internally using the standard CWFile::Glob extension. Of course, the shortest way to do the above is:

    chmod 0644, <*.c>;

A (file)glob evaluates its (embedded) argument only when it is starting a new list. All values must be read before it will start over. In list context, this isn't important because you automatically get them all anyway. However, in scalar context the operator returns the next value each time it's called, or CWundef when the list has run out. As with filehandle reads, an automatic CWdefined is generated when the glob occurs in the test part of a CWwhile, because legal glob returns (e.g. a file called 0) would otherwise terminate the loop. Again, CWundef is returned only once. So if you're expecting a single value from a glob, it is much better to say

    ($file) = <blurch*>;

than

    $file = <blurch*>;

because the latter will alternate between returning a filename and returning false.

If you're trying to do variable interpolation, it's definitely better to use the glob() function, because the older notation can cause people to become confused with the indirect filehandle notation.

    @files = glob("$dir/*.[ch]");
    @files = glob($files[$i]);

Like C, Perl does a certain amount of expression evaluation at compile time whenever it determines that all arguments to an operator are static and have no side effects. In particular, string concatenation happens at compile time between literals that don't do variable substitution. Backslash interpolation also happens at compile time. You can say

    'Now is the time for all' . "\n" .
        'good men to come to.'

and this all reduces to one string internally. Likewise, if you say

    foreach $file (@filenames) {
        if (-s $file > 5 + 100 * 2**16) {  }
    }

the compiler will precompute the number which that expression represents so that the interpreter won't have to.

Perl doesn't officially have a no-op operator, but the bare constants CW0 and CW1 are special-cased to not produce a warning in a void context, so you can for example safely do

    1 while foo();

Bitstrings of any size may be manipulated by the bitwise operators (CW~ | & ^).

If the operands to a binary bitwise op are strings of different sizes, | and ^ ops act as though the shorter operand had additional zero bits on the right, while the & op acts as though the longer operand were truncated to the length of the shorter. The granularity for such extension or truncation is one or more bytes.

    # ASCII-based examples 
    print "j p \n" ^ " a h";            # prints "JAPH\n"
    print "JA" | "  ph\n";              # prints "japh\n"
    print "japh\nJunk" & '_____';       # prints "JAPH\n";
    print 'p N$' ^ " E<H\n";            # prints "Perl\n";

If you are intending to manipulate bitstrings, be certain that you're supplying bitstrings: If an operand is a number, that will imply a numeric bitwise operation. You may explicitly show which type of operation you intend by using CW"" or CW0+, as in the examples below.

    $foo =  150  |  105 ;       # yields 255  (0x96 | 0x69 is 0xFF)
    $foo = '150' |  105 ;       # yields 255
    $foo =  150  | '105';       # yields 255
    $foo = '150' | '105';       # yields string '155' (under ASCII)

    $baz = 0+$foo & 0+$bar;     # both ops explicitly numeric
    $biz = "$foo" ^ "$bar";     # both ops explicitly stringy

See vec in perlfunc for information on how to manipulate individual bits in a bit vector.

By default, Perl assumes that it must do most of its arithmetic in floating point. But by saying

    use integer;

you may tell the compiler that it's okay to use integer operations (if it feels like it) from here to the end of the enclosing BLOCK. An inner BLOCK may countermand this by saying

    no integer;

which lasts until the end of that BLOCK. Note that this doesn't mean everything is only an integer, merely that Perl may use integer operations if it is so inclined. For example, even under CWuse integer, if you take the CWsqrt(2), you'll still get CW1.4142135623731 or so.

Used on numbers, the bitwise operators (&, |, ^, ~, <<, and >>) always produce integral results. (But see also Bitwise String Operators.) However, CWuse integer still has meaning for them. By default, their results are interpreted as unsigned integers, but if CWuse integer is in effect, their results are interpreted as signed integers. For example, CW~0 usually evaluates to a large integral value. However, CWuse integer; ~0 is CW-1 on twos-complement machines.

While CWuse integer provides integer-only arithmetic, there is no analogous mechanism to provide automatic rounding or truncation to a certain number of decimal places. For rounding to a certain number of digits, sprintf() or printf() is usually the easiest route. See perlfaq4.

Floating-point numbers are only approximations to what a mathematician would call real numbers. There are infinitely more reals than floats, so some corners must be cut. For example:

    printf "%.20g\n", 123456789123456789;
    #        produces 123456789123456784

Testing for exact equality of floating-point equality or inequality is not a good idea. Here's a (relatively expensive) work-around to compare whether two floating-point numbers are equal to a particular number of decimal places. See Knuth, volume II, for a more robust treatment of this topic.

    sub fp_equal {
        my ($X, $Y, $POINTS) = @_;
        my ($tX, $tY);
        $tX = sprintf("%.${POINTS}g", $X);
        $tY = sprintf("%.${POINTS}g", $Y);
        return $tX eq $tY;
    }

The POSIX module (part of the standard perl distribution) implements ceil(), floor(), and other mathematical and trigonometric functions. The Math::Complex module (part of the standard perl distribution) defines mathematical functions that work on both the reals and the imaginary numbers. Math::Complex not as efficient as POSIX, but POSIX can't work with complex numbers.

Rounding in financial applications can have serious implications, and the rounding method used should be specified precisely. In these cases, it probably pays not to trust whichever system rounding is being used by Perl, but to instead implement the rounding function you need yourself.

The standard Math::BigInt and Math::BigFloat modules provide variable-precision arithmetic and overloaded operators, although they're currently pretty slow. At the cost of some space and considerable speed, they avoid the normal pitfalls associated with limited-precision representations.

    use Math::BigInt;
    $x = Math::BigInt->new('123456789123456789');
    print $x * $x;

    # prints +15241578780673678515622620750190521

There are several modules that let you calculate with (bound only by memory and cpu-time) unlimited or fixed precision. There are also some non-standard modules that provide faster implementations via external C libraries.

Here is a short, but incomplete summary:

        Math::Fraction          big, unlimited fractions like 9973 / 12967
        Math::String            treat string sequences like numbers
        Math::FixedPrecision    calculate with a fixed precision
        Math::Currency          for currency calculations
        Bit::Vector             manipulate bit vectors fast (uses C)
        Math::BigIntFast        Bit::Vector wrapper for big numbers
        Math::Pari              provides access to the Pari C library
        Math::BigInteger        uses an external C library
        Math::Cephes            uses external Cephes C library (no big numbers)
        Math::Cephes::Fraction  fractions via the Cephes library
        Math::GMP               another one using an external C library

Choose wisely.