4 perlop - Perl operators and precedence
8 =head2 Operator Precedence and Associativity
9 X<operator, precedence> X<precedence> X<associativity>
11 Operator precedence and associativity work in Perl more or less like
12 they do in mathematics.
14 I<Operator precedence> means some operators are evaluated before
15 others. For example, in C<2 + 4 * 5>, the multiplication has higher
16 precedence so C<4 * 5> is evaluated first yielding C<2 + 20 ==
17 22> and not C<6 * 5 == 30>.
19 I<Operator associativity> defines what happens if a sequence of the
20 same operators is used one after another: whether the evaluator will
21 evaluate the left operations first or the right. For example, in C<8
22 - 4 - 2>, subtraction is left associative so Perl evaluates the
23 expression left to right. C<8 - 4> is evaluated first making the
24 expression C<4 - 2 == 2> and not C<8 - 2 == 6>.
26 Perl operators have the following associativity and precedence,
27 listed from highest precedence to lowest. Operators borrowed from
28 C keep the same precedence relationship with each other, even where
29 C's precedence is slightly screwy. (This makes learning Perl easier
30 for C folks.) With very few exceptions, these all operate on scalar
31 values only, not array values.
33 left terms and list operators (leftward)
37 right ! ~ \ and unary + and -
42 nonassoc named unary operators
43 nonassoc < > <= >= lt gt le ge
44 nonassoc == != <=> eq ne cmp ~~
53 nonassoc list operators (rightward)
58 In the following sections, these operators are covered in precedence order.
60 Many operators can be overloaded for objects. See L<overload>.
62 =head2 Terms and List Operators (Leftward)
63 X<list operator> X<operator, list> X<term>
65 A TERM has the highest precedence in Perl. They include variables,
66 quote and quote-like operators, any expression in parentheses,
67 and any function whose arguments are parenthesized. Actually, there
68 aren't really functions in this sense, just list operators and unary
69 operators behaving as functions because you put parentheses around
70 the arguments. These are all documented in L<perlfunc>.
72 If any list operator (print(), etc.) or any unary operator (chdir(), etc.)
73 is followed by a left parenthesis as the next token, the operator and
74 arguments within parentheses are taken to be of highest precedence,
75 just like a normal function call.
77 In the absence of parentheses, the precedence of list operators such as
78 C<print>, C<sort>, or C<chmod> is either very high or very low depending on
79 whether you are looking at the left side or the right side of the operator.
82 @ary = (1, 3, sort 4, 2);
83 print @ary; # prints 1324
85 the commas on the right of the sort are evaluated before the sort,
86 but the commas on the left are evaluated after. In other words,
87 list operators tend to gobble up all arguments that follow, and
88 then act like a simple TERM with regard to the preceding expression.
89 Be careful with parentheses:
91 # These evaluate exit before doing the print:
92 print($foo, exit); # Obviously not what you want.
93 print $foo, exit; # Nor is this.
95 # These do the print before evaluating exit:
96 (print $foo), exit; # This is what you want.
97 print($foo), exit; # Or this.
98 print ($foo), exit; # Or even this.
102 print ($foo & 255) + 1, "\n";
104 probably doesn't do what you expect at first glance. The parentheses
105 enclose the argument list for C<print> which is evaluated (printing
106 the result of C<$foo & 255>). Then one is added to the return value
107 of C<print> (usually 1). The result is something like this:
109 1 + 1, "\n"; # Obviously not what you meant.
111 To do what you meant properly, you must write:
113 print(($foo & 255) + 1, "\n");
115 See L<Named Unary Operators> for more discussion of this.
117 Also parsed as terms are the C<do {}> and C<eval {}> constructs, as
118 well as subroutine and method calls, and the anonymous
119 constructors C<[]> and C<{}>.
121 See also L<Quote and Quote-like Operators> toward the end of this section,
122 as well as L</"I/O Operators">.
124 =head2 The Arrow Operator
125 X<arrow> X<dereference> X<< -> >>
127 "C<< -> >>" is an infix dereference operator, just as it is in C
128 and C++. If the right side is either a C<[...]>, C<{...}>, or a
129 C<(...)> subscript, then the left side must be either a hard or
130 symbolic reference to an array, a hash, or a subroutine respectively.
131 (Or technically speaking, a location capable of holding a hard
132 reference, if it's an array or hash reference being used for
133 assignment.) See L<perlreftut> and L<perlref>.
135 Otherwise, the right side is a method name or a simple scalar
136 variable containing either the method name or a subroutine reference,
137 and the left side must be either an object (a blessed reference)
138 or a class name (that is, a package name). See L<perlobj>.
140 =head2 Auto-increment and Auto-decrement
141 X<increment> X<auto-increment> X<++> X<decrement> X<auto-decrement> X<-->
143 "++" and "--" work as in C. That is, if placed before a variable,
144 they increment or decrement the variable by one before returning the
145 value, and if placed after, increment or decrement after returning the
149 print $i++; # prints 0
150 print ++$j; # prints 1
152 Note that just as in C, Perl doesn't define B<when> the variable is
153 incremented or decremented. You just know it will be done sometime
154 before or after the value is returned. This also means that modifying
155 a variable twice in the same statement will lead to undefined behaviour.
156 Avoid statements like:
161 Perl will not guarantee what the result of the above statements is.
163 The auto-increment operator has a little extra builtin magic to it. If
164 you increment a variable that is numeric, or that has ever been used in
165 a numeric context, you get a normal increment. If, however, the
166 variable has been used in only string contexts since it was set, and
167 has a value that is not the empty string and matches the pattern
168 C</^[a-zA-Z]*[0-9]*\z/>, the increment is done as a string, preserving each
169 character within its range, with carry:
171 print ++($foo = '99'); # prints '100'
172 print ++($foo = 'a0'); # prints 'a1'
173 print ++($foo = 'Az'); # prints 'Ba'
174 print ++($foo = 'zz'); # prints 'aaa'
176 C<undef> is always treated as numeric, and in particular is changed
177 to C<0> before incrementing (so that a post-increment of an undef value
178 will return C<0> rather than C<undef>).
180 The auto-decrement operator is not magical.
182 =head2 Exponentiation
183 X<**> X<exponentiation> X<power>
185 Binary "**" is the exponentiation operator. It binds even more
186 tightly than unary minus, so -2**4 is -(2**4), not (-2)**4. (This is
187 implemented using C's pow(3) function, which actually works on doubles
190 =head2 Symbolic Unary Operators
191 X<unary operator> X<operator, unary>
193 Unary "!" performs logical negation, i.e., "not". See also C<not> for a lower
194 precedence version of this.
197 Unary "-" performs arithmetic negation if the operand is numeric,
198 including any string that looks like a number. If the operand is
199 an identifier, a string consisting of a minus sign concatenated
200 with the identifier is returned. Otherwise, if the string starts
201 with a plus or minus, a string starting with the opposite sign is
202 returned. One effect of these rules is that -bareword is equivalent
203 to the string "-bareword". If, however, the string begins with a
204 non-alphabetic character (excluding "+" or "-"), Perl will attempt to convert
205 the string to a numeric and the arithmetic negation is performed. If the
206 string cannot be cleanly converted to a numeric, Perl will give the warning
207 B<Argument "the string" isn't numeric in negation (-) at ...>.
208 X<-> X<negation, arithmetic>
210 Unary "~" performs bitwise negation, i.e., 1's complement. For
211 example, C<0666 & ~027> is 0640. (See also L<Integer Arithmetic> and
212 L<Bitwise String Operators>.) Note that the width of the result is
213 platform-dependent: ~0 is 32 bits wide on a 32-bit platform, but 64
214 bits wide on a 64-bit platform, so if you are expecting a certain bit
215 width, remember to use the & operator to mask off the excess bits.
216 X<~> X<negation, binary>
218 Unary "+" has no effect whatsoever, even on strings. It is useful
219 syntactically for separating a function name from a parenthesized expression
220 that would otherwise be interpreted as the complete list of function
221 arguments. (See examples above under L<Terms and List Operators (Leftward)>.)
224 Unary "\" creates a reference to whatever follows it. See L<perlreftut>
225 and L<perlref>. Do not confuse this behavior with the behavior of
226 backslash within a string, although both forms do convey the notion
227 of protecting the next thing from interpolation.
228 X<\> X<reference> X<backslash>
230 =head2 Binding Operators
231 X<binding> X<operator, binding> X<=~> X<!~>
233 Binary "=~" binds a scalar expression to a pattern match. Certain operations
234 search or modify the string $_ by default. This operator makes that kind
235 of operation work on some other string. The right argument is a search
236 pattern, substitution, or transliteration. The left argument is what is
237 supposed to be searched, substituted, or transliterated instead of the default
238 $_. When used in scalar context, the return value generally indicates the
239 success of the operation. The exception is substitution with the C</r>
240 (non-destructive) option, which causes the return value to be the result of
241 the substition. Behavior in list context depends on the particular operator.
242 See L</"Regexp Quote-Like Operators"> for details and L<perlretut> for
243 examples using these operators.
245 If the right argument is an expression rather than a search pattern,
246 substitution, or transliteration, it is interpreted as a search pattern at run
247 time. Note that this means that its contents will be interpolated twice, so
251 is not ok, as the regex engine will end up trying to compile the
252 pattern C<\>, which it will consider a syntax error.
254 Binary "!~" is just like "=~" except the return value is negated in
257 Binary "!~" with a non-destructive substitution (s///r) is a syntax error.
259 =head2 Multiplicative Operators
260 X<operator, multiplicative>
262 Binary "*" multiplies two numbers.
265 Binary "/" divides two numbers.
268 Binary "%" is the modulo operator, which computes the division
269 remainder of its first argument with respect to its second argument.
271 operands C<$a> and C<$b>: If C<$b> is positive, then C<$a % $b> is
272 C<$a> minus the largest multiple of C<$b> less than or equal to
273 C<$a>. If C<$b> is negative, then C<$a % $b> is C<$a> minus the
274 smallest multiple of C<$b> that is not less than C<$a> (i.e. the
275 result will be less than or equal to zero). If the operands
276 C<$a> and C<$b> are floating point values and the absolute value of
277 C<$b> (that is C<abs($b)>) is less than C<(UV_MAX + 1)>, only
278 the integer portion of C<$a> and C<$b> will be used in the operation
279 (Note: here C<UV_MAX> means the maximum of the unsigned integer type).
280 If the absolute value of the right operand (C<abs($b)>) is greater than
281 or equal to C<(UV_MAX + 1)>, "%" computes the floating-point remainder
282 C<$r> in the equation C<($r = $a - $i*$b)> where C<$i> is a certain
283 integer that makes C<$r> have the same sign as the right operand
284 C<$b> (B<not> as the left operand C<$a> like C function C<fmod()>)
285 and the absolute value less than that of C<$b>.
286 Note that when C<use integer> is in scope, "%" gives you direct access
287 to the modulo operator as implemented by your C compiler. This
288 operator is not as well defined for negative operands, but it will
290 X<%> X<remainder> X<modulo> X<mod>
292 Binary "x" is the repetition operator. In scalar context or if the left
293 operand is not enclosed in parentheses, it returns a string consisting
294 of the left operand repeated the number of times specified by the right
295 operand. In list context, if the left operand is enclosed in
296 parentheses or is a list formed by C<qw/STRING/>, it repeats the list.
297 If the right operand is zero or negative, it returns an empty string
298 or an empty list, depending on the context.
301 print '-' x 80; # print row of dashes
303 print "\t" x ($tab/8), ' ' x ($tab%8); # tab over
305 @ones = (1) x 80; # a list of 80 1's
306 @ones = (5) x @ones; # set all elements to 5
309 =head2 Additive Operators
310 X<operator, additive>
312 Binary "+" returns the sum of two numbers.
315 Binary "-" returns the difference of two numbers.
318 Binary "." concatenates two strings.
319 X<string, concatenation> X<concatenation>
320 X<cat> X<concat> X<concatenate> X<.>
322 =head2 Shift Operators
323 X<shift operator> X<operator, shift> X<<< << >>>
324 X<<< >> >>> X<right shift> X<left shift> X<bitwise shift>
325 X<shl> X<shr> X<shift, right> X<shift, left>
327 Binary "<<" returns the value of its left argument shifted left by the
328 number of bits specified by the right argument. Arguments should be
329 integers. (See also L<Integer Arithmetic>.)
331 Binary ">>" returns the value of its left argument shifted right by
332 the number of bits specified by the right argument. Arguments should
333 be integers. (See also L<Integer Arithmetic>.)
335 Note that both "<<" and ">>" in Perl are implemented directly using
336 "<<" and ">>" in C. If C<use integer> (see L<Integer Arithmetic>) is
337 in force then signed C integers are used, else unsigned C integers are
338 used. Either way, the implementation isn't going to generate results
339 larger than the size of the integer type Perl was built with (32 bits
342 The result of overflowing the range of the integers is undefined
343 because it is undefined also in C. In other words, using 32-bit
344 integers, C<< 1 << 32 >> is undefined. Shifting by a negative number
345 of bits is also undefined.
347 =head2 Named Unary Operators
348 X<operator, named unary>
350 The various named unary operators are treated as functions with one
351 argument, with optional parentheses.
353 If any list operator (print(), etc.) or any unary operator (chdir(), etc.)
354 is followed by a left parenthesis as the next token, the operator and
355 arguments within parentheses are taken to be of highest precedence,
356 just like a normal function call. For example,
357 because named unary operators are higher precedence than ||:
359 chdir $foo || die; # (chdir $foo) || die
360 chdir($foo) || die; # (chdir $foo) || die
361 chdir ($foo) || die; # (chdir $foo) || die
362 chdir +($foo) || die; # (chdir $foo) || die
364 but, because * is higher precedence than named operators:
366 chdir $foo * 20; # chdir ($foo * 20)
367 chdir($foo) * 20; # (chdir $foo) * 20
368 chdir ($foo) * 20; # (chdir $foo) * 20
369 chdir +($foo) * 20; # chdir ($foo * 20)
371 rand 10 * 20; # rand (10 * 20)
372 rand(10) * 20; # (rand 10) * 20
373 rand (10) * 20; # (rand 10) * 20
374 rand +(10) * 20; # rand (10 * 20)
376 Regarding precedence, the filetest operators, like C<-f>, C<-M>, etc. are
377 treated like named unary operators, but they don't follow this functional
378 parenthesis rule. That means, for example, that C<-f($file).".bak"> is
379 equivalent to C<-f "$file.bak">.
380 X<-X> X<filetest> X<operator, filetest>
382 See also L<"Terms and List Operators (Leftward)">.
384 =head2 Relational Operators
385 X<relational operator> X<operator, relational>
387 Binary "<" returns true if the left argument is numerically less than
391 Binary ">" returns true if the left argument is numerically greater
392 than the right argument.
395 Binary "<=" returns true if the left argument is numerically less than
396 or equal to the right argument.
399 Binary ">=" returns true if the left argument is numerically greater
400 than or equal to the right argument.
403 Binary "lt" returns true if the left argument is stringwise less than
407 Binary "gt" returns true if the left argument is stringwise greater
408 than the right argument.
411 Binary "le" returns true if the left argument is stringwise less than
412 or equal to the right argument.
415 Binary "ge" returns true if the left argument is stringwise greater
416 than or equal to the right argument.
419 =head2 Equality Operators
420 X<equality> X<equal> X<equals> X<operator, equality>
422 Binary "==" returns true if the left argument is numerically equal to
426 Binary "!=" returns true if the left argument is numerically not equal
427 to the right argument.
430 Binary "<=>" returns -1, 0, or 1 depending on whether the left
431 argument is numerically less than, equal to, or greater than the right
432 argument. If your platform supports NaNs (not-a-numbers) as numeric
433 values, using them with "<=>" returns undef. NaN is not "<", "==", ">",
434 "<=" or ">=" anything (even NaN), so those 5 return false. NaN != NaN
435 returns true, as does NaN != anything else. If your platform doesn't
436 support NaNs then NaN is just a string with numeric value 0.
437 X<< <=> >> X<spaceship>
439 perl -le '$a = "NaN"; print "No NaN support here" if $a == $a'
440 perl -le '$a = "NaN"; print "NaN support here" if $a != $a'
442 Binary "eq" returns true if the left argument is stringwise equal to
446 Binary "ne" returns true if the left argument is stringwise not equal
447 to the right argument.
450 Binary "cmp" returns -1, 0, or 1 depending on whether the left
451 argument is stringwise less than, equal to, or greater than the right
455 Binary "~~" does a smart match between its arguments. Smart matching
456 is described in L<perlsyn/"Smart matching in detail">.
459 "lt", "le", "ge", "gt" and "cmp" use the collation (sort) order specified
460 by the current locale if C<use locale> is in effect. See L<perllocale>.
463 X<operator, bitwise, and> X<bitwise and> X<&>
465 Binary "&" returns its operands ANDed together bit by bit.
466 (See also L<Integer Arithmetic> and L<Bitwise String Operators>.)
468 Note that "&" has lower priority than relational operators, so for example
469 the brackets are essential in a test like
471 print "Even\n" if ($x & 1) == 0;
473 =head2 Bitwise Or and Exclusive Or
474 X<operator, bitwise, or> X<bitwise or> X<|> X<operator, bitwise, xor>
477 Binary "|" returns its operands ORed together bit by bit.
478 (See also L<Integer Arithmetic> and L<Bitwise String Operators>.)
480 Binary "^" returns its operands XORed together bit by bit.
481 (See also L<Integer Arithmetic> and L<Bitwise String Operators>.)
483 Note that "|" and "^" have lower priority than relational operators, so
484 for example the brackets are essential in a test like
486 print "false\n" if (8 | 2) != 10;
488 =head2 C-style Logical And
489 X<&&> X<logical and> X<operator, logical, and>
491 Binary "&&" performs a short-circuit logical AND operation. That is,
492 if the left operand is false, the right operand is not even evaluated.
493 Scalar or list context propagates down to the right operand if it
496 =head2 C-style Logical Or
497 X<||> X<operator, logical, or>
499 Binary "||" performs a short-circuit logical OR operation. That is,
500 if the left operand is true, the right operand is not even evaluated.
501 Scalar or list context propagates down to the right operand if it
504 =head2 C-style Logical Defined-Or
505 X<//> X<operator, logical, defined-or>
507 Although it has no direct equivalent in C, Perl's C<//> operator is related
508 to its C-style or. In fact, it's exactly the same as C<||>, except that it
509 tests the left hand side's definedness instead of its truth. Thus, C<$a // $b>
510 is similar to C<defined($a) || $b> (except that it returns the value of C<$a>
511 rather than the value of C<defined($a)>) and is exactly equivalent to
512 C<defined($a) ? $a : $b>. This is very useful for providing default values
513 for variables. If you actually want to test if at least one of C<$a> and
514 C<$b> is defined, use C<defined($a // $b)>.
516 The C<||>, C<//> and C<&&> operators return the last value evaluated
517 (unlike C's C<||> and C<&&>, which return 0 or 1). Thus, a reasonably
518 portable way to find out the home directory might be:
520 $home = $ENV{'HOME'} // $ENV{'LOGDIR'} //
521 (getpwuid($<))[7] // die "You're homeless!\n";
523 In particular, this means that you shouldn't use this
524 for selecting between two aggregates for assignment:
526 @a = @b || @c; # this is wrong
527 @a = scalar(@b) || @c; # really meant this
528 @a = @b ? @b : @c; # this works fine, though
530 As more readable alternatives to C<&&> and C<||> when used for
531 control flow, Perl provides the C<and> and C<or> operators (see below).
532 The short-circuit behavior is identical. The precedence of "and"
533 and "or" is much lower, however, so that you can safely use them after a
534 list operator without the need for parentheses:
536 unlink "alpha", "beta", "gamma"
537 or gripe(), next LINE;
539 With the C-style operators that would have been written like this:
541 unlink("alpha", "beta", "gamma")
542 || (gripe(), next LINE);
544 Using "or" for assignment is unlikely to do what you want; see below.
546 =head2 Range Operators
547 X<operator, range> X<range> X<..> X<...>
549 Binary ".." is the range operator, which is really two different
550 operators depending on the context. In list context, it returns a
551 list of values counting (up by ones) from the left value to the right
552 value. If the left value is greater than the right value then it
553 returns the empty list. The range operator is useful for writing
554 C<foreach (1..10)> loops and for doing slice operations on arrays. In
555 the current implementation, no temporary array is created when the
556 range operator is used as the expression in C<foreach> loops, but older
557 versions of Perl might burn a lot of memory when you write something
560 for (1 .. 1_000_000) {
564 The range operator also works on strings, using the magical
565 auto-increment, see below.
567 In scalar context, ".." returns a boolean value. The operator is
568 bistable, like a flip-flop, and emulates the line-range (comma)
569 operator of B<sed>, B<awk>, and various editors. Each ".." operator
570 maintains its own boolean state, even across calls to a subroutine
571 that contains it. It is false as long as its left operand is false.
572 Once the left operand is true, the range operator stays true until the
573 right operand is true, I<AFTER> which the range operator becomes false
574 again. It doesn't become false till the next time the range operator
575 is evaluated. It can test the right operand and become false on the
576 same evaluation it became true (as in B<awk>), but it still returns
577 true once. If you don't want it to test the right operand until the
578 next evaluation, as in B<sed>, just use three dots ("...") instead of
579 two. In all other regards, "..." behaves just like ".." does.
581 The right operand is not evaluated while the operator is in the
582 "false" state, and the left operand is not evaluated while the
583 operator is in the "true" state. The precedence is a little lower
584 than || and &&. The value returned is either the empty string for
585 false, or a sequence number (beginning with 1) for true. The sequence
586 number is reset for each range encountered. The final sequence number
587 in a range has the string "E0" appended to it, which doesn't affect
588 its numeric value, but gives you something to search for if you want
589 to exclude the endpoint. You can exclude the beginning point by
590 waiting for the sequence number to be greater than 1.
592 If either operand of scalar ".." is a constant expression,
593 that operand is considered true if it is equal (C<==>) to the current
594 input line number (the C<$.> variable).
596 To be pedantic, the comparison is actually C<int(EXPR) == int(EXPR)>,
597 but that is only an issue if you use a floating point expression; when
598 implicitly using C<$.> as described in the previous paragraph, the
599 comparison is C<int(EXPR) == int($.)> which is only an issue when C<$.>
600 is set to a floating point value and you are not reading from a file.
601 Furthermore, C<"span" .. "spat"> or C<2.18 .. 3.14> will not do what
602 you want in scalar context because each of the operands are evaluated
603 using their integer representation.
607 As a scalar operator:
609 if (101 .. 200) { print; } # print 2nd hundred lines, short for
610 # if ($. == 101 .. $. == 200) { print; }
612 next LINE if (1 .. /^$/); # skip header lines, short for
613 # next LINE if ($. == 1 .. /^$/);
614 # (typically in a loop labeled LINE)
616 s/^/> / if (/^$/ .. eof()); # quote body
618 # parse mail messages
620 $in_header = 1 .. /^$/;
621 $in_body = /^$/ .. eof;
628 close ARGV if eof; # reset $. each file
631 Here's a simple example to illustrate the difference between
632 the two range operators:
645 This program will print only the line containing "Bar". If
646 the range operator is changed to C<...>, it will also print the
649 And now some examples as a list operator:
651 for (101 .. 200) { print; } # print $_ 100 times
652 @foo = @foo[0 .. $#foo]; # an expensive no-op
653 @foo = @foo[$#foo-4 .. $#foo]; # slice last 5 items
655 The range operator (in list context) makes use of the magical
656 auto-increment algorithm if the operands are strings. You
659 @alphabet = ('A' .. 'Z');
661 to get all normal letters of the English alphabet, or
663 $hexdigit = (0 .. 9, 'a' .. 'f')[$num & 15];
665 to get a hexadecimal digit, or
667 @z2 = ('01' .. '31'); print $z2[$mday];
669 to get dates with leading zeros.
671 If the final value specified is not in the sequence that the magical
672 increment would produce, the sequence goes until the next value would
673 be longer than the final value specified.
675 If the initial value specified isn't part of a magical increment
676 sequence (that is, a non-empty string matching "/^[a-zA-Z]*[0-9]*\z/"),
677 only the initial value will be returned. So the following will only
680 use charnames 'greek';
681 my @greek_small = ("\N{alpha}" .. "\N{omega}");
683 To get lower-case greek letters, use this instead:
685 my @greek_small = map { chr } ( ord("\N{alpha}") ..
688 Because each operand is evaluated in integer form, C<2.18 .. 3.14> will
689 return two elements in list context.
691 @list = (2.18 .. 3.14); # same as @list = (2 .. 3);
693 =head2 Conditional Operator
694 X<operator, conditional> X<operator, ternary> X<ternary> X<?:>
696 Ternary "?:" is the conditional operator, just as in C. It works much
697 like an if-then-else. If the argument before the ? is true, the
698 argument before the : is returned, otherwise the argument after the :
699 is returned. For example:
701 printf "I have %d dog%s.\n", $n,
702 ($n == 1) ? '' : "s";
704 Scalar or list context propagates downward into the 2nd
705 or 3rd argument, whichever is selected.
707 $a = $ok ? $b : $c; # get a scalar
708 @a = $ok ? @b : @c; # get an array
709 $a = $ok ? @b : @c; # oops, that's just a count!
711 The operator may be assigned to if both the 2nd and 3rd arguments are
712 legal lvalues (meaning that you can assign to them):
714 ($a_or_b ? $a : $b) = $c;
716 Because this operator produces an assignable result, using assignments
717 without parentheses will get you in trouble. For example, this:
719 $a % 2 ? $a += 10 : $a += 2
723 (($a % 2) ? ($a += 10) : $a) += 2
727 ($a % 2) ? ($a += 10) : ($a += 2)
729 That should probably be written more simply as:
731 $a += ($a % 2) ? 10 : 2;
733 =head2 Assignment Operators
734 X<assignment> X<operator, assignment> X<=> X<**=> X<+=> X<*=> X<&=>
735 X<<< <<= >>> X<&&=> X<-=> X</=> X<|=> X<<< >>= >>> X<||=> X<//=> X<.=>
738 "=" is the ordinary assignment operator.
740 Assignment operators work as in C. That is,
748 although without duplicating any side effects that dereferencing the lvalue
749 might trigger, such as from tie(). Other assignment operators work similarly.
750 The following are recognized:
757 Although these are grouped by family, they all have the precedence
760 Unlike in C, the scalar assignment operator produces a valid lvalue.
761 Modifying an assignment is equivalent to doing the assignment and
762 then modifying the variable that was assigned to. This is useful
763 for modifying a copy of something, like this:
765 ($tmp = $global) =~ tr [A-Z] [a-z];
776 Similarly, a list assignment in list context produces the list of
777 lvalues assigned to, and a list assignment in scalar context returns
778 the number of elements produced by the expression on the right hand
779 side of the assignment.
781 =head2 Comma Operator
782 X<comma> X<operator, comma> X<,>
784 Binary "," is the comma operator. In scalar context it evaluates
785 its left argument, throws that value away, then evaluates its right
786 argument and returns that value. This is just like C's comma operator.
788 In list context, it's just the list argument separator, and inserts
789 both its arguments into the list. These arguments are also evaluated
792 The C<< => >> operator is a synonym for the comma except that it causes
793 its left operand to be interpreted as a string if it begins with a letter
794 or underscore and is composed only of letters, digits and underscores.
795 This includes operands that might otherwise be interpreted as operators,
796 constants, single number v-strings or function calls. If in doubt about
797 this behaviour, the left operand can be quoted explicitly.
799 Otherwise, the C<< => >> operator behaves exactly as the comma operator
800 or list argument separator, according to context.
804 use constant FOO => "something";
806 my %h = ( FOO => 23 );
814 my %h = ("something", 23);
816 The C<< => >> operator is helpful in documenting the correspondence
817 between keys and values in hashes, and other paired elements in lists.
819 %hash = ( $key => $value );
820 login( $username => $password );
822 =head2 Yada Yada Operator
823 X<...> X<... operator> X<yada yada operator>
825 The yada yada operator (noted C<...>) is a placeholder for code. Perl
826 parses it without error, but when you try to execute a yada yada, it
827 throws an exception with the text C<Unimplemented>:
829 sub unimplemented { ... }
831 eval { unimplemented() };
832 if( $@ eq 'Unimplemented' ) {
833 print "I found the yada yada!\n";
836 You can only use the yada yada to stand in for a complete statement.
837 These examples of the yada yada work:
853 do { my $n; ...; print 'Hurrah!' };
855 The yada yada cannot stand in for an expression that is part of a
856 larger statement since the C<...> is also the three-dot version of the
857 range operator (see L<Range Operators>). These examples of the yada
858 yada are still syntax errors:
862 open my($fh), '>', '/dev/passwd' or ...;
864 if( $condition && ... ) { print "Hello\n" };
866 There are some cases where Perl can't immediately tell the difference
867 between an expression and a statement. For instance, the syntax for a
868 block and an anonymous hash reference constructor look the same unless
869 there's something in the braces that give Perl a hint. The yada yada
870 is a syntax error if Perl doesn't guess that the C<{ ... }> is a
871 block. In that case, it doesn't think the C<...> is the yada yada
872 because it's expecting an expression instead of a statement:
874 my @transformed = map { ... } @input; # syntax error
876 You can use a C<;> inside your block to denote that the C<{ ... }> is
877 a block and not a hash reference constructor. Now the yada yada works:
879 my @transformed = map {; ... } @input; # ; disambiguates
881 my @transformed = map { ...; } @input; # ; disambiguates
883 =head2 List Operators (Rightward)
884 X<operator, list, rightward> X<list operator>
886 On the right side of a list operator, it has very low precedence,
887 such that it controls all comma-separated expressions found there.
888 The only operators with lower precedence are the logical operators
889 "and", "or", and "not", which may be used to evaluate calls to list
890 operators without the need for extra parentheses:
892 open HANDLE, "filename"
893 or die "Can't open: $!\n";
895 See also discussion of list operators in L<Terms and List Operators (Leftward)>.
898 X<operator, logical, not> X<not>
900 Unary "not" returns the logical negation of the expression to its right.
901 It's the equivalent of "!" except for the very low precedence.
904 X<operator, logical, and> X<and>
906 Binary "and" returns the logical conjunction of the two surrounding
907 expressions. It's equivalent to && except for the very low
908 precedence. This means that it short-circuits: i.e., the right
909 expression is evaluated only if the left expression is true.
911 =head2 Logical or, Defined or, and Exclusive Or
912 X<operator, logical, or> X<operator, logical, xor>
913 X<operator, logical, defined or> X<operator, logical, exclusive or>
916 Binary "or" returns the logical disjunction of the two surrounding
917 expressions. It's equivalent to || except for the very low precedence.
918 This makes it useful for control flow
920 print FH $data or die "Can't write to FH: $!";
922 This means that it short-circuits: i.e., the right expression is evaluated
923 only if the left expression is false. Due to its precedence, you should
924 probably avoid using this for assignment, only for control flow.
926 $a = $b or $c; # bug: this is wrong
927 ($a = $b) or $c; # really means this
928 $a = $b || $c; # better written this way
930 However, when it's a list-context assignment and you're trying to use
931 "||" for control flow, you probably need "or" so that the assignment
932 takes higher precedence.
934 @info = stat($file) || die; # oops, scalar sense of stat!
935 @info = stat($file) or die; # better, now @info gets its due
937 Then again, you could always use parentheses.
939 Binary "xor" returns the exclusive-OR of the two surrounding expressions.
940 It cannot short circuit, of course.
942 =head2 C Operators Missing From Perl
943 X<operator, missing from perl> X<&> X<*>
944 X<typecasting> X<(TYPE)>
946 Here is what C has that Perl doesn't:
952 Address-of operator. (But see the "\" operator for taking a reference.)
956 Dereference-address operator. (Perl's prefix dereferencing
957 operators are typed: $, @, %, and &.)
961 Type-casting operator.
965 =head2 Quote and Quote-like Operators
966 X<operator, quote> X<operator, quote-like> X<q> X<qq> X<qx> X<qw> X<m>
967 X<qr> X<s> X<tr> X<'> X<''> X<"> X<""> X<//> X<`> X<``> X<<< << >>>
968 X<escape sequence> X<escape>
971 While we usually think of quotes as literal values, in Perl they
972 function as operators, providing various kinds of interpolating and
973 pattern matching capabilities. Perl provides customary quote characters
974 for these behaviors, but also provides a way for you to choose your
975 quote character for any of them. In the following table, a C<{}> represents
976 any pair of delimiters you choose.
978 Customary Generic Meaning Interpolates
983 // m{} Pattern match yes*
985 s{}{} Substitution yes*
986 tr{}{} Transliteration no (but see below)
989 * unless the delimiter is ''.
991 Non-bracketing delimiters use the same character fore and aft, but the four
992 sorts of brackets (round, angle, square, curly) will all nest, which means
1001 Note, however, that this does not always work for quoting Perl code:
1003 $s = q{ if($a eq "}") ... }; # WRONG
1005 is a syntax error. The C<Text::Balanced> module (from CPAN, and
1006 starting from Perl 5.8 part of the standard distribution) is able
1007 to do this properly.
1009 There can be whitespace between the operator and the quoting
1010 characters, except when C<#> is being used as the quoting character.
1011 C<q#foo#> is parsed as the string C<foo>, while C<q #foo#> is the
1012 operator C<q> followed by a comment. Its argument will be taken
1013 from the next line. This allows you to write:
1015 s {foo} # Replace foo
1018 The following escape sequences are available in constructs that interpolate
1019 and in transliterations.
1020 X<\t> X<\n> X<\r> X<\f> X<\b> X<\a> X<\e> X<\x> X<\0> X<\c> X<\N> X<\N{}>
1023 Sequence Note Description
1029 \a alarm (bell) (BEL)
1031 \x{263a} [1,8] hex char (example: SMILEY)
1032 \x1b [2,8] restricted range hex char (example: ESC)
1033 \N{name} [3] named Unicode character or character sequence
1034 \N{U+263D} [4,8] Unicode character (example: FIRST QUARTER MOON)
1035 \c[ [5] control char (example: chr(27))
1036 \o{23072} [6,8] octal char (example: SMILEY)
1037 \033 [7,8] restricted range octal char (example: ESC)
1043 The result is the character specified by the hexadecimal number between
1044 the braces. See L</[8]> below for details on which character.
1046 Only hexadecimal digits are valid between the braces. If an invalid
1047 character is encountered, a warning will be issued and the invalid
1048 character and all subsequent characters (valid or invalid) within the
1049 braces will be discarded.
1051 If there are no valid digits between the braces, the generated character is
1052 the NULL character (C<\x{00}>). However, an explicit empty brace (C<\x{}>)
1053 will not cause a warning.
1057 The result is the character specified by the hexadecimal number in the range
1058 0x00 to 0xFF. See L</[8]> below for details on which character.
1060 Only hexadecimal digits are valid following C<\x>. When C<\x> is followed
1061 by fewer than two valid digits, any valid digits will be zero-padded. This
1062 means that C<\x7> will be interpreted as C<\x07> and C<\x> alone will be
1063 interpreted as C<\x00>. Except at the end of a string, having fewer than
1064 two valid digits will result in a warning. Note that while the warning
1065 says the illegal character is ignored, it is only ignored as part of the
1066 escape and will still be used as the subsequent character in the string.
1069 Original Result Warns?
1077 The result is the Unicode character or character sequence given by I<name>.
1082 C<\N{U+I<hexadecimal number>}> means the Unicode character whose Unicode code
1083 point is I<hexadecimal number>.
1087 The character following C<\c> is mapped to some other character as shown in the
1104 Also, C<\c\I<X>> yields C< chr(28) . "I<X>"> for any I<X>, but cannot come at the
1105 end of a string, because the backslash would be parsed as escaping the end
1108 On ASCII platforms, the resulting characters from the list above are the
1109 complete set of ASCII controls. This isn't the case on EBCDIC platforms; see
1110 L<perlebcdic/OPERATOR DIFFERENCES> for the complete list of what these
1111 sequences mean on both ASCII and EBCDIC platforms.
1113 Use of any other character following the "c" besides those listed above is
1114 discouraged, and may become deprecated or forbidden. What happens for those
1115 other characters currently though, is that the value is derived by inverting
1118 To get platform independent controls, you can use C<\N{...}>.
1122 The result is the character specified by the octal number between the braces.
1123 See L</[8]> below for details on which character.
1125 If a character that isn't an octal digit is encountered, a warning is raised,
1126 and the value is based on the octal digits before it, discarding it and all
1127 following characters up to the closing brace. It is a fatal error if there are
1128 no octal digits at all.
1132 The result is the character specified by the three digit octal number in the
1133 range 000 to 777 (but best to not use above 077, see next paragraph). See
1134 L</[8]> below for details on which character.
1136 Some contexts allow 2 or even 1 digit, but any usage without exactly
1137 three digits, the first being a zero, may give unintended results. (For
1138 example, see L<perlrebackslash/Octal escapes>.) Starting in Perl 5.14, you may
1139 use C<\o{}> instead which avoids all these problems. Otherwise, it is best to
1140 use this construct only for ordinals C<\077> and below, remembering to pad to
1141 the left with zeros to make three digits. For larger ordinals, either use
1142 C<\o{}> , or convert to someething else, such as to hex and use C<\x{}>
1145 Having fewer than 3 digits may lead to a misleading warning message that says
1146 that what follows is ignored. For example, C<"\128"> in the ASCII character set
1147 is equivalent to the two characters C<"\n8">, but the warning C<Illegal octal
1148 digit '8' ignored> will be thrown. To avoid this warning, make sure to pad
1149 your octal number with C<0>s: C<"\0128">.
1153 Several of the constructs above specify a character by a number. That number
1154 gives the character's position in the character set encoding (indexed from 0).
1155 This is called synonymously its ordinal, code position, or code point). Perl
1156 works on platforms that have a native encoding currently of either ASCII/Latin1
1157 or EBCDIC, each of which allow specification of 256 characters. In general, if
1158 the number is 255 (0xFF, 0377) or below, Perl interprets this in the platform's
1159 native encoding. If the number is 256 (0x100, 0400) or above, Perl interprets
1160 it as as a Unicode code point and the result is the corresponding Unicode
1161 character. For example C<\x{50}> and C<\o{120}> both are the number 80 in
1162 decimal, which is less than 256, so the number is interpreted in the native
1163 character set encoding. In ASCII the character in the 80th position (indexed
1164 from 0) is the letter "P", and in EBCDIC it is the ampersand symbol "&".
1165 C<\x{100}> and C<\o{400}> are both 256 in decimal, so the number is interpreted
1166 as a Unicode code point no matter what the native encoding is. The name of the
1167 character in the 100th position (indexed by 0) in Unicode is
1168 C<LATIN CAPITAL LETTER A WITH MACRON>.
1170 There are a couple of exceptions to the above rule. C<\N{U+I<hex number>}> is
1171 always interpreted as a Unicode code point, so that C<\N{U+0050}> is "P" even
1172 on EBCDIC platforms. And if L<C<S<use encoding>>|encoding> is in effect, the
1173 number is considered to be in that encoding, and is translated from that into
1174 the platform's native encoding if there is a corresponding native character;
1175 otherwise to Unicode.
1179 B<NOTE>: Unlike C and other languages, Perl has no C<\v> escape sequence for
1180 the vertical tab (VT - ASCII 11), but you may use C<\ck> or C<\x0b>. (C<\v>
1181 does have meaning in regular expression patterns in Perl, see L<perlre>.)
1183 The following escape sequences are available in constructs that interpolate,
1184 but not in transliterations.
1185 X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q>
1187 \l lowercase next char
1188 \u uppercase next char
1189 \L lowercase till \E
1190 \U uppercase till \E
1191 \E end case modification
1192 \Q quote non-word characters till \E
1194 If C<use locale> is in effect, the case map used by C<\l>, C<\L>,
1195 C<\u> and C<\U> is taken from the current locale. See L<perllocale>.
1196 If Unicode (for example, C<\N{}> or wide hex characters of 0x100 or
1197 beyond) is being used, the case map used by C<\l>, C<\L>, C<\u> and
1198 C<\U> is as defined by Unicode.
1200 All systems use the virtual C<"\n"> to represent a line terminator,
1201 called a "newline". There is no such thing as an unvarying, physical
1202 newline character. It is only an illusion that the operating system,
1203 device drivers, C libraries, and Perl all conspire to preserve. Not all
1204 systems read C<"\r"> as ASCII CR and C<"\n"> as ASCII LF. For example,
1205 on a Mac, these are reversed, and on systems without line terminator,
1206 printing C<"\n"> may emit no actual data. In general, use C<"\n"> when
1207 you mean a "newline" for your system, but use the literal ASCII when you
1208 need an exact character. For example, most networking protocols expect
1209 and prefer a CR+LF (C<"\015\012"> or C<"\cM\cJ">) for line terminators,
1210 and although they often accept just C<"\012">, they seldom tolerate just
1211 C<"\015">. If you get in the habit of using C<"\n"> for networking,
1212 you may be burned some day.
1213 X<newline> X<line terminator> X<eol> X<end of line>
1216 For constructs that do interpolate, variables beginning with "C<$>"
1217 or "C<@>" are interpolated. Subscripted variables such as C<$a[3]> or
1218 C<< $href->{key}[0] >> are also interpolated, as are array and hash slices.
1219 But method calls such as C<< $obj->meth >> are not.
1221 Interpolating an array or slice interpolates the elements in order,
1222 separated by the value of C<$">, so is equivalent to interpolating
1223 C<join $", @array>. "Punctuation" arrays such as C<@*> are only
1224 interpolated if the name is enclosed in braces C<@{*}>, but special
1225 arrays C<@_>, C<@+>, and C<@-> are interpolated, even without braces.
1227 For double-quoted strings, the quoting from C<\Q> is applied after
1228 interpolation and escapes are processed.
1230 "abc\Qfoo\tbar$s\Exyz"
1234 "abc" . quotemeta("foo\tbar$s") . "xyz"
1236 For the pattern of regex operators (C<qr//>, C<m//> and C<s///>),
1237 the quoting from C<\Q> is applied after interpolation is processed,
1238 but before escapes are processed. This allows the pattern to match
1239 literally (except for C<$> and C<@>). For example, the following matches:
1243 Because C<$> or C<@> trigger interpolation, you'll need to use something
1244 like C</\Quser\E\@\Qhost/> to match them literally.
1246 Patterns are subject to an additional level of interpretation as a
1247 regular expression. This is done as a second pass, after variables are
1248 interpolated, so that regular expressions may be incorporated into the
1249 pattern from the variables. If this is not what you want, use C<\Q> to
1250 interpolate a variable literally.
1252 Apart from the behavior described above, Perl does not expand
1253 multiple levels of interpolation. In particular, contrary to the
1254 expectations of shell programmers, back-quotes do I<NOT> interpolate
1255 within double quotes, nor do single quotes impede evaluation of
1256 variables when used within double quotes.
1258 =head2 Regexp Quote-Like Operators
1261 Here are the quote-like operators that apply to pattern
1262 matching and related activities.
1266 =item qr/STRING/msixpo
1267 X<qr> X</i> X</m> X</o> X</s> X</x> X</p>
1269 This operator quotes (and possibly compiles) its I<STRING> as a regular
1270 expression. I<STRING> is interpolated the same way as I<PATTERN>
1271 in C<m/PATTERN/>. If "'" is used as the delimiter, no interpolation
1272 is done. Returns a Perl value which may be used instead of the
1273 corresponding C</STRING/msixpo> expression. The returned value is a
1274 normalized version of the original pattern. It magically differs from
1275 a string containing the same characters: C<ref(qr/x/)> returns "Regexp",
1276 even though dereferencing the result returns undef.
1280 $rex = qr/my.STRING/is;
1281 print $rex; # prints (?si-xm:my.STRING)
1288 The result may be used as a subpattern in a match:
1291 $string =~ /foo${re}bar/; # can be interpolated in other patterns
1292 $string =~ $re; # or used standalone
1293 $string =~ /$re/; # or this way
1295 Since Perl may compile the pattern at the moment of execution of qr()
1296 operator, using qr() may have speed advantages in some situations,
1297 notably if the result of qr() is used standalone:
1300 my $patterns = shift;
1301 my @compiled = map qr/$_/i, @$patterns;
1304 foreach my $pat (@compiled) {
1305 $success = 1, last if /$pat/;
1311 Precompilation of the pattern into an internal representation at
1312 the moment of qr() avoids a need to recompile the pattern every
1313 time a match C</$pat/> is attempted. (Perl has many other internal
1314 optimizations, but none would be triggered in the above example if
1315 we did not use qr() operator.)
1319 m Treat string as multiple lines.
1320 s Treat string as single line. (Make . match a newline)
1321 i Do case-insensitive pattern matching.
1322 x Use extended regular expressions.
1323 p When matching preserve a copy of the matched string so
1324 that ${^PREMATCH}, ${^MATCH}, ${^POSTMATCH} will be defined.
1325 o Compile pattern only once.
1327 If a precompiled pattern is embedded in a larger pattern then the effect
1328 of 'msixp' will be propagated appropriately. The effect of the 'o'
1329 modifier has is not propagated, being restricted to those patterns
1330 explicitly using it.
1332 See L<perlre> for additional information on valid syntax for STRING, and
1333 for a detailed look at the semantics of regular expressions.
1335 =item m/PATTERN/msixpogc
1336 X<m> X<operator, match>
1337 X<regexp, options> X<regexp> X<regex, options> X<regex>
1338 X</m> X</s> X</i> X</x> X</p> X</o> X</g> X</c>
1340 =item /PATTERN/msixpogc
1342 Searches a string for a pattern match, and in scalar context returns
1343 true if it succeeds, false if it fails. If no string is specified
1344 via the C<=~> or C<!~> operator, the $_ string is searched. (The
1345 string specified with C<=~> need not be an lvalue--it may be the
1346 result of an expression evaluation, but remember the C<=~> binds
1347 rather tightly.) See also L<perlre>. See L<perllocale> for
1348 discussion of additional considerations that apply when C<use locale>
1351 Options are as described in C<qr//>; in addition, the following match
1352 process modifiers are available:
1354 g Match globally, i.e., find all occurrences.
1355 c Do not reset search position on a failed match when /g is in effect.
1357 If "/" is the delimiter then the initial C<m> is optional. With the C<m>
1358 you can use any pair of non-whitespace characters
1359 as delimiters. This is particularly useful for matching path names
1360 that contain "/", to avoid LTS (leaning toothpick syndrome). If "?" is
1361 the delimiter, then the match-only-once rule of C<?PATTERN?> applies.
1362 If "'" is the delimiter, no interpolation is performed on the PATTERN.
1363 When using a character valid in an identifier, whitespace is required
1366 PATTERN may contain variables, which will be interpolated (and the
1367 pattern recompiled) every time the pattern search is evaluated, except
1368 for when the delimiter is a single quote. (Note that C<$(>, C<$)>, and
1369 C<$|> are not interpolated because they look like end-of-string tests.)
1370 If you want such a pattern to be compiled only once, add a C</o> after
1371 the trailing delimiter. This avoids expensive run-time recompilations,
1372 and is useful when the value you are interpolating won't change over
1373 the life of the script. However, mentioning C</o> constitutes a promise
1374 that you won't change the variables in the pattern. If you change them,
1375 Perl won't even notice. See also L<"qr/STRING/msixpo">.
1377 =item The empty pattern //
1379 If the PATTERN evaluates to the empty string, the last
1380 I<successfully> matched regular expression is used instead. In this
1381 case, only the C<g> and C<c> flags on the empty pattern is honoured -
1382 the other flags are taken from the original pattern. If no match has
1383 previously succeeded, this will (silently) act instead as a genuine
1384 empty pattern (which will always match).
1386 Note that it's possible to confuse Perl into thinking C<//> (the empty
1387 regex) is really C<//> (the defined-or operator). Perl is usually pretty
1388 good about this, but some pathological cases might trigger this, such as
1389 C<$a///> (is that C<($a) / (//)> or C<$a // />?) and C<print $fh //>
1390 (C<print $fh(//> or C<print($fh //>?). In all of these examples, Perl
1391 will assume you meant defined-or. If you meant the empty regex, just
1392 use parentheses or spaces to disambiguate, or even prefix the empty
1393 regex with an C<m> (so C<//> becomes C<m//>).
1395 =item Matching in list context
1397 If the C</g> option is not used, C<m//> in list context returns a
1398 list consisting of the subexpressions matched by the parentheses in the
1399 pattern, i.e., (C<$1>, C<$2>, C<$3>...). (Note that here C<$1> etc. are
1400 also set, and that this differs from Perl 4's behavior.) When there are
1401 no parentheses in the pattern, the return value is the list C<(1)> for
1402 success. With or without parentheses, an empty list is returned upon
1407 open(TTY, '/dev/tty');
1408 <TTY> =~ /^y/i && foo(); # do foo if desired
1410 if (/Version: *([0-9.]*)/) { $version = $1; }
1412 next if m#^/usr/spool/uucp#;
1417 print if /$arg/o; # compile only once
1420 if (($F1, $F2, $Etc) = ($foo =~ /^(\S+)\s+(\S+)\s*(.*)/))
1422 This last example splits $foo into the first two words and the
1423 remainder of the line, and assigns those three fields to $F1, $F2, and
1424 $Etc. The conditional is true if any variables were assigned, i.e., if
1425 the pattern matched.
1427 The C</g> modifier specifies global pattern matching--that is,
1428 matching as many times as possible within the string. How it behaves
1429 depends on the context. In list context, it returns a list of the
1430 substrings matched by any capturing parentheses in the regular
1431 expression. If there are no parentheses, it returns a list of all
1432 the matched strings, as if there were parentheses around the whole
1435 In scalar context, each execution of C<m//g> finds the next match,
1436 returning true if it matches, and false if there is no further match.
1437 The position after the last match can be read or set using the C<pos()>
1438 function; see L<perlfunc/pos>. A failed match normally resets the
1439 search position to the beginning of the string, but you can avoid that
1440 by adding the C</c> modifier (e.g. C<m//gc>). Modifying the target
1441 string also resets the search position.
1445 You can intermix C<m//g> matches with C<m/\G.../g>, where C<\G> is a
1446 zero-width assertion that matches the exact position where the
1447 previous C<m//g>, if any, left off. Without the C</g> modifier, the
1448 C<\G> assertion still anchors at C<pos()> as it was at the start of
1449 the operation (see L<perlfunc/pos>), but the match is of course only
1450 attempted once. Using C<\G> without C</g> on a target string that has
1451 not previously had a C</g> match applied to it is the same as using
1452 the C<\A> assertion to match the beginning of the string. Note also
1453 that, currently, C<\G> is only properly supported when anchored at the
1454 very beginning of the pattern.
1459 ($one,$five,$fifteen) = (`uptime` =~ /(\d+\.\d+)/g);
1463 while (defined($paragraph = <>)) {
1464 while ($paragraph =~ /[a-z]['")]*[.!?]+['")]*\s/g) {
1468 print "$sentences\n";
1470 # using m//gc with \G
1474 print $1 while /(o)/gc; print "', pos=", pos, "\n";
1476 print $1 if /\G(q)/gc; print "', pos=", pos, "\n";
1478 print $1 while /(p)/gc; print "', pos=", pos, "\n";
1480 print "Final: '$1', pos=",pos,"\n" if /\G(.)/;
1482 The last example should print:
1492 Notice that the final match matched C<q> instead of C<p>, which a match
1493 without the C<\G> anchor would have done. Also note that the final match
1494 did not update C<pos>. C<pos> is only updated on a C</g> match. If the
1495 final match did indeed match C<p>, it's a good bet that you're running an
1496 older (pre-5.6.0) Perl.
1498 A useful idiom for C<lex>-like scanners is C</\G.../gc>. You can
1499 combine several regexps like this to process a string part-by-part,
1500 doing different actions depending on which regexp matched. Each
1501 regexp tries to match where the previous one leaves off.
1504 $url = URI::URL->new( "http://example.com/" ); die if $url eq "xXx";
1508 print(" digits"), redo LOOP if /\G\d+\b[,.;]?\s*/gc;
1509 print(" lowercase"), redo LOOP if /\G[a-z]+\b[,.;]?\s*/gc;
1510 print(" UPPERCASE"), redo LOOP if /\G[A-Z]+\b[,.;]?\s*/gc;
1511 print(" Capitalized"), redo LOOP if /\G[A-Z][a-z]+\b[,.;]?\s*/gc;
1512 print(" MiXeD"), redo LOOP if /\G[A-Za-z]+\b[,.;]?\s*/gc;
1513 print(" alphanumeric"), redo LOOP if /\G[A-Za-z0-9]+\b[,.;]?\s*/gc;
1514 print(" line-noise"), redo LOOP if /\G[^A-Za-z0-9]+/gc;
1515 print ". That's all!\n";
1518 Here is the output (split into several lines):
1520 line-noise lowercase line-noise UPPERCASE line-noise UPPERCASE
1521 line-noise lowercase line-noise lowercase line-noise lowercase
1522 lowercase line-noise lowercase lowercase line-noise lowercase
1523 lowercase line-noise MiXeD line-noise. That's all!
1528 This is just like the C</pattern/> search, except that it matches only
1529 once between calls to the reset() operator. This is a useful
1530 optimization when you want to see only the first occurrence of
1531 something in each file of a set of files, for instance. Only C<??>
1532 patterns local to the current package are reset.
1536 # blank line between header and body
1539 reset if eof; # clear ?? status for next file
1542 This usage is vaguely deprecated, which means it just might possibly
1543 be removed in some distant future version of Perl, perhaps somewhere
1544 around the year 2168.
1546 =item s/PATTERN/REPLACEMENT/msixpogcer
1547 X<substitute> X<substitution> X<replace> X<regexp, replace>
1548 X<regexp, substitute> X</m> X</s> X</i> X</x> X</p> X</o> X</g> X</c> X</e> X</r>
1550 Searches a string for a pattern, and if found, replaces that pattern
1551 with the replacement text and returns the number of substitutions
1552 made. Otherwise it returns false (specifically, the empty string).
1554 If the C</r> (non-destructive) option is used then it will perform the
1555 substitution on a copy of the string and return the copy whether or not a
1556 substitution occurred. The original string will always remain unchanged in
1557 this case. The copy will always be a plain string, even if the input is an
1558 object or a tied variable.
1560 If no string is specified via the C<=~> or C<!~> operator, the C<$_>
1561 variable is searched and modified. (The string specified with C<=~> must
1562 be scalar variable, an array element, a hash element, or an assignment
1563 to one of those, i.e., an lvalue.)
1565 If the delimiter chosen is a single quote, no interpolation is
1566 done on either the PATTERN or the REPLACEMENT. Otherwise, if the
1567 PATTERN contains a $ that looks like a variable rather than an
1568 end-of-string test, the variable will be interpolated into the pattern
1569 at run-time. If you want the pattern compiled only once the first time
1570 the variable is interpolated, use the C</o> option. If the pattern
1571 evaluates to the empty string, the last successfully executed regular
1572 expression is used instead. See L<perlre> for further explanation on these.
1573 See L<perllocale> for discussion of additional considerations that apply
1574 when C<use locale> is in effect.
1576 Options are as with m// with the addition of the following replacement
1579 e Evaluate the right side as an expression.
1580 ee Evaluate the right side as a string then eval the result.
1581 r Return substitution and leave the original string untouched.
1583 Any non-whitespace delimiter may replace the slashes. Add space after
1584 the C<s> when using a character allowed in identifiers. If single quotes
1585 are used, no interpretation is done on the replacement string (the C</e>
1586 modifier overrides this, however). Unlike Perl 4, Perl 5 treats backticks
1587 as normal delimiters; the replacement text is not evaluated as a command.
1588 If the PATTERN is delimited by bracketing quotes, the REPLACEMENT has
1589 its own pair of quotes, which may or may not be bracketing quotes, e.g.,
1590 C<s(foo)(bar)> or C<< s<foo>/bar/ >>. A C</e> will cause the
1591 replacement portion to be treated as a full-fledged Perl expression
1592 and evaluated right then and there. It is, however, syntax checked at
1593 compile-time. A second C<e> modifier will cause the replacement portion
1594 to be C<eval>ed before being run as a Perl expression.
1598 s/\bgreen\b/mauve/g; # don't change wintergreen
1600 $path =~ s|/usr/bin|/usr/local/bin|;
1602 s/Login: $foo/Login: $bar/; # run-time pattern
1604 ($foo = $bar) =~ s/this/that/; # copy first, then change
1605 ($foo = "$bar") =~ s/this/that/; # convert to string, copy, then change
1606 $foo = $bar =~ s/this/that/r; # Same as above using /r
1607 $foo = $bar =~ s/this/that/r
1608 =~ s/that/the other/r; # Chained substitutes using /r
1609 @foo = map { s/this/that/r } @bar # /r is very useful in maps
1611 $count = ($paragraph =~ s/Mister\b/Mr./g); # get change-count
1614 s/\d+/$&*2/e; # yields 'abc246xyz'
1615 s/\d+/sprintf("%5d",$&)/e; # yields 'abc 246xyz'
1616 s/\w/$& x 2/eg; # yields 'aabbcc 224466xxyyzz'
1618 s/%(.)/$percent{$1}/g; # change percent escapes; no /e
1619 s/%(.)/$percent{$1} || $&/ge; # expr now, so /e
1620 s/^=(\w+)/pod($1)/ge; # use function call
1623 $a = s/abc/def/r; # $a is 'def123xyz' and
1624 # $_ remains 'abc123xyz'.
1626 # expand variables in $_, but dynamics only, using
1627 # symbolic dereferencing
1630 # Add one to the value of any numbers in the string
1633 # This will expand any embedded scalar variable
1634 # (including lexicals) in $_ : First $1 is interpolated
1635 # to the variable name, and then evaluated
1638 # Delete (most) C comments.
1640 /\* # Match the opening delimiter.
1641 .*? # Match a minimal number of characters.
1642 \*/ # Match the closing delimiter.
1645 s/^\s*(.*?)\s*$/$1/; # trim whitespace in $_, expensively
1647 for ($variable) { # trim whitespace in $variable, cheap
1652 s/([^ ]*) *([^ ]*)/$2 $1/; # reverse 1st two fields
1654 Note the use of $ instead of \ in the last example. Unlike
1655 B<sed>, we use the \<I<digit>> form in only the left hand side.
1656 Anywhere else it's $<I<digit>>.
1658 Occasionally, you can't use just a C</g> to get all the changes
1659 to occur that you might want. Here are two common cases:
1661 # put commas in the right places in an integer
1662 1 while s/(\d)(\d\d\d)(?!\d)/$1,$2/g;
1664 # expand tabs to 8-column spacing
1665 1 while s/\t+/' ' x (length($&)*8 - length($`)%8)/e;
1669 =head2 Quote-Like Operators
1670 X<operator, quote-like>
1675 X<q> X<quote, single> X<'> X<''>
1679 A single-quoted, literal string. A backslash represents a backslash
1680 unless followed by the delimiter or another backslash, in which case
1681 the delimiter or backslash is interpolated.
1683 $foo = q!I said, "You said, 'She said it.'"!;
1684 $bar = q('This is it.');
1685 $baz = '\n'; # a two-character string
1688 X<qq> X<quote, double> X<"> X<"">
1692 A double-quoted, interpolated string.
1695 (*** The previous line contains the naughty word "$1".\n)
1696 if /\b(tcl|java|python)\b/i; # :-)
1697 $baz = "\n"; # a one-character string
1700 X<qx> X<`> X<``> X<backtick>
1704 A string which is (possibly) interpolated and then executed as a
1705 system command with C</bin/sh> or its equivalent. Shell wildcards,
1706 pipes, and redirections will be honored. The collected standard
1707 output of the command is returned; standard error is unaffected. In
1708 scalar context, it comes back as a single (potentially multi-line)
1709 string, or undef if the command failed. In list context, returns a
1710 list of lines (however you've defined lines with $/ or
1711 $INPUT_RECORD_SEPARATOR), or an empty list if the command failed.
1713 Because backticks do not affect standard error, use shell file descriptor
1714 syntax (assuming the shell supports this) if you care to address this.
1715 To capture a command's STDERR and STDOUT together:
1717 $output = `cmd 2>&1`;
1719 To capture a command's STDOUT but discard its STDERR:
1721 $output = `cmd 2>/dev/null`;
1723 To capture a command's STDERR but discard its STDOUT (ordering is
1726 $output = `cmd 2>&1 1>/dev/null`;
1728 To exchange a command's STDOUT and STDERR in order to capture the STDERR
1729 but leave its STDOUT to come out the old STDERR:
1731 $output = `cmd 3>&1 1>&2 2>&3 3>&-`;
1733 To read both a command's STDOUT and its STDERR separately, it's easiest
1734 to redirect them separately to files, and then read from those files
1735 when the program is done:
1737 system("program args 1>program.stdout 2>program.stderr");
1739 The STDIN filehandle used by the command is inherited from Perl's STDIN.
1742 open BLAM, "blam" || die "Can't open: $!";
1743 open STDIN, "<&BLAM";
1746 will print the sorted contents of the file "blam".
1748 Using single-quote as a delimiter protects the command from Perl's
1749 double-quote interpolation, passing it on to the shell instead:
1751 $perl_info = qx(ps $$); # that's Perl's $$
1752 $shell_info = qx'ps $$'; # that's the new shell's $$
1754 How that string gets evaluated is entirely subject to the command
1755 interpreter on your system. On most platforms, you will have to protect
1756 shell metacharacters if you want them treated literally. This is in
1757 practice difficult to do, as it's unclear how to escape which characters.
1758 See L<perlsec> for a clean and safe example of a manual fork() and exec()
1759 to emulate backticks safely.
1761 On some platforms (notably DOS-like ones), the shell may not be
1762 capable of dealing with multiline commands, so putting newlines in
1763 the string may not get you what you want. You may be able to evaluate
1764 multiple commands in a single line by separating them with the command
1765 separator character, if your shell supports that (e.g. C<;> on many Unix
1766 shells; C<&> on the Windows NT C<cmd> shell).
1768 Beginning with v5.6.0, Perl will attempt to flush all files opened for
1769 output before starting the child process, but this may not be supported
1770 on some platforms (see L<perlport>). To be safe, you may need to set
1771 C<$|> ($AUTOFLUSH in English) or call the C<autoflush()> method of
1772 C<IO::Handle> on any open handles.
1774 Beware that some command shells may place restrictions on the length
1775 of the command line. You must ensure your strings don't exceed this
1776 limit after any necessary interpolations. See the platform-specific
1777 release notes for more details about your particular environment.
1779 Using this operator can lead to programs that are difficult to port,
1780 because the shell commands called vary between systems, and may in
1781 fact not be present at all. As one example, the C<type> command under
1782 the POSIX shell is very different from the C<type> command under DOS.
1783 That doesn't mean you should go out of your way to avoid backticks
1784 when they're the right way to get something done. Perl was made to be
1785 a glue language, and one of the things it glues together is commands.
1786 Just understand what you're getting yourself into.
1788 See L</"I/O Operators"> for more discussion.
1791 X<qw> X<quote, list> X<quote, words>
1793 Evaluates to a list of the words extracted out of STRING, using embedded
1794 whitespace as the word delimiters. It can be understood as being roughly
1797 split(' ', q/STRING/);
1799 the differences being that it generates a real list at compile time, and
1800 in scalar context it returns the last element in the list. So
1805 is semantically equivalent to the list:
1809 Some frequently seen examples:
1811 use POSIX qw( setlocale localeconv )
1812 @EXPORT = qw( foo bar baz );
1814 A common mistake is to try to separate the words with comma or to
1815 put comments into a multi-line C<qw>-string. For this reason, the
1816 C<use warnings> pragma and the B<-w> switch (that is, the C<$^W> variable)
1817 produces warnings if the STRING contains the "," or the "#" character.
1820 =item tr/SEARCHLIST/REPLACEMENTLIST/cds
1821 X<tr> X<y> X<transliterate> X</c> X</d> X</s>
1823 =item y/SEARCHLIST/REPLACEMENTLIST/cds
1825 Transliterates all occurrences of the characters found in the search list
1826 with the corresponding character in the replacement list. It returns
1827 the number of characters replaced or deleted. If no string is
1828 specified via the =~ or !~ operator, the $_ string is transliterated. (The
1829 string specified with =~ must be a scalar variable, an array element, a
1830 hash element, or an assignment to one of those, i.e., an lvalue.)
1832 A character range may be specified with a hyphen, so C<tr/A-J/0-9/>
1833 does the same replacement as C<tr/ACEGIBDFHJ/0246813579/>.
1834 For B<sed> devotees, C<y> is provided as a synonym for C<tr>. If the
1835 SEARCHLIST is delimited by bracketing quotes, the REPLACEMENTLIST has
1836 its own pair of quotes, which may or may not be bracketing quotes,
1837 e.g., C<tr[A-Z][a-z]> or C<tr(+\-*/)/ABCD/>.
1839 Note that C<tr> does B<not> do regular expression character classes
1840 such as C<\d> or C<[:lower:]>. The C<tr> operator is not equivalent to
1841 the tr(1) utility. If you want to map strings between lower/upper
1842 cases, see L<perlfunc/lc> and L<perlfunc/uc>, and in general consider
1843 using the C<s> operator if you need regular expressions.
1845 Note also that the whole range idea is rather unportable between
1846 character sets--and even within character sets they may cause results
1847 you probably didn't expect. A sound principle is to use only ranges
1848 that begin from and end at either alphabets of equal case (a-e, A-E),
1849 or digits (0-4). Anything else is unsafe. If in doubt, spell out the
1850 character sets in full.
1854 c Complement the SEARCHLIST.
1855 d Delete found but unreplaced characters.
1856 s Squash duplicate replaced characters.
1858 If the C</c> modifier is specified, the SEARCHLIST character set
1859 is complemented. If the C</d> modifier is specified, any characters
1860 specified by SEARCHLIST not found in REPLACEMENTLIST are deleted.
1861 (Note that this is slightly more flexible than the behavior of some
1862 B<tr> programs, which delete anything they find in the SEARCHLIST,
1863 period.) If the C</s> modifier is specified, sequences of characters
1864 that were transliterated to the same character are squashed down
1865 to a single instance of the character.
1867 If the C</d> modifier is used, the REPLACEMENTLIST is always interpreted
1868 exactly as specified. Otherwise, if the REPLACEMENTLIST is shorter
1869 than the SEARCHLIST, the final character is replicated till it is long
1870 enough. If the REPLACEMENTLIST is empty, the SEARCHLIST is replicated.
1871 This latter is useful for counting characters in a class or for
1872 squashing character sequences in a class.
1876 $ARGV[1] =~ tr/A-Z/a-z/; # canonicalize to lower case
1878 $cnt = tr/*/*/; # count the stars in $_
1880 $cnt = $sky =~ tr/*/*/; # count the stars in $sky
1882 $cnt = tr/0-9//; # count the digits in $_
1884 tr/a-zA-Z//s; # bookkeeper -> bokeper
1886 ($HOST = $host) =~ tr/a-z/A-Z/;
1888 tr/a-zA-Z/ /cs; # change non-alphas to single space
1891 [\000-\177]; # delete 8th bit
1893 If multiple transliterations are given for a character, only the
1898 will transliterate any A to X.
1900 Because the transliteration table is built at compile time, neither
1901 the SEARCHLIST nor the REPLACEMENTLIST are subjected to double quote
1902 interpolation. That means that if you want to use variables, you
1905 eval "tr/$oldlist/$newlist/";
1908 eval "tr/$oldlist/$newlist/, 1" or die $@;
1911 X<here-doc> X<heredoc> X<here-document> X<<< << >>>
1913 A line-oriented form of quoting is based on the shell "here-document"
1914 syntax. Following a C<< << >> you specify a string to terminate
1915 the quoted material, and all lines following the current line down to
1916 the terminating string are the value of the item.
1918 The terminating string may be either an identifier (a word), or some
1919 quoted text. An unquoted identifier works like double quotes.
1920 There may not be a space between the C<< << >> and the identifier,
1921 unless the identifier is explicitly quoted. (If you put a space it
1922 will be treated as a null identifier, which is valid, and matches the
1923 first empty line.) The terminating string must appear by itself
1924 (unquoted and with no surrounding whitespace) on the terminating line.
1926 If the terminating string is quoted, the type of quotes used determine
1927 the treatment of the text.
1933 Double quotes indicate that the text will be interpolated using exactly
1934 the same rules as normal double quoted strings.
1937 The price is $Price.
1940 print << "EOF"; # same as above
1941 The price is $Price.
1947 Single quotes indicate the text is to be treated literally with no
1948 interpolation of its content. This is similar to single quoted
1949 strings except that backslashes have no special meaning, with C<\\>
1950 being treated as two backslashes and not one as they would in every
1951 other quoting construct.
1953 This is the only form of quoting in perl where there is no need
1954 to worry about escaping content, something that code generators
1955 can and do make good use of.
1959 The content of the here doc is treated just as it would be if the
1960 string were embedded in backticks. Thus the content is interpolated
1961 as though it were double quoted and then executed via the shell, with
1962 the results of the execution returned.
1964 print << `EOC`; # execute command and get results
1970 It is possible to stack multiple here-docs in a row:
1972 print <<"foo", <<"bar"; # you can stack them
1978 myfunc(<< "THIS", 23, <<'THAT');
1985 Just don't forget that you have to put a semicolon on the end
1986 to finish the statement, as Perl doesn't know you're not going to
1994 If you want to remove the line terminator from your here-docs,
1997 chomp($string = <<'END');
2001 If you want your here-docs to be indented with the rest of the code,
2002 you'll need to remove leading whitespace from each line manually:
2004 ($quote = <<'FINIS') =~ s/^\s+//gm;
2005 The Road goes ever on and on,
2006 down from the door where it began.
2009 If you use a here-doc within a delimited construct, such as in C<s///eg>,
2010 the quoted material must come on the lines following the final delimiter.
2025 If the terminating identifier is on the last line of the program, you
2026 must be sure there is a newline after it; otherwise, Perl will give the
2027 warning B<Can't find string terminator "END" anywhere before EOF...>.
2029 Additionally, the quoting rules for the end of string identifier are not
2030 related to Perl's quoting rules. C<q()>, C<qq()>, and the like are not
2031 supported in place of C<''> and C<"">, and the only interpolation is for
2032 backslashing the quoting character:
2034 print << "abc\"def";
2038 Finally, quoted strings cannot span multiple lines. The general rule is
2039 that the identifier must be a string literal. Stick with that, and you
2044 =head2 Gory details of parsing quoted constructs
2045 X<quote, gory details>
2047 When presented with something that might have several different
2048 interpretations, Perl uses the B<DWIM> (that's "Do What I Mean")
2049 principle to pick the most probable interpretation. This strategy
2050 is so successful that Perl programmers often do not suspect the
2051 ambivalence of what they write. But from time to time, Perl's
2052 notions differ substantially from what the author honestly meant.
2054 This section hopes to clarify how Perl handles quoted constructs.
2055 Although the most common reason to learn this is to unravel labyrinthine
2056 regular expressions, because the initial steps of parsing are the
2057 same for all quoting operators, they are all discussed together.
2059 The most important Perl parsing rule is the first one discussed
2060 below: when processing a quoted construct, Perl first finds the end
2061 of that construct, then interprets its contents. If you understand
2062 this rule, you may skip the rest of this section on the first
2063 reading. The other rules are likely to contradict the user's
2064 expectations much less frequently than this first one.
2066 Some passes discussed below are performed concurrently, but because
2067 their results are the same, we consider them individually. For different
2068 quoting constructs, Perl performs different numbers of passes, from
2069 one to four, but these passes are always performed in the same order.
2073 =item Finding the end
2075 The first pass is finding the end of the quoted construct, where
2076 the information about the delimiters is used in parsing.
2077 During this search, text between the starting and ending delimiters
2078 is copied to a safe location. The text copied gets delimiter-independent.
2080 If the construct is a here-doc, the ending delimiter is a line
2081 that has a terminating string as the content. Therefore C<<<EOF> is
2082 terminated by C<EOF> immediately followed by C<"\n"> and starting
2083 from the first column of the terminating line.
2084 When searching for the terminating line of a here-doc, nothing
2085 is skipped. In other words, lines after the here-doc syntax
2086 are compared with the terminating string line by line.
2088 For the constructs except here-docs, single characters are used as starting
2089 and ending delimiters. If the starting delimiter is an opening punctuation
2090 (that is C<(>, C<[>, C<{>, or C<< < >>), the ending delimiter is the
2091 corresponding closing punctuation (that is C<)>, C<]>, C<}>, or C<< > >>).
2092 If the starting delimiter is an unpaired character like C</> or a closing
2093 punctuation, the ending delimiter is same as the starting delimiter.
2094 Therefore a C</> terminates a C<qq//> construct, while a C<]> terminates
2095 C<qq[]> and C<qq]]> constructs.
2097 When searching for single-character delimiters, escaped delimiters
2098 and C<\\> are skipped. For example, while searching for terminating C</>,
2099 combinations of C<\\> and C<\/> are skipped. If the delimiters are
2100 bracketing, nested pairs are also skipped. For example, while searching
2101 for closing C<]> paired with the opening C<[>, combinations of C<\\>, C<\]>,
2102 and C<\[> are all skipped, and nested C<[> and C<]> are skipped as well.
2103 However, when backslashes are used as the delimiters (like C<qq\\> and
2104 C<tr\\\>), nothing is skipped.
2105 During the search for the end, backslashes that escape delimiters
2106 are removed (exactly speaking, they are not copied to the safe location).
2108 For constructs with three-part delimiters (C<s///>, C<y///>, and
2109 C<tr///>), the search is repeated once more.
2110 If the first delimiter is not an opening punctuation, three delimiters must
2111 be same such as C<s!!!> and C<tr)))>, in which case the second delimiter
2112 terminates the left part and starts the right part at once.
2113 If the left part is delimited by bracketing punctuations (that is C<()>,
2114 C<[]>, C<{}>, or C<< <> >>), the right part needs another pair of
2115 delimiters such as C<s(){}> and C<tr[]//>. In these cases, whitespaces
2116 and comments are allowed between both parts, though the comment must follow
2117 at least one whitespace; otherwise a character expected as the start of
2118 the comment may be regarded as the starting delimiter of the right part.
2120 During this search no attention is paid to the semantics of the construct.
2123 "$hash{"$foo/$bar"}"
2128 bar # NOT a comment, this slash / terminated m//!
2131 do not form legal quoted expressions. The quoted part ends on the
2132 first C<"> and C</>, and the rest happens to be a syntax error.
2133 Because the slash that terminated C<m//> was followed by a C<SPACE>,
2134 the example above is not C<m//x>, but rather C<m//> with no C</x>
2135 modifier. So the embedded C<#> is interpreted as a literal C<#>.
2137 Also no attention is paid to C<\c\> (multichar control char syntax) during
2138 this search. Thus the second C<\> in C<qq/\c\/> is interpreted as a part
2139 of C<\/>, and the following C</> is not recognized as a delimiter.
2140 Instead, use C<\034> or C<\x1c> at the end of quoted constructs.
2145 The next step is interpolation in the text obtained, which is now
2146 delimiter-independent. There are multiple cases.
2152 No interpolation is performed.
2153 Note that the combination C<\\> is left intact, since escaped delimiters
2154 are not available for here-docs.
2156 =item C<m''>, the pattern of C<s'''>
2158 No interpolation is performed at this stage.
2159 Any backslashed sequences including C<\\> are treated at the stage
2160 to L</"parsing regular expressions">.
2162 =item C<''>, C<q//>, C<tr'''>, C<y'''>, the replacement of C<s'''>
2164 The only interpolation is removal of C<\> from pairs of C<\\>.
2165 Therefore C<-> in C<tr'''> and C<y'''> is treated literally
2166 as a hyphen and no character range is available.
2167 C<\1> in the replacement of C<s'''> does not work as C<$1>.
2169 =item C<tr///>, C<y///>
2171 No variable interpolation occurs. String modifying combinations for
2172 case and quoting such as C<\Q>, C<\U>, and C<\E> are not recognized.
2173 The other escape sequences such as C<\200> and C<\t> and backslashed
2174 characters such as C<\\> and C<\-> are converted to appropriate literals.
2175 The character C<-> is treated specially and therefore C<\-> is treated
2178 =item C<"">, C<``>, C<qq//>, C<qx//>, C<< <file*glob> >>, C<<<"EOF">
2180 C<\Q>, C<\U>, C<\u>, C<\L>, C<\l> (possibly paired with C<\E>) are
2181 converted to corresponding Perl constructs. Thus, C<"$foo\Qbaz$bar">
2182 is converted to C<$foo . (quotemeta("baz" . $bar))> internally.
2183 The other escape sequences such as C<\200> and C<\t> and backslashed
2184 characters such as C<\\> and C<\-> are replaced with appropriate
2187 Let it be stressed that I<whatever falls between C<\Q> and C<\E>>
2188 is interpolated in the usual way. Something like C<"\Q\\E"> has
2189 no C<\E> inside. instead, it has C<\Q>, C<\\>, and C<E>, so the
2190 result is the same as for C<"\\\\E">. As a general rule, backslashes
2191 between C<\Q> and C<\E> may lead to counterintuitive results. So,
2192 C<"\Q\t\E"> is converted to C<quotemeta("\t")>, which is the same
2193 as C<"\\\t"> (since TAB is not alphanumeric). Note also that:
2198 may be closer to the conjectural I<intention> of the writer of C<"\Q\t\E">.
2200 Interpolated scalars and arrays are converted internally to the C<join> and
2201 C<.> catenation operations. Thus, C<"$foo XXX '@arr'"> becomes:
2203 $foo . " XXX '" . (join $", @arr) . "'";
2205 All operations above are performed simultaneously, left to right.
2207 Because the result of C<"\Q STRING \E"> has all metacharacters
2208 quoted, there is no way to insert a literal C<$> or C<@> inside a
2209 C<\Q\E> pair. If protected by C<\>, C<$> will be quoted to became
2210 C<"\\\$">; if not, it is interpreted as the start of an interpolated
2213 Note also that the interpolation code needs to make a decision on
2214 where the interpolated scalar ends. For instance, whether
2215 C<< "a $b -> {c}" >> really means:
2217 "a " . $b . " -> {c}";
2223 Most of the time, the longest possible text that does not include
2224 spaces between components and which contains matching braces or
2225 brackets. because the outcome may be determined by voting based
2226 on heuristic estimators, the result is not strictly predictable.
2227 Fortunately, it's usually correct for ambiguous cases.
2229 =item the replacement of C<s///>
2231 Processing of C<\Q>, C<\U>, C<\u>, C<\L>, C<\l>, and interpolation
2232 happens as with C<qq//> constructs.
2234 It is at this step that C<\1> is begrudgingly converted to C<$1> in
2235 the replacement text of C<s///>, in order to correct the incorrigible
2236 I<sed> hackers who haven't picked up the saner idiom yet. A warning
2237 is emitted if the C<use warnings> pragma or the B<-w> command-line flag
2238 (that is, the C<$^W> variable) was set.
2240 =item C<RE> in C<?RE?>, C</RE/>, C<m/RE/>, C<s/RE/foo/>,
2242 Processing of C<\Q>, C<\U>, C<\u>, C<\L>, C<\l>, C<\E>,
2243 and interpolation happens (almost) as with C<qq//> constructs.
2245 Processing of C<\N{...}> is also done here, and compiled into an intermediate
2246 form for the regex compiler. (This is because, as mentioned below, the regex
2247 compilation may be done at execution time, and C<\N{...}> is a compile-time
2250 However any other combinations of C<\> followed by a character
2251 are not substituted but only skipped, in order to parse them
2252 as regular expressions at the following step.
2253 As C<\c> is skipped at this step, C<@> of C<\c@> in RE is possibly
2254 treated as an array symbol (for example C<@foo>),
2255 even though the same text in C<qq//> gives interpolation of C<\c@>.
2257 Moreover, inside C<(?{BLOCK})>, C<(?# comment )>, and
2258 a C<#>-comment in a C<//x>-regular expression, no processing is
2259 performed whatsoever. This is the first step at which the presence
2260 of the C<//x> modifier is relevant.
2262 Interpolation in patterns has several quirks: C<$|>, C<$(>, C<$)>, C<@+>
2263 and C<@-> are not interpolated, and constructs C<$var[SOMETHING]> are
2264 voted (by several different estimators) to be either an array element
2265 or C<$var> followed by an RE alternative. This is where the notation
2266 C<${arr[$bar]}> comes handy: C</${arr[0-9]}/> is interpreted as
2267 array element C<-9>, not as a regular expression from the variable
2268 C<$arr> followed by a digit, which would be the interpretation of
2269 C</$arr[0-9]/>. Since voting among different estimators may occur,
2270 the result is not predictable.
2272 The lack of processing of C<\\> creates specific restrictions on
2273 the post-processed text. If the delimiter is C</>, one cannot get
2274 the combination C<\/> into the result of this step. C</> will
2275 finish the regular expression, C<\/> will be stripped to C</> on
2276 the previous step, and C<\\/> will be left as is. Because C</> is
2277 equivalent to C<\/> inside a regular expression, this does not
2278 matter unless the delimiter happens to be character special to the
2279 RE engine, such as in C<s*foo*bar*>, C<m[foo]>, or C<?foo?>; or an
2280 alphanumeric char, as in:
2284 In the RE above, which is intentionally obfuscated for illustration, the
2285 delimiter is C<m>, the modifier is C<mx>, and after delimiter-removal the
2286 RE is the same as for C<m/ ^ a \s* b /mx>. There's more than one
2287 reason you're encouraged to restrict your delimiters to non-alphanumeric,
2288 non-whitespace choices.
2292 This step is the last one for all constructs except regular expressions,
2293 which are processed further.
2295 =item parsing regular expressions
2298 Previous steps were performed during the compilation of Perl code,
2299 but this one happens at run time, although it may be optimized to
2300 be calculated at compile time if appropriate. After preprocessing
2301 described above, and possibly after evaluation if concatenation,
2302 joining, casing translation, or metaquoting are involved, the
2303 resulting I<string> is passed to the RE engine for compilation.
2305 Whatever happens in the RE engine might be better discussed in L<perlre>,
2306 but for the sake of continuity, we shall do so here.
2308 This is another step where the presence of the C<//x> modifier is
2309 relevant. The RE engine scans the string from left to right and
2310 converts it to a finite automaton.
2312 Backslashed characters are either replaced with corresponding
2313 literal strings (as with C<\{>), or else they generate special nodes
2314 in the finite automaton (as with C<\b>). Characters special to the
2315 RE engine (such as C<|>) generate corresponding nodes or groups of
2316 nodes. C<(?#...)> comments are ignored. All the rest is either
2317 converted to literal strings to match, or else is ignored (as is
2318 whitespace and C<#>-style comments if C<//x> is present).
2320 Parsing of the bracketed character class construct, C<[...]>, is
2321 rather different than the rule used for the rest of the pattern.
2322 The terminator of this construct is found using the same rules as
2323 for finding the terminator of a C<{}>-delimited construct, the only
2324 exception being that C<]> immediately following C<[> is treated as
2325 though preceded by a backslash. Similarly, the terminator of
2326 C<(?{...})> is found using the same rules as for finding the
2327 terminator of a C<{}>-delimited construct.
2329 It is possible to inspect both the string given to RE engine and the
2330 resulting finite automaton. See the arguments C<debug>/C<debugcolor>
2331 in the C<use L<re>> pragma, as well as Perl's B<-Dr> command-line
2332 switch documented in L<perlrun/"Command Switches">.
2334 =item Optimization of regular expressions
2335 X<regexp, optimization>
2337 This step is listed for completeness only. Since it does not change
2338 semantics, details of this step are not documented and are subject
2339 to change without notice. This step is performed over the finite
2340 automaton that was generated during the previous pass.
2342 It is at this stage that C<split()> silently optimizes C</^/> to
2347 =head2 I/O Operators
2348 X<operator, i/o> X<operator, io> X<io> X<while> X<filehandle>
2351 There are several I/O operators you should know about.
2353 A string enclosed by backticks (grave accents) first undergoes
2354 double-quote interpolation. It is then interpreted as an external
2355 command, and the output of that command is the value of the
2356 backtick string, like in a shell. In scalar context, a single string
2357 consisting of all output is returned. In list context, a list of
2358 values is returned, one per line of output. (You can set C<$/> to use
2359 a different line terminator.) The command is executed each time the
2360 pseudo-literal is evaluated. The status value of the command is
2361 returned in C<$?> (see L<perlvar> for the interpretation of C<$?>).
2362 Unlike in B<csh>, no translation is done on the return data--newlines
2363 remain newlines. Unlike in any of the shells, single quotes do not
2364 hide variable names in the command from interpretation. To pass a
2365 literal dollar-sign through to the shell you need to hide it with a
2366 backslash. The generalized form of backticks is C<qx//>. (Because
2367 backticks always undergo shell expansion as well, see L<perlsec> for
2369 X<qx> X<`> X<``> X<backtick> X<glob>
2371 In scalar context, evaluating a filehandle in angle brackets yields
2372 the next line from that file (the newline, if any, included), or
2373 C<undef> at end-of-file or on error. When C<$/> is set to C<undef>
2374 (sometimes known as file-slurp mode) and the file is empty, it
2375 returns C<''> the first time, followed by C<undef> subsequently.
2377 Ordinarily you must assign the returned value to a variable, but
2378 there is one situation where an automatic assignment happens. If
2379 and only if the input symbol is the only thing inside the conditional
2380 of a C<while> statement (even if disguised as a C<for(;;)> loop),
2381 the value is automatically assigned to the global variable $_,
2382 destroying whatever was there previously. (This may seem like an
2383 odd thing to you, but you'll use the construct in almost every Perl
2384 script you write.) The $_ variable is not implicitly localized.
2385 You'll have to put a C<local $_;> before the loop if you want that
2388 The following lines are equivalent:
2390 while (defined($_ = <STDIN>)) { print; }
2391 while ($_ = <STDIN>) { print; }
2392 while (<STDIN>) { print; }
2393 for (;<STDIN>;) { print; }
2394 print while defined($_ = <STDIN>);
2395 print while ($_ = <STDIN>);
2396 print while <STDIN>;
2398 This also behaves similarly, but avoids $_ :
2400 while (my $line = <STDIN>) { print $line }
2402 In these loop constructs, the assigned value (whether assignment
2403 is automatic or explicit) is then tested to see whether it is
2404 defined. The defined test avoids problems where line has a string
2405 value that would be treated as false by Perl, for example a "" or
2406 a "0" with no trailing newline. If you really mean for such values
2407 to terminate the loop, they should be tested for explicitly:
2409 while (($_ = <STDIN>) ne '0') { ... }
2410 while (<STDIN>) { last unless $_; ... }
2412 In other boolean contexts, C<< <filehandle> >> without an
2413 explicit C<defined> test or comparison elicits a warning if the
2414 C<use warnings> pragma or the B<-w>
2415 command-line switch (the C<$^W> variable) is in effect.
2417 The filehandles STDIN, STDOUT, and STDERR are predefined. (The
2418 filehandles C<stdin>, C<stdout>, and C<stderr> will also work except
2419 in packages, where they would be interpreted as local identifiers
2420 rather than global.) Additional filehandles may be created with
2421 the open() function, amongst others. See L<perlopentut> and
2422 L<perlfunc/open> for details on this.
2423 X<stdin> X<stdout> X<sterr>
2425 If a <FILEHANDLE> is used in a context that is looking for
2426 a list, a list comprising all input lines is returned, one line per
2427 list element. It's easy to grow to a rather large data space this
2428 way, so use with care.
2430 <FILEHANDLE> may also be spelled C<readline(*FILEHANDLE)>.
2431 See L<perlfunc/readline>.
2433 The null filehandle <> is special: it can be used to emulate the
2434 behavior of B<sed> and B<awk>. Input from <> comes either from
2435 standard input, or from each file listed on the command line. Here's
2436 how it works: the first time <> is evaluated, the @ARGV array is
2437 checked, and if it is empty, C<$ARGV[0]> is set to "-", which when opened
2438 gives you standard input. The @ARGV array is then processed as a list
2439 of filenames. The loop
2442 ... # code for each line
2445 is equivalent to the following Perl-like pseudo code:
2447 unshift(@ARGV, '-') unless @ARGV;
2448 while ($ARGV = shift) {
2451 ... # code for each line
2455 except that it isn't so cumbersome to say, and will actually work.
2456 It really does shift the @ARGV array and put the current filename
2457 into the $ARGV variable. It also uses filehandle I<ARGV>
2458 internally. <> is just a synonym for <ARGV>, which
2459 is magical. (The pseudo code above doesn't work because it treats
2460 <ARGV> as non-magical.)
2462 Since the null filehandle uses the two argument form of L<perlfunc/open>
2463 it interprets special characters, so if you have a script like this:
2469 and call it with C<perl dangerous.pl 'rm -rfv *|'>, it actually opens a
2470 pipe, executes the C<rm> command and reads C<rm>'s output from that pipe.
2471 If you want all items in C<@ARGV> to be interpreted as file names, you
2472 can use the module C<ARGV::readonly> from CPAN.
2474 You can modify @ARGV before the first <> as long as the array ends up
2475 containing the list of filenames you really want. Line numbers (C<$.>)
2476 continue as though the input were one big happy file. See the example
2477 in L<perlfunc/eof> for how to reset line numbers on each file.
2479 If you want to set @ARGV to your own list of files, go right ahead.
2480 This sets @ARGV to all plain text files if no @ARGV was given:
2482 @ARGV = grep { -f && -T } glob('*') unless @ARGV;
2484 You can even set them to pipe commands. For example, this automatically
2485 filters compressed arguments through B<gzip>:
2487 @ARGV = map { /\.(gz|Z)$/ ? "gzip -dc < $_ |" : $_ } @ARGV;
2489 If you want to pass switches into your script, you can use one of the
2490 Getopts modules or put a loop on the front like this:
2492 while ($_ = $ARGV[0], /^-/) {
2495 if (/^-D(.*)/) { $debug = $1 }
2496 if (/^-v/) { $verbose++ }
2497 # ... # other switches
2501 # ... # code for each line
2504 The <> symbol will return C<undef> for end-of-file only once.
2505 If you call it again after this, it will assume you are processing another
2506 @ARGV list, and if you haven't set @ARGV, will read input from STDIN.
2508 If what the angle brackets contain is a simple scalar variable (e.g.,
2509 <$foo>), then that variable contains the name of the
2510 filehandle to input from, or its typeglob, or a reference to the
2516 If what's within the angle brackets is neither a filehandle nor a simple
2517 scalar variable containing a filehandle name, typeglob, or typeglob
2518 reference, it is interpreted as a filename pattern to be globbed, and
2519 either a list of filenames or the next filename in the list is returned,
2520 depending on context. This distinction is determined on syntactic
2521 grounds alone. That means C<< <$x> >> is always a readline() from
2522 an indirect handle, but C<< <$hash{key}> >> is always a glob().
2523 That's because $x is a simple scalar variable, but C<$hash{key}> is
2524 not--it's a hash element. Even C<< <$x > >> (note the extra space)
2525 is treated as C<glob("$x ")>, not C<readline($x)>.
2527 One level of double-quote interpretation is done first, but you can't
2528 say C<< <$foo> >> because that's an indirect filehandle as explained
2529 in the previous paragraph. (In older versions of Perl, programmers
2530 would insert curly brackets to force interpretation as a filename glob:
2531 C<< <${foo}> >>. These days, it's considered cleaner to call the
2532 internal function directly as C<glob($foo)>, which is probably the right
2533 way to have done it in the first place.) For example:
2539 is roughly equivalent to:
2541 open(FOO, "echo *.c | tr -s ' \t\r\f' '\\012\\012\\012\\012'|");
2547 except that the globbing is actually done internally using the standard
2548 C<File::Glob> extension. Of course, the shortest way to do the above is:
2552 A (file)glob evaluates its (embedded) argument only when it is
2553 starting a new list. All values must be read before it will start
2554 over. In list context, this isn't important because you automatically
2555 get them all anyway. However, in scalar context the operator returns
2556 the next value each time it's called, or C<undef> when the list has
2557 run out. As with filehandle reads, an automatic C<defined> is
2558 generated when the glob occurs in the test part of a C<while>,
2559 because legal glob returns (e.g. a file called F<0>) would otherwise
2560 terminate the loop. Again, C<undef> is returned only once. So if
2561 you're expecting a single value from a glob, it is much better to
2564 ($file) = <blurch*>;
2570 because the latter will alternate between returning a filename and
2573 If you're trying to do variable interpolation, it's definitely better
2574 to use the glob() function, because the older notation can cause people
2575 to become confused with the indirect filehandle notation.
2577 @files = glob("$dir/*.[ch]");
2578 @files = glob($files[$i]);
2580 =head2 Constant Folding
2581 X<constant folding> X<folding>
2583 Like C, Perl does a certain amount of expression evaluation at
2584 compile time whenever it determines that all arguments to an
2585 operator are static and have no side effects. In particular, string
2586 concatenation happens at compile time between literals that don't do
2587 variable substitution. Backslash interpolation also happens at
2588 compile time. You can say
2590 'Now is the time for all' . "\n" .
2591 'good men to come to.'
2593 and this all reduces to one string internally. Likewise, if
2596 foreach $file (@filenames) {
2597 if (-s $file > 5 + 100 * 2**16) { }
2600 the compiler will precompute the number which that expression
2601 represents so that the interpreter won't have to.
2606 Perl doesn't officially have a no-op operator, but the bare constants
2607 C<0> and C<1> are special-cased to not produce a warning in a void
2608 context, so you can for example safely do
2612 =head2 Bitwise String Operators
2613 X<operator, bitwise, string>
2615 Bitstrings of any size may be manipulated by the bitwise operators
2618 If the operands to a binary bitwise op are strings of different
2619 sizes, B<|> and B<^> ops act as though the shorter operand had
2620 additional zero bits on the right, while the B<&> op acts as though
2621 the longer operand were truncated to the length of the shorter.
2622 The granularity for such extension or truncation is one or more
2625 # ASCII-based examples
2626 print "j p \n" ^ " a h"; # prints "JAPH\n"
2627 print "JA" | " ph\n"; # prints "japh\n"
2628 print "japh\nJunk" & '_____'; # prints "JAPH\n";
2629 print 'p N$' ^ " E<H\n"; # prints "Perl\n";
2631 If you are intending to manipulate bitstrings, be certain that
2632 you're supplying bitstrings: If an operand is a number, that will imply
2633 a B<numeric> bitwise operation. You may explicitly show which type of
2634 operation you intend by using C<""> or C<0+>, as in the examples below.
2636 $foo = 150 | 105; # yields 255 (0x96 | 0x69 is 0xFF)
2637 $foo = '150' | 105; # yields 255
2638 $foo = 150 | '105'; # yields 255
2639 $foo = '150' | '105'; # yields string '155' (under ASCII)
2641 $baz = 0+$foo & 0+$bar; # both ops explicitly numeric
2642 $biz = "$foo" ^ "$bar"; # both ops explicitly stringy
2644 See L<perlfunc/vec> for information on how to manipulate individual bits
2647 =head2 Integer Arithmetic
2650 By default, Perl assumes that it must do most of its arithmetic in
2651 floating point. But by saying
2655 you may tell the compiler to use integer operations
2656 (see L<integer> for a detailed explanation) from here to the end of
2657 the enclosing BLOCK. An inner BLOCK may countermand this by saying
2661 which lasts until the end of that BLOCK. Note that this doesn't
2662 mean everything is an integer, merely that Perl will use integer
2663 operations for arithmetic, comparison, and bitwise operators. For
2664 example, even under C<use integer>, if you take the C<sqrt(2)>, you'll
2665 still get C<1.4142135623731> or so.
2667 Used on numbers, the bitwise operators ("&", "|", "^", "~", "<<",
2668 and ">>") always produce integral results. (But see also
2669 L<Bitwise String Operators>.) However, C<use integer> still has meaning for
2670 them. By default, their results are interpreted as unsigned integers, but
2671 if C<use integer> is in effect, their results are interpreted
2672 as signed integers. For example, C<~0> usually evaluates to a large
2673 integral value. However, C<use integer; ~0> is C<-1> on two's-complement
2676 =head2 Floating-point Arithmetic
2677 X<floating-point> X<floating point> X<float> X<real>
2679 While C<use integer> provides integer-only arithmetic, there is no
2680 analogous mechanism to provide automatic rounding or truncation to a
2681 certain number of decimal places. For rounding to a certain number
2682 of digits, sprintf() or printf() is usually the easiest route.
2685 Floating-point numbers are only approximations to what a mathematician
2686 would call real numbers. There are infinitely more reals than floats,
2687 so some corners must be cut. For example:
2689 printf "%.20g\n", 123456789123456789;
2690 # produces 123456789123456784
2692 Testing for exact floating-point equality or inequality is not a
2693 good idea. Here's a (relatively expensive) work-around to compare
2694 whether two floating-point numbers are equal to a particular number of
2695 decimal places. See Knuth, volume II, for a more robust treatment of
2699 my ($X, $Y, $POINTS) = @_;
2701 $tX = sprintf("%.${POINTS}g", $X);
2702 $tY = sprintf("%.${POINTS}g", $Y);
2706 The POSIX module (part of the standard perl distribution) implements
2707 ceil(), floor(), and other mathematical and trigonometric functions.
2708 The Math::Complex module (part of the standard perl distribution)
2709 defines mathematical functions that work on both the reals and the
2710 imaginary numbers. Math::Complex not as efficient as POSIX, but
2711 POSIX can't work with complex numbers.
2713 Rounding in financial applications can have serious implications, and
2714 the rounding method used should be specified precisely. In these
2715 cases, it probably pays not to trust whichever system rounding is
2716 being used by Perl, but to instead implement the rounding function you
2719 =head2 Bigger Numbers
2720 X<number, arbitrary precision>
2722 The standard Math::BigInt and Math::BigFloat modules provide
2723 variable-precision arithmetic and overloaded operators, although
2724 they're currently pretty slow. At the cost of some space and
2725 considerable speed, they avoid the normal pitfalls associated with
2726 limited-precision representations.
2729 $x = Math::BigInt->new('123456789123456789');
2732 # prints +15241578780673678515622620750190521
2734 There are several modules that let you calculate with (bound only by
2735 memory and cpu-time) unlimited or fixed precision. There are also
2736 some non-standard modules that provide faster implementations via
2737 external C libraries.
2739 Here is a short, but incomplete summary:
2741 Math::Fraction big, unlimited fractions like 9973 / 12967
2742 Math::String treat string sequences like numbers
2743 Math::FixedPrecision calculate with a fixed precision
2744 Math::Currency for currency calculations
2745 Bit::Vector manipulate bit vectors fast (uses C)
2746 Math::BigIntFast Bit::Vector wrapper for big numbers
2747 Math::Pari provides access to the Pari C library
2748 Math::BigInteger uses an external C library
2749 Math::Cephes uses external Cephes C library (no big numbers)
2750 Math::Cephes::Fraction fractions via the Cephes library
2751 Math::GMP another one using an external C library