This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
CommitLineData
a0d0e21e 1=head1 NAME
d74e8afc 2X<operator>
a0d0e21e
LW
3
4perlop - Perl operators and precedence
5
d042e63d
MS
6=head1 DESCRIPTION
7
89d205f2 8=head2 Operator Precedence and Associativity
d74e8afc 9X<operator, precedence> X<precedence> X<associativity>
d042e63d
MS
10
11Operator precedence and associativity work in Perl more or less like
12they do in mathematics.
13
14I<Operator precedence> means some operators are evaluated before
15others. For example, in C<2 + 4 * 5>, the multiplication has higher
16precedence so C<4 * 5> is evaluated first yielding C<2 + 20 ==
1722> and not C<6 * 5 == 30>.
18
19I<Operator associativity> defines what happens if a sequence of the
20same operators is used one after another: whether the evaluator will
21evaluate the left operations first or the right. For example, in C<8
22- 4 - 2>, subtraction is left associative so Perl evaluates the
23expression left to right. C<8 - 4> is evaluated first making the
24expression C<4 - 2 == 2> and not C<8 - 2 == 6>.
a0d0e21e
LW
25
26Perl operators have the following associativity and precedence,
19799a22
GS
27listed from highest precedence to lowest. Operators borrowed from
28C keep the same precedence relationship with each other, even where
29C's precedence is slightly screwy. (This makes learning Perl easier
30for C folks.) With very few exceptions, these all operate on scalar
31values only, not array values.
a0d0e21e
LW
32
33 left terms and list operators (leftward)
34 left ->
35 nonassoc ++ --
36 right **
37 right ! ~ \ and unary + and -
54310121 38 left =~ !~
a0d0e21e
LW
39 left * / % x
40 left + - .
41 left << >>
42 nonassoc named unary operators
43 nonassoc < > <= >= lt gt le ge
0d863452 44 nonassoc == != <=> eq ne cmp ~~
a0d0e21e
LW
45 left &
46 left | ^
47 left &&
c963b151 48 left || //
137443ea 49 nonassoc .. ...
a0d0e21e
LW
50 right ?:
51 right = += -= *= etc.
52 left , =>
53 nonassoc list operators (rightward)
a5f75d66 54 right not
a0d0e21e 55 left and
f23102e2 56 left or xor
a0d0e21e
LW
57
58In the following sections, these operators are covered in precedence order.
59
5a964f20
TC
60Many operators can be overloaded for objects. See L<overload>.
61
a0d0e21e 62=head2 Terms and List Operators (Leftward)
d74e8afc 63X<list operator> X<operator, list> X<term>
a0d0e21e 64
62c18ce2 65A TERM has the highest precedence in Perl. They include variables,
5f05dabc 66quote and quote-like operators, any expression in parentheses,
a0d0e21e
LW
67and any function whose arguments are parenthesized. Actually, there
68aren't really functions in this sense, just list operators and unary
69operators behaving as functions because you put parentheses around
70the arguments. These are all documented in L<perlfunc>.
71
72If any list operator (print(), etc.) or any unary operator (chdir(), etc.)
73is followed by a left parenthesis as the next token, the operator and
74arguments within parentheses are taken to be of highest precedence,
75just like a normal function call.
76
77In the absence of parentheses, the precedence of list operators such as
78C<print>, C<sort>, or C<chmod> is either very high or very low depending on
54310121 79whether you are looking at the left side or the right side of the operator.
a0d0e21e
LW
80For example, in
81
82 @ary = (1, 3, sort 4, 2);
83 print @ary; # prints 1324
84
19799a22
GS
85the commas on the right of the sort are evaluated before the sort,
86but the commas on the left are evaluated after. In other words,
87list operators tend to gobble up all arguments that follow, and
a0d0e21e 88then act like a simple TERM with regard to the preceding expression.
19799a22 89Be careful with parentheses:
a0d0e21e
LW
90
91 # These evaluate exit before doing the print:
92 print(\$foo, exit); # Obviously not what you want.
93 print \$foo, exit; # Nor is this.
94
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.
99
100Also note that
101
102 print (\$foo & 255) + 1, "\n";
103
d042e63d
MS
104probably doesn't do what you expect at first glance. The parentheses
105enclose the argument list for C<print> which is evaluated (printing
106the result of C<\$foo & 255>). Then one is added to the return value
107of C<print> (usually 1). The result is something like this:
108
109 1 + 1, "\n"; # Obviously not what you meant.
110
111To do what you meant properly, you must write:
112
113 print((\$foo & 255) + 1, "\n");
114
115See L<Named Unary Operators> for more discussion of this.
a0d0e21e
LW
116
117Also parsed as terms are the C<do {}> and C<eval {}> constructs, as
54310121 118well as subroutine and method calls, and the anonymous
a0d0e21e
LW
119constructors C<[]> and C<{}>.
120
2ae324a7 121See also L<Quote and Quote-like Operators> toward the end of this section,
da87341d 122as well as L</"I/O Operators">.
a0d0e21e
LW
123
124=head2 The Arrow Operator
d74e8afc 125X<arrow> X<dereference> X<< -> >>
a0d0e21e 126
35f2feb0 127"C<< -> >>" is an infix dereference operator, just as it is in C
19799a22
GS
128and C++. If the right side is either a C<[...]>, C<{...}>, or a
129C<(...)> subscript, then the left side must be either a hard or
130symbolic reference to an array, a hash, or a subroutine respectively.
131(Or technically speaking, a location capable of holding a hard
132reference, if it's an array or hash reference being used for
133assignment.) See L<perlreftut> and L<perlref>.
a0d0e21e 134
19799a22
GS
135Otherwise, the right side is a method name or a simple scalar
136variable containing either the method name or a subroutine reference,
137and the left side must be either an object (a blessed reference)
138or a class name (that is, a package name). See L<perlobj>.
a0d0e21e 139
5f05dabc 140=head2 Auto-increment and Auto-decrement
d74e8afc 141X<increment> X<auto-increment> X<++> X<decrement> X<auto-decrement> X<-->
a0d0e21e 142
d042e63d
MS
143"++" and "--" work as in C. That is, if placed before a variable,
144they increment or decrement the variable by one before returning the
145value, and if placed after, increment or decrement after returning the
146value.
147
148 \$i = 0; \$j = 0;
149 print \$i++; # prints 0
150 print ++\$j; # prints 1
a0d0e21e 151
b033823e 152Note that just as in C, Perl doesn't define B<when> the variable is
89d205f2 153incremented or decremented. You just know it will be done sometime
b033823e 154before or after the value is returned. This also means that modifying
c543c01b 155a variable twice in the same statement will lead to undefined behavior.
b033823e
A
156Avoid statements like:
157
158 \$i = \$i ++;
159 print ++ \$i + \$i ++;
160
161Perl will not guarantee what the result of the above statements is.
162
54310121 163The auto-increment operator has a little extra builtin magic to it. If
a0d0e21e
LW
164you increment a variable that is numeric, or that has ever been used in
165a numeric context, you get a normal increment. If, however, the
5f05dabc 166variable has been used in only string contexts since it was set, and
5a964f20 167has a value that is not the empty string and matches the pattern
9c0670e1 168C</^[a-zA-Z]*[0-9]*\z/>, the increment is done as a string, preserving each
a0d0e21e
LW
169character within its range, with carry:
170
c543c01b
TC
171 print ++(\$foo = "99"); # prints "100"
172 print ++(\$foo = "a0"); # prints "a1"
173 print ++(\$foo = "Az"); # prints "Ba"
174 print ++(\$foo = "zz"); # prints "aaa"
a0d0e21e 175
6a61d433
HS
176C<undef> is always treated as numeric, and in particular is changed
177to C<0> before incrementing (so that a post-increment of an undef value
178will return C<0> rather than C<undef>).
179
5f05dabc 180The auto-decrement operator is not magical.
a0d0e21e
LW
181
182=head2 Exponentiation
d74e8afc 183X<**> X<exponentiation> X<power>
a0d0e21e 184
19799a22 185Binary "**" is the exponentiation operator. It binds even more
cb1a09d0
AD
186tightly than unary minus, so -2**4 is -(2**4), not (-2)**4. (This is
187implemented using C's pow(3) function, which actually works on doubles
188internally.)
a0d0e21e
LW
189
190=head2 Symbolic Unary Operators
d74e8afc 191X<unary operator> X<operator, unary>
a0d0e21e 192
1ca345ed 193Unary "!" performs logical negation, that is, "not". See also C<not> for a lower
a0d0e21e 194precedence version of this.
d74e8afc 195X<!>
a0d0e21e 196
da2f94c5
FC
197Unary "-" performs arithmetic negation if the operand is numeric,
198including any string that looks like a number. If the operand is
199an identifier, a string consisting of a minus sign concatenated
200with the identifier is returned. Otherwise, if the string starts
201with a plus or minus, a string starting with the opposite sign is
202returned. One effect of these rules is that -bareword is equivalent
8705167b 203to the string "-bareword". If, however, the string begins with a
353c6505 204non-alphabetic character (excluding "+" or "-"), Perl will attempt to convert
06705523
SP
205the string to a numeric and the arithmetic negation is performed. If the
206string cannot be cleanly converted to a numeric, Perl will give the warning
207B<Argument "the string" isn't numeric in negation (-) at ...>.
d74e8afc 208X<-> X<negation, arithmetic>
a0d0e21e 209
1ca345ed 210Unary "~" performs bitwise negation, that is, 1's complement. For
972b05a9
JH
211example, C<0666 & ~027> is 0640. (See also L<Integer Arithmetic> and
212L<Bitwise String Operators>.) Note that the width of the result is
213platform-dependent: ~0 is 32 bits wide on a 32-bit platform, but 64
214bits wide on a 64-bit platform, so if you are expecting a certain bit
f113cf86 215width, remember to use the "&" operator to mask off the excess bits.
d74e8afc 216X<~> X<negation, binary>
a0d0e21e 217
f113cf86
TC
218When complementing strings, if all characters have ordinal values under
219256, then their complements will, also. But if they do not, all
220characters will be in either 32- or 64-bit complements, depending on your
221architecture. So for example, C<~"\x{3B1}"> is C<"\x{FFFF_FC4E}"> on
22232-bit machines and C<"\x{FFFF_FFFF_FFFF_FC4E}"> on 64-bit machines.
223
a0d0e21e
LW
224Unary "+" has no effect whatsoever, even on strings. It is useful
225syntactically for separating a function name from a parenthesized expression
226that would otherwise be interpreted as the complete list of function
5ba421f6 227arguments. (See examples above under L<Terms and List Operators (Leftward)>.)
d74e8afc 228X<+>
a0d0e21e 229
19799a22
GS
230Unary "\" creates a reference to whatever follows it. See L<perlreftut>
231and L<perlref>. Do not confuse this behavior with the behavior of
232backslash within a string, although both forms do convey the notion
233of protecting the next thing from interpolation.
d74e8afc 234X<\> X<reference> X<backslash>
a0d0e21e
LW
235
236=head2 Binding Operators
d74e8afc 237X<binding> X<operator, binding> X<=~> X<!~>
a0d0e21e 238
c07a80fd 239Binary "=~" binds a scalar expression to a pattern match. Certain operations
cb1a09d0
AD
240search or modify the string \$_ by default. This operator makes that kind
241of operation work on some other string. The right argument is a search
2c268ad5
TP
242pattern, substitution, or transliteration. The left argument is what is
243supposed to be searched, substituted, or transliterated instead of the default
f8bab1e9 244\$_. When used in scalar context, the return value generally indicates the
8ff32507
FC
245success of the operation. The exceptions are substitution (s///)
246and transliteration (y///) with the C</r> (non-destructive) option,
247which cause the B<r>eturn value to be the result of the substitution.
248Behavior in list context depends on the particular operator.
000c65fc
DG
249See L</"Regexp Quote-Like Operators"> for details and L<perlretut> for
250examples using these operators.
f8bab1e9
GS
251
252If the right argument is an expression rather than a search pattern,
2c268ad5 253substitution, or transliteration, it is interpreted as a search pattern at run
89d205f2
YO
254time. Note that this means that its contents will be interpolated twice, so
255
1ca345ed 256 '\\' =~ q'\\';
89d205f2
YO
257
258is not ok, as the regex engine will end up trying to compile the
259pattern C<\>, which it will consider a syntax error.
a0d0e21e
LW
260
261Binary "!~" is just like "=~" except the return value is negated in
262the logical sense.
263
8ff32507
FC
264Binary "!~" with a non-destructive substitution (s///r) or transliteration
265(y///r) is a syntax error.
4f4d7508 266
a0d0e21e 267=head2 Multiplicative Operators
d74e8afc 268X<operator, multiplicative>
a0d0e21e
LW
269
270Binary "*" multiplies two numbers.
d74e8afc 271X<*>
a0d0e21e
LW
272
273Binary "/" divides two numbers.
d74e8afc 274X</> X<slash>
a0d0e21e 275
f7918450
KW
276Binary "%" is the modulo operator, which computes the division
277remainder of its first argument with respect to its second argument.
278Given integer
54310121 279operands C<\$a> and C<\$b>: If C<\$b> is positive, then C<\$a % \$b> is
f7918450 280C<\$a> minus the largest multiple of C<\$b> less than or equal to
54310121 281C<\$a>. If C<\$b> is negative, then C<\$a % \$b> is C<\$a> minus the
1ca345ed 282smallest multiple of C<\$b> that is not less than C<\$a> (that is, the
89b4f0ad 283result will be less than or equal to zero). If the operands
4848a83b
TS
284C<\$a> and C<\$b> are floating point values and the absolute value of
285C<\$b> (that is C<abs(\$b)>) is less than C<(UV_MAX + 1)>, only
286the integer portion of C<\$a> and C<\$b> will be used in the operation
287(Note: here C<UV_MAX> means the maximum of the unsigned integer type).
288If the absolute value of the right operand (C<abs(\$b)>) is greater than
289or equal to C<(UV_MAX + 1)>, "%" computes the floating-point remainder
290C<\$r> in the equation C<(\$r = \$a - \$i*\$b)> where C<\$i> is a certain
f7918450 291integer that makes C<\$r> have the same sign as the right operand
4848a83b
TS
292C<\$b> (B<not> as the left operand C<\$a> like C function C<fmod()>)
293and the absolute value less than that of C<\$b>.
0412d526 294Note that when C<use integer> is in scope, "%" gives you direct access
f7918450 295to the modulo operator as implemented by your C compiler. This
55d729e4
GS
296operator is not as well defined for negative operands, but it will
297execute faster.
f7918450 298X<%> X<remainder> X<modulo> X<mod>
55d729e4 299
62d10b70
GS
300Binary "x" is the repetition operator. In scalar context or if the left
301operand is not enclosed in parentheses, it returns a string consisting
302of the left operand repeated the number of times specified by the right
303operand. In list context, if the left operand is enclosed in
3585017f
YST
304parentheses or is a list formed by C<qw/STRING/>, it repeats the list.
305If the right operand is zero or negative, it returns an empty string
306or an empty list, depending on the context.
d74e8afc 307X<x>
a0d0e21e
LW
308
309 print '-' x 80; # print row of dashes
310
311 print "\t" x (\$tab/8), ' ' x (\$tab%8); # tab over
312
313 @ones = (1) x 80; # a list of 80 1's
314 @ones = (5) x @ones; # set all elements to 5
315
316
317=head2 Additive Operators
d74e8afc 318X<operator, additive>
a0d0e21e 319
1ca345ed 320Binary C<+> returns the sum of two numbers.
d74e8afc 321X<+>
a0d0e21e 322
1ca345ed 323Binary C<-> returns the difference of two numbers.
d74e8afc 324X<->
a0d0e21e 325
1ca345ed 326Binary C<.> concatenates two strings.
d74e8afc
ITB
327X<string, concatenation> X<concatenation>
328X<cat> X<concat> X<concatenate> X<.>
a0d0e21e
LW
329
330=head2 Shift Operators
d74e8afc
ITB
331X<shift operator> X<operator, shift> X<<< << >>>
332X<<< >> >>> X<right shift> X<left shift> X<bitwise shift>
333X<shl> X<shr> X<shift, right> X<shift, left>
a0d0e21e 334
1ca345ed 335Binary C<<< << >>> returns the value of its left argument shifted left by the
55497cff 336number of bits specified by the right argument. Arguments should be
982ce180 337integers. (See also L<Integer Arithmetic>.)
a0d0e21e 338
1ca345ed 339Binary C<<< >> >>> returns the value of its left argument shifted right by
55497cff 340the number of bits specified by the right argument. Arguments should
982ce180 341be integers. (See also L<Integer Arithmetic>.)
a0d0e21e 342
1ca345ed
TC
343Note that both C<<< << >>> and C<<< >> >>> in Perl are implemented directly using
344C<<< << >>> and C<<< >> >>> in C. If C<use integer> (see L<Integer Arithmetic>) is
b16cf6df
JH
345in force then signed C integers are used, else unsigned C integers are
346used. Either way, the implementation isn't going to generate results
347larger than the size of the integer type Perl was built with (32 bits
348or 64 bits).
349
350The result of overflowing the range of the integers is undefined
351because it is undefined also in C. In other words, using 32-bit
352integers, C<< 1 << 32 >> is undefined. Shifting by a negative number
353of bits is also undefined.
354
1ca345ed
TC
355If you get tired of being subject to your platform's native integers,
356the C<use bigint> pragma neatly sidesteps the issue altogether:
357
358 print 20 << 20; # 20971520
359 print 20 << 40; # 5120 on 32-bit machines,
360 # 21990232555520 on 64-bit machines
361 use bigint;
362 print 20 << 100; # 25353012004564588029934064107520
363
a0d0e21e 364=head2 Named Unary Operators
d74e8afc 365X<operator, named unary>
a0d0e21e
LW
366
367The various named unary operators are treated as functions with one
568e6d8b 368argument, with optional parentheses.
a0d0e21e
LW
369
370If any list operator (print(), etc.) or any unary operator (chdir(), etc.)
371is followed by a left parenthesis as the next token, the operator and
372arguments within parentheses are taken to be of highest precedence,
3981b0eb 373just like a normal function call. For example,
1ca345ed 374because named unary operators are higher precedence than C<||>:
a0d0e21e
LW
375
376 chdir \$foo || die; # (chdir \$foo) || die
377 chdir(\$foo) || die; # (chdir \$foo) || die
378 chdir (\$foo) || die; # (chdir \$foo) || die
379 chdir +(\$foo) || die; # (chdir \$foo) || die
380
3981b0eb 381but, because * is higher precedence than named operators:
a0d0e21e
LW
382
383 chdir \$foo * 20; # chdir (\$foo * 20)
384 chdir(\$foo) * 20; # (chdir \$foo) * 20
385 chdir (\$foo) * 20; # (chdir \$foo) * 20
386 chdir +(\$foo) * 20; # chdir (\$foo * 20)
387
388 rand 10 * 20; # rand (10 * 20)
389 rand(10) * 20; # (rand 10) * 20
390 rand (10) * 20; # (rand 10) * 20
391 rand +(10) * 20; # rand (10 * 20)
392
568e6d8b
RGS
393Regarding precedence, the filetest operators, like C<-f>, C<-M>, etc. are
394treated like named unary operators, but they don't follow this functional
395parenthesis rule. That means, for example, that C<-f(\$file).".bak"> is
396equivalent to C<-f "\$file.bak">.
d74e8afc 397X<-X> X<filetest> X<operator, filetest>
568e6d8b 398
5ba421f6 399See also L<"Terms and List Operators (Leftward)">.
a0d0e21e
LW
400
401=head2 Relational Operators
d74e8afc 402X<relational operator> X<operator, relational>
a0d0e21e 403
1ca345ed
TC
404Perl operators that return true or false generally return values
405that can be safely used as numbers. For example, the relational
406operators in this section and the equality operators in the next
407one return C<1> for true and a special version of the defined empty
408string, C<"">, which counts as a zero but is exempt from warnings
409about improper numeric conversions, just as C<"0 but true"> is.
410
35f2feb0 411Binary "<" returns true if the left argument is numerically less than
a0d0e21e 412the right argument.
d74e8afc 413X<< < >>
a0d0e21e 414
35f2feb0 415Binary ">" returns true if the left argument is numerically greater
a0d0e21e 416than the right argument.
d74e8afc 417X<< > >>
a0d0e21e 418
35f2feb0 419Binary "<=" returns true if the left argument is numerically less than
a0d0e21e 420or equal to the right argument.
d74e8afc 421X<< <= >>
a0d0e21e 422
35f2feb0 423Binary ">=" returns true if the left argument is numerically greater
a0d0e21e 424than or equal to the right argument.
d74e8afc 425X<< >= >>
a0d0e21e
LW
426
427Binary "lt" returns true if the left argument is stringwise less than
428the right argument.
d74e8afc 429X<< lt >>
a0d0e21e
LW
430
431Binary "gt" returns true if the left argument is stringwise greater
432than the right argument.
d74e8afc 433X<< gt >>
a0d0e21e
LW
434
435Binary "le" returns true if the left argument is stringwise less than
436or equal to the right argument.
d74e8afc 437X<< le >>
a0d0e21e
LW
438
439Binary "ge" returns true if the left argument is stringwise greater
440than or equal to the right argument.
d74e8afc 441X<< ge >>
a0d0e21e
LW
442
443=head2 Equality Operators
d74e8afc 444X<equality> X<equal> X<equals> X<operator, equality>
a0d0e21e
LW
445
446Binary "==" returns true if the left argument is numerically equal to
447the right argument.
d74e8afc 448X<==>
a0d0e21e
LW
449
450Binary "!=" returns true if the left argument is numerically not equal
451to the right argument.
d74e8afc 452X<!=>
a0d0e21e 453
35f2feb0 454Binary "<=>" returns -1, 0, or 1 depending on whether the left
6ee5d4e7 455argument is numerically less than, equal to, or greater than the right
d4ad863d 456argument. If your platform supports NaNs (not-a-numbers) as numeric
7d3a9d88
NC
457values, using them with "<=>" returns undef. NaN is not "<", "==", ">",
458"<=" or ">=" anything (even NaN), so those 5 return false. NaN != NaN
459returns true, as does NaN != anything else. If your platform doesn't
460support NaNs then NaN is just a string with numeric value 0.
d74e8afc 461X<< <=> >> X<spaceship>
7d3a9d88 462
1ca345ed
TC
463 \$ perl -le '\$a = "NaN"; print "No NaN support here" if \$a == \$a'
464 \$ perl -le '\$a = "NaN"; print "NaN support here" if \$a != \$a'
465
466(Note that the L<bigint>, L<bigrat>, and L<bignum> pragmas all
467support "NaN".)
a0d0e21e
LW
468
469Binary "eq" returns true if the left argument is stringwise equal to
470the right argument.
d74e8afc 471X<eq>
a0d0e21e
LW
472
473Binary "ne" returns true if the left argument is stringwise not equal
474to the right argument.
d74e8afc 475X<ne>
a0d0e21e 476
d4ad863d
JH
477Binary "cmp" returns -1, 0, or 1 depending on whether the left
478argument is stringwise less than, equal to, or greater than the right
479argument.
d74e8afc 480X<cmp>
a0d0e21e 481
1ca345ed
TC
482Binary "~~" does a smartmatch between its arguments. Smart matching
483is described in the next section.
0d863452
RH
484X<~~>
485
a034a98d 486"lt", "le", "ge", "gt" and "cmp" use the collation (sort) order specified
66cbab2c
KW
487by the current locale if a legacy C<use locale> (but not
488C<use locale ':not_characters'>) is in effect. See
1ca345ed
TC
489L<perllocale>. Do not mix these with Unicode, only with legacy binary
490encodings. The standard L<Unicode::Collate> and
491L<Unicode::Collate::Locale> modules offer much more powerful solutions to
492collation issues.
493
494=head2 Smartmatch Operator
495
496First available in Perl 5.10.1 (the 5.10.0 version behaved differently),
497binary C<~~> does a "smartmatch" between its arguments. This is mostly
498used implicitly in the C<when> construct described in L<perlsyn>, although
499not all C<when> clauses call the smartmatch operator. Unique among all of
500Perl's operators, the smartmatch operator can recurse.
501
502It is also unique in that all other Perl operators impose a context
503(usually string or numeric context) on their operands, autoconverting
504those operands to those imposed contexts. In contrast, smartmatch
505I<infers> contexts from the actual types of its operands and uses that
506type information to select a suitable comparison mechanism.
507
508The C<~~> operator compares its operands "polymorphically", determining how
509to compare them according to their actual types (numeric, string, array,
510hash, etc.) Like the equality operators with which it shares the same
511precedence, C<~~> returns 1 for true and C<""> for false. It is often best
512read aloud as "in", "inside of", or "is contained in", because the left
513operand is often looked for I<inside> the right operand. That makes the
40bec8a5 514order of the operands to the smartmatch operand often opposite that of
1ca345ed
TC
515the regular match operator. In other words, the "smaller" thing is usually
516placed in the left operand and the larger one in the right.
517
518The behavior of a smartmatch depends on what type of things its arguments
519are, as determined by the following table. The first row of the table
520whose types apply determines the smartmatch behavior. Because what
521actually happens is mostly determined by the type of the second operand,
522the table is sorted on the right operand instead of on the left.
523
524 Left Right Description and pseudocode
525 ===============================================================
526 Any undef check whether Any is undefined
527 like: !defined Any
528
529 Any Object invoke ~~ overloading on Object, or die
530
531 Right operand is an ARRAY:
532
533 Left Right Description and pseudocode
534 ===============================================================
535 ARRAY1 ARRAY2 recurse on paired elements of ARRAY1 and ARRAY2[2]
536 like: (ARRAY1[0] ~~ ARRAY2[0])
537 && (ARRAY1[1] ~~ ARRAY2[1]) && ...
538 HASH ARRAY any ARRAY elements exist as HASH keys
539 like: grep { exists HASH->{\$_} } ARRAY
540 Regexp ARRAY any ARRAY elements pattern match Regexp
541 like: grep { /Regexp/ } ARRAY
542 undef ARRAY undef in ARRAY
543 like: grep { !defined } ARRAY
40bec8a5 544 Any ARRAY smartmatch each ARRAY element[3]
1ca345ed
TC
545 like: grep { Any ~~ \$_ } ARRAY
546
547 Right operand is a HASH:
548
549 Left Right Description and pseudocode
550 ===============================================================
551 HASH1 HASH2 all same keys in both HASHes
552 like: keys HASH1 ==
553 grep { exists HASH2->{\$_} } keys HASH1
554 ARRAY HASH any ARRAY elements exist as HASH keys
555 like: grep { exists HASH->{\$_} } ARRAY
556 Regexp HASH any HASH keys pattern match Regexp
557 like: grep { /Regexp/ } keys HASH
558 undef HASH always false (undef can't be a key)
559 like: 0 == 1
560 Any HASH HASH key existence
561 like: exists HASH->{Any}
562
563 Right operand is CODE:
564
565 Left Right Description and pseudocode
566 ===============================================================
567 ARRAY CODE sub returns true on all ARRAY elements[1]
568 like: !grep { !CODE->(\$_) } ARRAY
569 HASH CODE sub returns true on all HASH keys[1]
570 like: !grep { !CODE->(\$_) } keys HASH
571 Any CODE sub passed Any returns true
572 like: CODE->(Any)
573
574Right operand is a Regexp:
575
576 Left Right Description and pseudocode
577 ===============================================================
578 ARRAY Regexp any ARRAY elements match Regexp
579 like: grep { /Regexp/ } ARRAY
580 HASH Regexp any HASH keys match Regexp
581 like: grep { /Regexp/ } keys HASH
582 Any Regexp pattern match
583 like: Any =~ /Regexp/
584
585 Other:
586
587 Left Right Description and pseudocode
588 ===============================================================
589 Object Any invoke ~~ overloading on Object,
590 or fall back to...
591
592 Any Num numeric equality
593 like: Any == Num
594 Num nummy[4] numeric equality
595 like: Num == nummy
596 undef Any check whether undefined
597 like: !defined(Any)
598 Any Any string equality
599 like: Any eq Any
600
601
602Notes:
603
604=over
605
606=item 1.
607Empty hashes or arrays match.
608
609=item 2.
40bec8a5 610That is, each element smartmatches the element of the same index in the other array.[3]
1ca345ed
TC
611
612=item 3.
613If a circular reference is found, fall back to referential equality.
614
615=item 4.
616Either an actual number, or a string that looks like one.
617
618=back
619
620The smartmatch implicitly dereferences any non-blessed hash or array
621reference, so the C<I<HASH>> and C<I<ARRAY>> entries apply in those cases.
622For blessed references, the C<I<Object>> entries apply. Smartmatches
623involving hashes only consider hash keys, never hash values.
624
625The "like" code entry is not always an exact rendition. For example, the
40bec8a5 626smartmatch operator short-circuits whenever possible, but C<grep> does
1ca345ed
TC
627not. Also, C<grep> in scalar context returns the number of matches, but
628C<~~> returns only true or false.
629
630Unlike most operators, the smartmatch operator knows to treat C<undef>
631specially:
632
633 use v5.10.1;
634 @array = (1, 2, 3, undef, 4, 5);
635 say "some elements undefined" if undef ~~ @array;
636
637Each operand is considered in a modified scalar context, the modification
638being that array and hash variables are passed by reference to the
639operator, which implicitly dereferences them. Both elements
640of each pair are the same:
641
642 use v5.10.1;
643
644 my %hash = (red => 1, blue => 2, green => 3,
645 orange => 4, yellow => 5, purple => 6,
646 black => 7, grey => 8, white => 9);
647
648 my @array = qw(red blue green);
649
650 say "some array elements in hash keys" if @array ~~ %hash;
651 say "some array elements in hash keys" if \@array ~~ \%hash;
652
653 say "red in array" if "red" ~~ @array;
654 say "red in array" if "red" ~~ \@array;
655
656 say "some keys end in e" if /e\$/ ~~ %hash;
657 say "some keys end in e" if /e\$/ ~~ \%hash;
658
40bec8a5
TC
659Two arrays smartmatch if each element in the first array smartmatches
660(that is, is "in") the corresponding element in the second array,
661recursively.
1ca345ed
TC
662
663 use v5.10.1;
664 my @little = qw(red blue green);
665 my @bigger = ("red", "blue", [ "orange", "green" ] );
666 if (@little ~~ @bigger) { # true!
667 say "little is contained in bigger";
668 }
669
670Because the smartmatch operator recurses on nested arrays, this
671will still report that "red" is in the array.
672
673 use v5.10.1;
674 my @array = qw(red blue green);
675 my \$nested_array = [[[[[[[ @array ]]]]]]];
676 say "red in array" if "red" ~~ \$nested_array;
677
678If two arrays smartmatch each other, then they are deep
679copies of each others' values, as this example reports:
680
681 use v5.12.0;
682 my @a = (0, 1, 2, [3, [4, 5], 6], 7);
683 my @b = (0, 1, 2, [3, [4, 5], 6], 7);
684
685 if (@a ~~ @b && @b ~~ @a) {
686 say "a and b are deep copies of each other";
687 }
688 elsif (@a ~~ @b) {
689 say "a smartmatches in b";
690 }
691 elsif (@b ~~ @a) {
692 say "b smartmatches in a";
693 }
694 else {
695 say "a and b don't smartmatch each other at all";
696 }
697
698
699If you were to set C<\$b[3] = 4>, then instead of reporting that "a and b
700are deep copies of each other", it now reports that "b smartmatches in a".
701That because the corresponding position in C<@a> contains an array that
702(eventually) has a 4 in it.
703
704Smartmatching one hash against another reports whether both contain the
705same keys, no more and no less. This could be used to see whether two
706records have the same field names, without caring what values those fields
707might have. For example:
708
709 use v5.10.1;
710 sub make_dogtag {
711 state \$REQUIRED_FIELDS = { name=>1, rank=>1, serial_num=>1 };
712
713 my (\$class, \$init_fields) = @_;
714
715 die "Must supply (only) name, rank, and serial number"
716 unless \$init_fields ~~ \$REQUIRED_FIELDS;
717
718 ...
719 }
720
721or, if other non-required fields are allowed, use ARRAY ~~ HASH:
722
723 use v5.10.1;
724 sub make_dogtag {
725 state \$REQUIRED_FIELDS = { name=>1, rank=>1, serial_num=>1 };
726
727 my (\$class, \$init_fields) = @_;
728
729 die "Must supply (at least) name, rank, and serial number"
730 unless [keys %{\$init_fields}] ~~ \$REQUIRED_FIELDS;
731
732 ...
733 }
734
735The smartmatch operator is most often used as the implicit operator of a
736C<when> clause. See the section on "Switch Statements" in L<perlsyn>.
737
738=head3 Smartmatching of Objects
739
40bec8a5
TC
740To avoid relying on an object's underlying representation, if the
741smartmatch's right operand is an object that doesn't overload C<~~>,
742it raises the exception "C<Smartmatching a non-overloaded object
743breaks encapsulation>". That's because one has no business digging
744around to see whether something is "in" an object. These are all
745illegal on objects without a C<~~> overload:
1ca345ed
TC
746
747 %hash ~~ \$object
748 42 ~~ \$object
749 "fred" ~~ \$object
750
751However, you can change the way an object is smartmatched by overloading
752the C<~~> operator. This is allowed to extend the usual smartmatch semantics.
753For objects that do have an C<~~> overload, see L<overload>.
754
755Using an object as the left operand is allowed, although not very useful.
756Smartmatching rules take precedence over overloading, so even if the
757object in the left operand has smartmatch overloading, this will be
758ignored. A left operand that is a non-overloaded object falls back on a
759string or numeric comparison of whatever the C<ref> operator returns. That
760means that
761
762 \$object ~~ X
763
764does I<not> invoke the overload method with C<I<X>> as an argument.
765Instead the above table is consulted as normal, and based on the type of
766C<I<X>>, overloading may or may not be invoked. For simple strings or
767numbers, in becomes equivalent to this:
768
769 \$object ~~ \$number ref(\$object) == \$number
770 \$object ~~ \$string ref(\$object) eq \$string
771
772For example, this reports that the handle smells IOish
773(but please don't really do this!):
774
775 use IO::Handle;
776 my \$fh = IO::Handle->new();
777 if (\$fh ~~ /\bIO\b/) {
778 say "handle smells IOish";
779 }
780
781That's because it treats C<\$fh> as a string like
782C<"IO::Handle=GLOB(0x8039e0)">, then pattern matches against that.
a034a98d 783
a0d0e21e 784=head2 Bitwise And
d74e8afc 785X<operator, bitwise, and> X<bitwise and> X<&>
a0d0e21e 786
2cdc098b 787Binary "&" returns its operands ANDed together bit by bit.
2c268ad5 788(See also L<Integer Arithmetic> and L<Bitwise String Operators>.)
a0d0e21e 789
2cdc098b 790Note that "&" has lower priority than relational operators, so for example
1ca345ed 791the parentheses are essential in a test like
2cdc098b 792
1ca345ed 793 print "Even\n" if (\$x & 1) == 0;
2cdc098b 794
a0d0e21e 795=head2 Bitwise Or and Exclusive Or
d74e8afc
ITB
796X<operator, bitwise, or> X<bitwise or> X<|> X<operator, bitwise, xor>
797X<bitwise xor> X<^>
a0d0e21e 798
2cdc098b 799Binary "|" returns its operands ORed together bit by bit.
2c268ad5 800(See also L<Integer Arithmetic> and L<Bitwise String Operators>.)
a0d0e21e 801
2cdc098b 802Binary "^" returns its operands XORed together bit by bit.
2c268ad5 803(See also L<Integer Arithmetic> and L<Bitwise String Operators>.)
a0d0e21e 804
2cdc098b
MG
805Note that "|" and "^" have lower priority than relational operators, so
806for example the brackets are essential in a test like
807
1ca345ed 808 print "false\n" if (8 | 2) != 10;
2cdc098b 809
a0d0e21e 810=head2 C-style Logical And
d74e8afc 811X<&&> X<logical and> X<operator, logical, and>
a0d0e21e
LW
812
813Binary "&&" performs a short-circuit logical AND operation. That is,
814if the left operand is false, the right operand is not even evaluated.
815Scalar or list context propagates down to the right operand if it
816is evaluated.
817
818=head2 C-style Logical Or
d74e8afc 819X<||> X<operator, logical, or>
a0d0e21e
LW
820
821Binary "||" performs a short-circuit logical OR operation. That is,
822if the left operand is true, the right operand is not even evaluated.
823Scalar or list context propagates down to the right operand if it
824is evaluated.
825
26d9d83b 826=head2 Logical Defined-Or
d74e8afc 827X<//> X<operator, logical, defined-or>
c963b151
BD
828
829Although it has no direct equivalent in C, Perl's C<//> operator is related
89d205f2 830to its C-style or. In fact, it's exactly the same as C<||>, except that it
95bee9ba
A
831tests the left hand side's definedness instead of its truth. Thus,
832C<< EXPR1 // EXPR2 >> returns the value of C<< EXPR1 >> if it's defined,
833otherwise, the value of C<< EXPR2 >> is returned. (C<< EXPR1 >> is evaluated
834in scalar context, C<< EXPR2 >> in the context of C<< // >> itself). Usually,
835this is the same result as C<< defined(EXPR1) ? EXPR1 : EXPR2 >> (except that
836the ternary-operator form can be used as a lvalue, while C<< EXPR1 // EXPR2 >>
837cannot). This is very useful for
bdc7923b
RGS
838providing default values for variables. If you actually want to test if
839at least one of C<\$a> and C<\$b> is defined, use C<defined(\$a // \$b)>.
c963b151 840
d042e63d
MS
841The C<||>, C<//> and C<&&> operators return the last value evaluated
842(unlike C's C<||> and C<&&>, which return 0 or 1). Thus, a reasonably
843portable way to find out the home directory might be:
a0d0e21e 844
c543c01b
TC
845 \$home = \$ENV{HOME}
846 // \$ENV{LOGDIR}
847 // (getpwuid(\$<))[7]
848 // die "You're homeless!\n";
a0d0e21e 849
5a964f20
TC
850In particular, this means that you shouldn't use this
851for selecting between two aggregates for assignment:
852
853 @a = @b || @c; # this is wrong
854 @a = scalar(@b) || @c; # really meant this
855 @a = @b ? @b : @c; # this works fine, though
856
1ca345ed 857As alternatives to C<&&> and C<||> when used for
f23102e2
RGS
858control flow, Perl provides the C<and> and C<or> operators (see below).
859The short-circuit behavior is identical. The precedence of "and"
c963b151 860and "or" is much lower, however, so that you can safely use them after a
5a964f20 861list operator without the need for parentheses:
a0d0e21e
LW
862
863 unlink "alpha", "beta", "gamma"
864 or gripe(), next LINE;
865
866With the C-style operators that would have been written like this:
867
868 unlink("alpha", "beta", "gamma")
869 || (gripe(), next LINE);
870
1ca345ed
TC
871It would be even more readable to write that this way:
872
873 unless(unlink("alpha", "beta", "gamma")) {
874 gripe();
875 next LINE;
876 }
877
eeb6a2c9 878Using "or" for assignment is unlikely to do what you want; see below.
5a964f20
TC
879
880=head2 Range Operators
d74e8afc 881X<operator, range> X<range> X<..> X<...>
a0d0e21e
LW
882
883Binary ".." is the range operator, which is really two different
fb53bbb2 884operators depending on the context. In list context, it returns a
54ae734e 885list of values counting (up by ones) from the left value to the right
2cdbc966 886value. If the left value is greater than the right value then it
fb53bbb2 887returns the empty list. The range operator is useful for writing
54ae734e 888C<foreach (1..10)> loops and for doing slice operations on arrays. In
2cdbc966
JD
889the current implementation, no temporary array is created when the
890range operator is used as the expression in C<foreach> loops, but older
891versions of Perl might burn a lot of memory when you write something
892like this:
a0d0e21e
LW
893
894 for (1 .. 1_000_000) {
895 # code
54310121 896 }
a0d0e21e 897
8f0f46f8 898The range operator also works on strings, using the magical
899auto-increment, see below.
54ae734e 900
5a964f20 901In scalar context, ".." returns a boolean value. The operator is
8f0f46f8 902bistable, like a flip-flop, and emulates the line-range (comma)
903operator of B<sed>, B<awk>, and various editors. Each ".." operator
904maintains its own boolean state, even across calls to a subroutine
905that contains it. It is false as long as its left operand is false.
a0d0e21e
LW
906Once the left operand is true, the range operator stays true until the
907right operand is true, I<AFTER> which the range operator becomes false
8f0f46f8 908again. It doesn't become false till the next time the range operator
909is evaluated. It can test the right operand and become false on the
910same evaluation it became true (as in B<awk>), but it still returns
911true once. If you don't want it to test the right operand until the
912next evaluation, as in B<sed>, just use three dots ("...") instead of
19799a22
GS
913two. In all other regards, "..." behaves just like ".." does.
914
915The right operand is not evaluated while the operator is in the
916"false" state, and the left operand is not evaluated while the
917operator is in the "true" state. The precedence is a little lower
918than || and &&. The value returned is either the empty string for
8f0f46f8 919false, or a sequence number (beginning with 1) for true. The sequence
920number is reset for each range encountered. The final sequence number
921in a range has the string "E0" appended to it, which doesn't affect
922its numeric value, but gives you something to search for if you want
923to exclude the endpoint. You can exclude the beginning point by
924waiting for the sequence number to be greater than 1.
df5f8116
CW
925
926If either operand of scalar ".." is a constant expression,
927that operand is considered true if it is equal (C<==>) to the current
928input line number (the C<\$.> variable).
929
930To be pedantic, the comparison is actually C<int(EXPR) == int(EXPR)>,
931but that is only an issue if you use a floating point expression; when
932implicitly using C<\$.> as described in the previous paragraph, the
933comparison is C<int(EXPR) == int(\$.)> which is only an issue when C<\$.>
934is set to a floating point value and you are not reading from a file.
935Furthermore, C<"span" .. "spat"> or C<2.18 .. 3.14> will not do what
936you want in scalar context because each of the operands are evaluated
937using their integer representation.
938
939Examples:
a0d0e21e
LW
940
941As a scalar operator:
942
df5f8116 943 if (101 .. 200) { print; } # print 2nd hundred lines, short for
950b09ed 944 # if (\$. == 101 .. \$. == 200) { print; }
9f10b797
RGS
945
946 next LINE if (1 .. /^\$/); # skip header lines, short for
f343f960 947 # next LINE if (\$. == 1 .. /^\$/);
9f10b797
RGS
948 # (typically in a loop labeled LINE)
949
950 s/^/> / if (/^\$/ .. eof()); # quote body
a0d0e21e 951
5a964f20
TC
952 # parse mail messages
953 while (<>) {
954 \$in_header = 1 .. /^\$/;
df5f8116
CW
955 \$in_body = /^\$/ .. eof;
956 if (\$in_header) {
f343f960 957 # do something
df5f8116 958 } else { # in body
f343f960 959 # do something else
df5f8116 960 }
5a964f20 961 } continue {
df5f8116 962 close ARGV if eof; # reset \$. each file
5a964f20
TC
963 }
964
acf31ca5
SF
965Here's a simple example to illustrate the difference between
966the two range operators:
967
968 @lines = (" - Foo",
969 "01 - Bar",
970 "1 - Baz",
971 " - Quux");
972
9f10b797
RGS
973 foreach (@lines) {
974 if (/0/ .. /1/) {
acf31ca5
SF
975 print "\$_\n";
976 }
977 }
978
9f10b797
RGS
979This program will print only the line containing "Bar". If
980the range operator is changed to C<...>, it will also print the
acf31ca5
SF
981"Baz" line.
982
983And now some examples as a list operator:
a0d0e21e 984
1ca345ed
TC
985 for (101 .. 200) { print } # print \$_ 100 times
986 @foo = @foo[0 .. \$#foo]; # an expensive no-op
987 @foo = @foo[\$#foo-4 .. \$#foo]; # slice last 5 items
a0d0e21e 988
5a964f20 989The range operator (in list context) makes use of the magical
5f05dabc 990auto-increment algorithm if the operands are strings. You
a0d0e21e
LW
991can say
992
c543c01b 993 @alphabet = ("A" .. "Z");
a0d0e21e 994
54ae734e 995to get all normal letters of the English alphabet, or
a0d0e21e 996
c543c01b 997 \$hexdigit = (0 .. 9, "a" .. "f")[\$num & 15];
a0d0e21e
LW
998
999to get a hexadecimal digit, or
1000
1ca345ed
TC
1001 @z2 = ("01" .. "31");
1002 print \$z2[\$mday];
a0d0e21e 1003
ea4f5703
YST
1004to get dates with leading zeros.
1005
1006If the final value specified is not in the sequence that the magical
1007increment would produce, the sequence goes until the next value would
1008be longer than the final value specified.
1009
1010If the initial value specified isn't part of a magical increment
c543c01b 1011sequence (that is, a non-empty string matching C</^[a-zA-Z]*[0-9]*\z/>),
ea4f5703
YST
1012only the initial value will be returned. So the following will only
1013return an alpha:
1014
c543c01b 1015 use charnames "greek";
ea4f5703
YST
1016 my @greek_small = ("\N{alpha}" .. "\N{omega}");
1017
c543c01b
TC
1018To get the 25 traditional lowercase Greek letters, including both sigmas,
1019you could use this instead:
ea4f5703 1020
c543c01b 1021 use charnames "greek";
1ca345ed
TC
1022 my @greek_small = map { chr } ( ord("\N{alpha}")
1023 ..
1024 ord("\N{omega}")
1025 );
c543c01b
TC
1026
1027However, because there are I<many> other lowercase Greek characters than
1028just those, to match lowercase Greek characters in a regular expression,
1029you would use the pattern C</(?:(?=\p{Greek})\p{Lower})+/>.
a0d0e21e 1030
df5f8116
CW
1031Because each operand is evaluated in integer form, C<2.18 .. 3.14> will
1032return two elements in list context.
1033
1034 @list = (2.18 .. 3.14); # same as @list = (2 .. 3);
1035
a0d0e21e 1036=head2 Conditional Operator
d74e8afc 1037X<operator, conditional> X<operator, ternary> X<ternary> X<?:>
a0d0e21e
LW
1038
1039Ternary "?:" is the conditional operator, just as in C. It works much
1040like an if-then-else. If the argument before the ? is true, the
1041argument before the : is returned, otherwise the argument after the :
cb1a09d0
AD
1042is returned. For example:
1043
54310121 1044 printf "I have %d dog%s.\n", \$n,
c543c01b 1045 (\$n == 1) ? "" : "s";
cb1a09d0
AD
1046
1047Scalar or list context propagates downward into the 2nd
54310121 1048or 3rd argument, whichever is selected.
cb1a09d0
AD
1049
1050 \$a = \$ok ? \$b : \$c; # get a scalar
1051 @a = \$ok ? @b : @c; # get an array
1052 \$a = \$ok ? @b : @c; # oops, that's just a count!
1053
1054The operator may be assigned to if both the 2nd and 3rd arguments are
1055legal lvalues (meaning that you can assign to them):
a0d0e21e
LW
1056
1057 (\$a_or_b ? \$a : \$b) = \$c;
1058
5a964f20
TC
1059Because this operator produces an assignable result, using assignments
1060without parentheses will get you in trouble. For example, this:
1061
1062 \$a % 2 ? \$a += 10 : \$a += 2
1063
1064Really means this:
1065
1066 ((\$a % 2) ? (\$a += 10) : \$a) += 2
1067
1068Rather than this:
1069
1070 (\$a % 2) ? (\$a += 10) : (\$a += 2)
1071
19799a22
GS
1072That should probably be written more simply as:
1073
1074 \$a += (\$a % 2) ? 10 : 2;
1075
4633a7c4 1076=head2 Assignment Operators
d74e8afc 1077X<assignment> X<operator, assignment> X<=> X<**=> X<+=> X<*=> X<&=>
5ac3b81c 1078X<<< <<= >>> X<&&=> X<-=> X</=> X<|=> X<<< >>= >>> X<||=> X<//=> X<.=>
d74e8afc 1079X<%=> X<^=> X<x=>
a0d0e21e
LW
1080
1081"=" is the ordinary assignment operator.
1082
1083Assignment operators work as in C. That is,
1084
1085 \$a += 2;
1086
1087is equivalent to
1088
1089 \$a = \$a + 2;
1090
1091although without duplicating any side effects that dereferencing the lvalue
54310121 1092might trigger, such as from tie(). Other assignment operators work similarly.
1093The following are recognized:
a0d0e21e
LW
1094
1095 **= += *= &= <<= &&=
9f10b797
RGS
1096 -= /= |= >>= ||=
1097 .= %= ^= //=
1098 x=
a0d0e21e 1099
19799a22 1100Although these are grouped by family, they all have the precedence
a0d0e21e
LW
1101of assignment.
1102
b350dd2f
GS
1103Unlike in C, the scalar assignment operator produces a valid lvalue.
1104Modifying an assignment is equivalent to doing the assignment and
1105then modifying the variable that was assigned to. This is useful
1106for modifying a copy of something, like this:
a0d0e21e 1107
1ca345ed
TC
1108 (\$tmp = \$global) =~ tr/13579/24680/;
1109
1110Although as of 5.14, that can be also be accomplished this way:
1111
1112 use v5.14;
1113 \$tmp = (\$global =~ tr/13579/24680/r);
a0d0e21e
LW
1114
1115Likewise,
1116
1117 (\$a += 2) *= 3;
1118
1119is equivalent to
1120
1121 \$a += 2;
1122 \$a *= 3;
1123
b350dd2f
GS
1124Similarly, a list assignment in list context produces the list of
1125lvalues assigned to, and a list assignment in scalar context returns
1126the number of elements produced by the expression on the right hand
1127side of the assignment.
1128
748a9306 1129=head2 Comma Operator
d74e8afc 1130X<comma> X<operator, comma> X<,>
a0d0e21e 1131
5a964f20 1132Binary "," is the comma operator. In scalar context it evaluates
a0d0e21e
LW
1133its left argument, throws that value away, then evaluates its right
1134argument and returns that value. This is just like C's comma operator.
1135
5a964f20 1136In list context, it's just the list argument separator, and inserts
ed5c6d31
PJ
1137both its arguments into the list. These arguments are also evaluated
1138from left to right.
a0d0e21e 1139
4e1988c6
FC
1140The C<< => >> operator is a synonym for the comma except that it causes a
1141word on its left to be interpreted as a string if it begins with a letter
344f2c40
IG
1142or underscore and is composed only of letters, digits and underscores.
1143This includes operands that might otherwise be interpreted as operators,
1144constants, single number v-strings or function calls. If in doubt about
c543c01b 1145this behavior, the left operand can be quoted explicitly.
344f2c40
IG
1146
1147Otherwise, the C<< => >> operator behaves exactly as the comma operator
1148or list argument separator, according to context.
1149
1150For example:
a44e5664
MS
1151
1152 use constant FOO => "something";
1153
1154 my %h = ( FOO => 23 );
1155
1156is equivalent to:
1157
1158 my %h = ("FOO", 23);
1159
1160It is I<NOT>:
1161
1162 my %h = ("something", 23);
1163
719b43e8
RGS
1164The C<< => >> operator is helpful in documenting the correspondence
1165between keys and values in hashes, and other paired elements in lists.
748a9306 1166
a12b8f3c
FC
1167 %hash = ( \$key => \$value );
1168 login( \$username => \$password );
a44e5664 1169
4e1988c6
FC
1170The special quoting behavior ignores precedence, and hence may apply to
1171I<part> of the left operand:
1172
1173 print time.shift => "bbb";
1174
1175That example prints something like "1314363215shiftbbb", because the
1176C<< => >> implicitly quotes the C<shift> immediately on its left, ignoring
1177the fact that C<time.shift> is the entire left operand.
1178
a0d0e21e 1179=head2 List Operators (Rightward)
d74e8afc 1180X<operator, list, rightward> X<list operator>
a0d0e21e 1181
c543c01b 1182On the right side of a list operator, the comma has very low precedence,
a0d0e21e
LW
1183such that it controls all comma-separated expressions found there.
1184The only operators with lower precedence are the logical operators
1185"and", "or", and "not", which may be used to evaluate calls to list
1ca345ed
TC
1186operators without the need for parentheses:
1187
1188 open HANDLE, "< :utf8", "filename" or die "Can't open: \$!\n";
1189
1190However, some people find that code harder to read than writing
1191it with parentheses:
1192
1193 open(HANDLE, "< :utf8", "filename") or die "Can't open: \$!\n";
1194
1195in which case you might as well just use the more customary "||" operator:
a0d0e21e 1196
1ca345ed 1197 open(HANDLE, "< :utf8", "filename") || die "Can't open: \$!\n";
a0d0e21e 1198
5ba421f6 1199See also discussion of list operators in L<Terms and List Operators (Leftward)>.
a0d0e21e
LW
1200
1201=head2 Logical Not
d74e8afc 1202X<operator, logical, not> X<not>
a0d0e21e
LW
1203
1204Unary "not" returns the logical negation of the expression to its right.
1205It's the equivalent of "!" except for the very low precedence.
1206
1207=head2 Logical And
d74e8afc 1208X<operator, logical, and> X<and>
a0d0e21e
LW
1209
1210Binary "and" returns the logical conjunction of the two surrounding
c543c01b
TC
1211expressions. It's equivalent to C<&&> except for the very low
1212precedence. This means that it short-circuits: the right
a0d0e21e
LW
1213expression is evaluated only if the left expression is true.
1214
59ab9d6e 1215=head2 Logical or and Exclusive Or
f23102e2 1216X<operator, logical, or> X<operator, logical, xor>
59ab9d6e 1217X<operator, logical, exclusive or>
f23102e2 1218X<or> X<xor>
a0d0e21e
LW
1219
1220Binary "or" returns the logical disjunction of the two surrounding
c543c01b
TC
1221expressions. It's equivalent to C<||> except for the very low precedence.
1222This makes it useful for control flow:
5a964f20
TC
1223
1224 print FH \$data or die "Can't write to FH: \$!";
1225
c543c01b
TC
1226This means that it short-circuits: the right expression is evaluated
1227only if the left expression is false. Due to its precedence, you must
1228be careful to avoid using it as replacement for the C<||> operator.
1229It usually works out better for flow control than in assignments:
5a964f20
TC
1230
1231 \$a = \$b or \$c; # bug: this is wrong
1232 (\$a = \$b) or \$c; # really means this
1233 \$a = \$b || \$c; # better written this way
1234
19799a22 1235However, when it's a list-context assignment and you're trying to use
c543c01b 1236C<||> for control flow, you probably need "or" so that the assignment
5a964f20
TC
1237takes higher precedence.
1238
1239 @info = stat(\$file) || die; # oops, scalar sense of stat!
1240 @info = stat(\$file) or die; # better, now @info gets its due
1241
c963b151
BD
1242Then again, you could always use parentheses.
1243
1ca345ed 1244Binary C<xor> returns the exclusive-OR of the two surrounding expressions.
c543c01b 1245It cannot short-circuit (of course).
a0d0e21e 1246
59ab9d6e
MB
1247There is no low precedence operator for defined-OR.
1248
a0d0e21e 1249=head2 C Operators Missing From Perl
d74e8afc
ITB
1250X<operator, missing from perl> X<&> X<*>
1251X<typecasting> X<(TYPE)>
a0d0e21e
LW
1252
1253Here is what C has that Perl doesn't:
1254
1255=over 8
1256
1257=item unary &
1258
1259Address-of operator. (But see the "\" operator for taking a reference.)
1260
1261=item unary *
1262
54310121 1263Dereference-address operator. (Perl's prefix dereferencing
a0d0e21e
LW
1264operators are typed: \$, @, %, and &.)
1265
1266=item (TYPE)
1267
19799a22 1268Type-casting operator.
a0d0e21e
LW
1269
1270=back
1271
5f05dabc 1272=head2 Quote and Quote-like Operators
89d205f2 1273X<operator, quote> X<operator, quote-like> X<q> X<qq> X<qx> X<qw> X<m>
d74e8afc
ITB
1274X<qr> X<s> X<tr> X<'> X<''> X<"> X<""> X<//> X<`> X<``> X<<< << >>>
1275X<escape sequence> X<escape>
1276
a0d0e21e
LW
1277While we usually think of quotes as literal values, in Perl they
1278function as operators, providing various kinds of interpolating and
1279pattern matching capabilities. Perl provides customary quote characters
1280for these behaviors, but also provides a way for you to choose your
1281quote character for any of them. In the following table, a C<{}> represents
9f10b797 1282any pair of delimiters you choose.
a0d0e21e 1283
2c268ad5
TP
1284 Customary Generic Meaning Interpolates
1285 '' q{} Literal no
1286 "" qq{} Literal yes
af9219ee 1287 `` qx{} Command yes*
2c268ad5 1288 qw{} Word list no
af9219ee
MG
1289 // m{} Pattern match yes*
1290 qr{} Pattern yes*
1291 s{}{} Substitution yes*
2c268ad5 1292 tr{}{} Transliteration no (but see below)
c543c01b 1293 y{}{} Transliteration no (but see below)
7e3b091d 1294 <<EOF here-doc yes*
a0d0e21e 1295
af9219ee
MG
1296 * unless the delimiter is ''.
1297
87275199 1298Non-bracketing delimiters use the same character fore and aft, but the four
c543c01b 1299sorts of ASCII brackets (round, angle, square, curly) all nest, which means
9f10b797 1300that
87275199 1301
c543c01b 1302 q{foo{bar}baz}
35f2feb0 1303
9f10b797 1304is the same as
87275199 1305
c543c01b 1306 'foo{bar}baz'
87275199
GS
1307
1308Note, however, that this does not always work for quoting Perl code:
1309
c543c01b 1310 \$s = q{ if(\$a eq "}") ... }; # WRONG
87275199 1311
c543c01b
TC
1312is a syntax error. The C<Text::Balanced> module (standard as of v5.8,
1313and from CPAN before then) is able to do this properly.
87275199 1314
19799a22 1315There can be whitespace between the operator and the quoting
fb73857a 1316characters, except when C<#> is being used as the quoting character.
19799a22
GS
1317C<q#foo#> is parsed as the string C<foo>, while C<q #foo#> is the
1318operator C<q> followed by a comment. Its argument will be taken
1319from the next line. This allows you to write:
fb73857a 1320
1321 s {foo} # Replace foo
1322 {bar} # with bar.
1323
c543c01b
TC
1324The following escape sequences are available in constructs that interpolate,
1325and in transliterations:
5691ca5f 1326X<\t> X<\n> X<\r> X<\f> X<\b> X<\a> X<\e> X<\x> X<\0> X<\c> X<\N> X<\N{}>
04341565 1327X<\o{}>
5691ca5f 1328
2c4c1ff2
KW
1329 Sequence Note Description
1330 \t tab (HT, TAB)
1331 \n newline (NL)
1332 \r return (CR)
1333 \f form feed (FF)
1334 \b backspace (BS)
1335 \a alarm (bell) (BEL)
1336 \e escape (ESC)
c543c01b 1337 \x{263A} [1,8] hex char (example: SMILEY)
2c4c1ff2 1338 \x1b [2,8] restricted range hex char (example: ESC)
fb121860 1339 \N{name} [3] named Unicode character or character sequence
2c4c1ff2
KW
1340 \N{U+263D} [4,8] Unicode character (example: FIRST QUARTER MOON)
1341 \c[ [5] control char (example: chr(27))
1342 \o{23072} [6,8] octal char (example: SMILEY)
1343 \033 [7,8] restricted range octal char (example: ESC)
5691ca5f
KW
1344
1345=over 4
1346
1347=item [1]
1348
2c4c1ff2
KW
1349The result is the character specified by the hexadecimal number between
1350the braces. See L</[8]> below for details on which character.
96448467
DG
1351
1352Only hexadecimal digits are valid between the braces. If an invalid
1353character is encountered, a warning will be issued and the invalid
1354character and all subsequent characters (valid or invalid) within the
1355braces will be discarded.
1356
1357If there are no valid digits between the braces, the generated character is
1358the NULL character (C<\x{00}>). However, an explicit empty brace (C<\x{}>)
c543c01b 1359will not cause a warning (currently).
40687185
KW
1360
1361=item [2]
1362
2c4c1ff2
KW
1363The result is the character specified by the hexadecimal number in the range
13640x00 to 0xFF. See L</[8]> below for details on which character.
96448467
DG
1365
1366Only hexadecimal digits are valid following C<\x>. When C<\x> is followed
2c4c1ff2 1367by fewer than two valid digits, any valid digits will be zero-padded. This
c543c01b 1368means that C<\x7> will be interpreted as C<\x07>, and a lone <\x> will be
2c4c1ff2 1369interpreted as C<\x00>. Except at the end of a string, having fewer than
c543c01b 1370two valid digits will result in a warning. Note that although the warning
96448467
DG
1371says the illegal character is ignored, it is only ignored as part of the
1372escape and will still be used as the subsequent character in the string.
1373For example:
1374
1375 Original Result Warns?
1376 "\x7" "\x07" no
1377 "\x" "\x00" no
1378 "\x7q" "\x07q" yes
1379 "\xq" "\x00q" yes
1380
40687185
KW
1381=item [3]
1382
fb121860 1383The result is the Unicode character or character sequence given by I<name>.
2c4c1ff2 1384See L<charnames>.
40687185
KW
1385
1386=item [4]
1387
2c4c1ff2
KW
1388C<\N{U+I<hexadecimal number>}> means the Unicode character whose Unicode code
1389point is I<hexadecimal number>.
40687185
KW
1390
1391=item [5]
1392
5691ca5f
KW
1393The character following C<\c> is mapped to some other character as shown in the
1394table:
1395
1396 Sequence Value
1397 \c@ chr(0)
1398 \cA chr(1)
1399 \ca chr(1)
1400 \cB chr(2)
1401 \cb chr(2)
1402 ...
1403 \cZ chr(26)
1404 \cz chr(26)
1405 \c[ chr(27)
1406 \c] chr(29)
1407 \c^ chr(30)
1408 \c? chr(127)
1409
d813941f
KW
1410In other words, it's the character whose code point has had 64 xor'd with
1411its uppercase. C<\c?> is DELETE because C<ord("@") ^ 64> is 127, and
1412C<\c@> is NULL because the ord of "@" is 64, so xor'ing 64 itself produces 0.
1413
5691ca5f
KW
1414Also, C<\c\I<X>> yields C< chr(28) . "I<X>"> for any I<X>, but cannot come at the
1415end of a string, because the backslash would be parsed as escaping the end
1416quote.
1417
1418On ASCII platforms, the resulting characters from the list above are the
1419complete set of ASCII controls. This isn't the case on EBCDIC platforms; see
1420L<perlebcdic/OPERATOR DIFFERENCES> for the complete list of what these
1421sequences mean on both ASCII and EBCDIC platforms.
1422
1423Use of any other character following the "c" besides those listed above is
17a3df4c 1424discouraged, and some are deprecated with the intention of removing
9fef6a0d 1425those in a later Perl version. What happens for any of these
d813941f
KW
1426other characters currently though, is that the value is derived by xor'ing
1427with the seventh bit, which is 64.
5691ca5f
KW
1428
1429To get platform independent controls, you can use C<\N{...}>.
1430
40687185
KW
1431=item [6]
1432
2c4c1ff2
KW
1433The result is the character specified by the octal number between the braces.
1434See L</[8]> below for details on which character.
04341565
DG
1435
1436If a character that isn't an octal digit is encountered, a warning is raised,
1437and the value is based on the octal digits before it, discarding it and all
1438following characters up to the closing brace. It is a fatal error if there are
1439no octal digits at all.
1440
1441=item [7]
1442
c543c01b 1443The result is the character specified by the three-digit octal number in the
2c4c1ff2
KW
1444range 000 to 777 (but best to not use above 077, see next paragraph). See
1445L</[8]> below for details on which character.
1446
1447Some contexts allow 2 or even 1 digit, but any usage without exactly
40687185 1448three digits, the first being a zero, may give unintended results. (For
5db3e519
FC
1449example, in a regular expression it may be confused with a backreference;
1450see L<perlrebackslash/Octal escapes>.) Starting in Perl 5.14, you may
c543c01b 1451use C<\o{}> instead, which avoids all these problems. Otherwise, it is best to
04341565
DG
1452use this construct only for ordinals C<\077> and below, remembering to pad to
1453the left with zeros to make three digits. For larger ordinals, either use
9fef6a0d 1454C<\o{}>, or convert to something else, such as to hex and use C<\x{}>
04341565 1455instead.
40687185 1456
40687185
KW
1457Having fewer than 3 digits may lead to a misleading warning message that says
1458that what follows is ignored. For example, C<"\128"> in the ASCII character set
1459is equivalent to the two characters C<"\n8">, but the warning C<Illegal octal
5db3e519
FC
1460digit '8' ignored> will be thrown. If C<"\n8"> is what you want, you can
1461avoid this warning by padding your octal number with C<0>'s: C<"\0128">.
5691ca5f 1462
2c4c1ff2
KW
1463=item [8]
1464
c543c01b 1465Several constructs above specify a character by a number. That number
2c4c1ff2 1466gives the character's position in the character set encoding (indexed from 0).
c543c01b 1467This is called synonymously its ordinal, code position, or code point. Perl
2c4c1ff2
KW
1468works on platforms that have a native encoding currently of either ASCII/Latin1
1469or EBCDIC, each of which allow specification of 256 characters. In general, if
1470the number is 255 (0xFF, 0377) or below, Perl interprets this in the platform's
1471native encoding. If the number is 256 (0x100, 0400) or above, Perl interprets
c543c01b 1472it as a Unicode code point and the result is the corresponding Unicode
2c4c1ff2
KW
1473character. For example C<\x{50}> and C<\o{120}> both are the number 80 in
1474decimal, which is less than 256, so the number is interpreted in the native
1475character set encoding. In ASCII the character in the 80th position (indexed
1476from 0) is the letter "P", and in EBCDIC it is the ampersand symbol "&".
1477C<\x{100}> and C<\o{400}> are both 256 in decimal, so the number is interpreted
1478as a Unicode code point no matter what the native encoding is. The name of the
9fef6a0d 1479character in the 256th position (indexed by 0) in Unicode is
2c4c1ff2
KW
1480C<LATIN CAPITAL LETTER A WITH MACRON>.
1481
9fef6a0d 1482There are a couple of exceptions to the above rule. S<C<\N{U+I<hex number>}>> is
2c4c1ff2
KW
1483always interpreted as a Unicode code point, so that C<\N{U+0050}> is "P" even
1484on EBCDIC platforms. And if L<C<S<use encoding>>|encoding> is in effect, the
1485number is considered to be in that encoding, and is translated from that into
1486the platform's native encoding if there is a corresponding native character;
1487otherwise to Unicode.
1488
5691ca5f 1489=back
4c77eaa2 1490
e526e8bb
KW
1491B<NOTE>: Unlike C and other languages, Perl has no C<\v> escape sequence for
1492the vertical tab (VT - ASCII 11), but you may use C<\ck> or C<\x0b>. (C<\v>
1493does have meaning in regular expression patterns in Perl, see L<perlre>.)
1494
1495The following escape sequences are available in constructs that interpolate,
904501ec 1496but not in transliterations.
628253b8 1497X<\l> X<\u> X<\L> X<\U> X<\E> X<\Q> X<\F>
904501ec 1498
c543c01b
TC
1499 \l lowercase next character only
1500 \u titlecase (not uppercase!) next character only
e4d34742
EB
1501 \L lowercase all characters till \E or end of string
1502 \U uppercase all characters till \E or end of string
628253b8 1503 \F foldcase all characters till \E or end of string
736fe711
KW
1504 \Q quote (disable) pattern metacharacters till \E or
1505 end of string
7e31b643 1506 \E end either case modification or quoted section
c543c01b
TC
1507 (whichever was last seen)
1508
736fe711
KW
1509See L<perlfunc/quotemeta> for the exact definition of characters that
1510are quoted by C<\Q>.
1511
628253b8 1512C<\L>, C<\U>, C<\F>, and C<\Q> can stack, in which case you need one
c543c01b
TC
1513C<\E> for each. For example:
1514
9fef6a0d
KW
1515 say"This \Qquoting \ubusiness \Uhere isn't quite\E done yet,\E is it?";
1516 This quoting\ Business\ HERE\ ISN\'T\ QUITE\ done\ yet\, is it?
a0d0e21e 1517
66cbab2c
KW
1518If C<use locale> is in effect (but not C<use locale ':not_characters'>),
1519the case map used by C<\l>, C<\L>,
c543c01b 1520C<\u>, and C<\U> is taken from the current locale. See L<perllocale>.
b6538e4f 1521If Unicode (for example, C<\N{}> or code points of 0x100 or
c543c01b
TC
1522beyond) is being used, the case map used by C<\l>, C<\L>, C<\u>, and
1523C<\U> is as defined by Unicode. That means that case-mapping
1524a single character can sometimes produce several characters.
628253b8 1525Under C<use locale>, C<\F> produces the same results as C<\L>.
a034a98d 1526
5a964f20
TC
1527All systems use the virtual C<"\n"> to represent a line terminator,
1528called a "newline". There is no such thing as an unvarying, physical
19799a22 1529newline character. It is only an illusion that the operating system,
5a964f20
TC
1530device drivers, C libraries, and Perl all conspire to preserve. Not all
1531systems read C<"\r"> as ASCII CR and C<"\n"> as ASCII LF. For example,
c543c01b
TC
1532on the ancient Macs (pre-MacOS X) of yesteryear, these used to be reversed,
1533and on systems without line terminator,
1534printing C<"\n"> might emit no actual data. In general, use C<"\n"> when
5a964f20
TC
1535you mean a "newline" for your system, but use the literal ASCII when you
1536need an exact character. For example, most networking protocols expect
2a380090 1537and prefer a CR+LF (C<"\015\012"> or C<"\cM\cJ">) for line terminators,
5a964f20
TC
1538and although they often accept just C<"\012">, they seldom tolerate just
1539C<"\015">. If you get in the habit of using C<"\n"> for networking,
1540you may be burned some day.
d74e8afc
ITB
1541X<newline> X<line terminator> X<eol> X<end of line>
1542X<\n> X<\r> X<\r\n>
5a964f20 1543
904501ec
MG
1544For constructs that do interpolate, variables beginning with "C<\$>"
1545or "C<@>" are interpolated. Subscripted variables such as C<\$a[3]> or
ad0f383a
A
1546C<< \$href->{key}[0] >> are also interpolated, as are array and hash slices.
1547But method calls such as C<< \$obj->meth >> are not.
af9219ee
MG
1548
1549Interpolating an array or slice interpolates the elements in order,
1550separated by the value of C<\$">, so is equivalent to interpolating
c543c01b
TC
1551C<join \$", @array>. "Punctuation" arrays such as C<@*> are usually
1552interpolated only if the name is enclosed in braces C<@{*}>, but the
1553arrays C<@_>, C<@+>, and C<@-> are interpolated even without braces.
af9219ee 1554
bc7b91c6
EB
1555For double-quoted strings, the quoting from C<\Q> is applied after
1556interpolation and escapes are processed.
1557
1558 "abc\Qfoo\tbar\$s\Exyz"
1559
1560is equivalent to
1561
1562 "abc" . quotemeta("foo\tbar\$s") . "xyz"
1563
1564For the pattern of regex operators (C<qr//>, C<m//> and C<s///>),
1565the quoting from C<\Q> is applied after interpolation is processed,
1566but before escapes are processed. This allows the pattern to match
1567literally (except for C<\$> and C<@>). For example, the following matches:
1568
1569 '\s\t' =~ /\Q\s\t/
1570
1571Because C<\$> or C<@> trigger interpolation, you'll need to use something
1572like C</\Quser\E\@\Qhost/> to match them literally.
1d2dff63 1573
a0d0e21e
LW
1574Patterns are subject to an additional level of interpretation as a
1575regular expression. This is done as a second pass, after variables are
1576interpolated, so that regular expressions may be incorporated into the
1577pattern from the variables. If this is not what you want, use C<\Q> to
1578interpolate a variable literally.
1579
19799a22
GS
1580Apart from the behavior described above, Perl does not expand
1581multiple levels of interpolation. In particular, contrary to the
1582expectations of shell programmers, back-quotes do I<NOT> interpolate
1583within double quotes, nor do single quotes impede evaluation of
1584variables when used within double quotes.
a0d0e21e 1585
5f05dabc 1586=head2 Regexp Quote-Like Operators
d74e8afc 1587X<operator, regexp>
cb1a09d0 1588
5f05dabc 1589Here are the quote-like operators that apply to pattern
cb1a09d0
AD
1590matching and related activities.
1591
a0d0e21e
LW
1592=over 8
1593
b6fa137b 1594=item qr/STRING/msixpodual
01c6f5f4 1595X<qr> X</i> X</m> X</o> X</s> X</x> X</p>
a0d0e21e 1596
87e95b7f
YO
1597This operator quotes (and possibly compiles) its I<STRING> as a regular
1598expression. I<STRING> is interpolated the same way as I<PATTERN>
1599in C<m/PATTERN/>. If "'" is used as the delimiter, no interpolation
1600is done. Returns a Perl value which may be used instead of the
f6050459 1601corresponding C</STRING/msixpodual> expression. The returned value is a
85dd5c8b 1602normalized version of the original pattern. It magically differs from
1c8ee595
CO
1603a string containing the same characters: C<ref(qr/x/)> returns "Regexp";
1604however, dereferencing it is not well defined (you currently get the
1605normalized version of the original pattern, but this may change).
1606
a0d0e21e 1607
87e95b7f
YO
1608For example,
1609
1610 \$rex = qr/my.STRING/is;
85dd5c8b 1611 print \$rex; # prints (?si-xm:my.STRING)
87e95b7f
YO
1612 s/\$rex/foo/;
1613
1614is equivalent to
1615
1616 s/my.STRING/foo/is;
1617
1618The result may be used as a subpattern in a match:
1619
1620 \$re = qr/\$pattern/;
1621 \$string =~ /foo\${re}bar/; # can be interpolated in other patterns
1622 \$string =~ \$re; # or used standalone
1623 \$string =~ /\$re/; # or this way
1624
f6050459 1625Since Perl may compile the pattern at the moment of execution of the qr()
87e95b7f
YO
1626operator, using qr() may have speed advantages in some situations,
1627notably if the result of qr() is used standalone:
1628
1629 sub match {
1630 my \$patterns = shift;
1631 my @compiled = map qr/\$_/i, @\$patterns;
1632 grep {
1633 my \$success = 0;
1634 foreach my \$pat (@compiled) {
1635 \$success = 1, last if /\$pat/;
1636 }
1637 \$success;
1638 } @_;
5a964f20
TC
1639 }
1640
87e95b7f
YO
1641Precompilation of the pattern into an internal representation at
1642the moment of qr() avoids a need to recompile the pattern every
1643time a match C</\$pat/> is attempted. (Perl has many other internal
1644optimizations, but none would be triggered in the above example if
1645we did not use qr() operator.)
1646
765fa144 1647Options (specified by the following modifiers) are:
87e95b7f
YO
1648
1649 m Treat string as multiple lines.
1650 s Treat string as single line. (Make . match a newline)
1651 i Do case-insensitive pattern matching.
1652 x Use extended regular expressions.
1653 p When matching preserve a copy of the matched string so
1654 that \${^PREMATCH}, \${^MATCH}, \${^POSTMATCH} will be defined.
1655 o Compile pattern only once.
b5c53fdc
KW
1656 a ASCII-restrict: Use ASCII for \d, \s, \w; specifying two a's
1657 further restricts /i matching so that no ASCII character will
1658 match a non-ASCII one
18509dec
KW
1659 l Use the locale
1660 u Use Unicode rules
b6fa137b 1661 d Use Unicode or native charset, as in 5.12 and earlier
87e95b7f
YO
1662
1663If a precompiled pattern is embedded in a larger pattern then the effect
c543c01b 1664of "msixpluad" will be propagated appropriately. The effect the "o"
87e95b7f
YO
1665modifier has is not propagated, being restricted to those patterns
1666explicitly using it.
1667
b6fa137b 1668The last four modifiers listed above, added in Perl 5.14,
18509dec
KW
1669control the character set semantics, but C</a> is the only one you are likely
1670to want to specify explicitly; the other three are selected
1671automatically by various pragmas.
da392a17 1672
87e95b7f 1673See L<perlre> for additional information on valid syntax for STRING, and
5e2aa8f5 1674for a detailed look at the semantics of regular expressions. In
1ca345ed
TC
1675particular, all modifiers except the largely obsolete C</o> are further
1676explained in L<perlre/Modifiers>. C</o> is described in the next section.
a0d0e21e 1677
b6fa137b 1678=item m/PATTERN/msixpodualgc
89d205f2
YO
1679X<m> X<operator, match>
1680X<regexp, options> X<regexp> X<regex, options> X<regex>
01c6f5f4 1681X</m> X</s> X</i> X</x> X</p> X</o> X</g> X</c>
a0d0e21e 1682
b6fa137b 1683=item /PATTERN/msixpodualgc
a0d0e21e 1684
5a964f20 1685Searches a string for a pattern match, and in scalar context returns
19799a22
GS
1686true if it succeeds, false if it fails. If no string is specified
1687via the C<=~> or C<!~> operator, the \$_ string is searched. (The
1688string specified with C<=~> need not be an lvalue--it may be the
1689result of an expression evaluation, but remember the C<=~> binds
006671a6 1690rather tightly.) See also L<perlre>.
a0d0e21e 1691
f6050459 1692Options are as described in C<qr//> above; in addition, the following match
01c6f5f4 1693process modifiers are available:
a0d0e21e 1694
950b09ed
KW
1695 g Match globally, i.e., find all occurrences.
1696 c Do not reset search position on a failed match when /g is in effect.
a0d0e21e 1697
725a61d7 1698If "/" is the delimiter then the initial C<m> is optional. With the C<m>
c543c01b 1699you can use any pair of non-whitespace (ASCII) characters
725a61d7
Z
1700as delimiters. This is particularly useful for matching path names
1701that contain "/", to avoid LTS (leaning toothpick syndrome). If "?" is
1702the delimiter, then a match-only-once rule applies,
1703described in C<m?PATTERN?> below.
19799a22 1704If "'" is the delimiter, no interpolation is performed on the PATTERN.
ed02a3bf
DN
1705When using a character valid in an identifier, whitespace is required
1706after the C<m>.
a0d0e21e 1707
532c9e80
KW
1708PATTERN may contain variables, which will be interpolated
1709every time the pattern search is evaluated, except
1f247705
GS
1710for when the delimiter is a single quote. (Note that C<\$(>, C<\$)>, and
1711C<\$|> are not interpolated because they look like end-of-string tests.)
532c9e80
KW
1712Perl will not recompile the pattern unless an interpolated
1713variable that it contains changes. You can force Perl to skip the
1714test and never recompile by adding a C</o> (which stands for "once")
1715after the trailing delimiter.
1716Once upon a time, Perl would recompile regular expressions
1717unnecessarily, and this modifier was useful to tell it not to do so, in the
1718interests of speed. But now, the only reasons to use C</o> are either:
1719
1720=over
1721
1722=item 1
1723
1724The variables are thousands of characters long and you know that they
1725don't change, and you need to wring out the last little bit of speed by
1726having Perl skip testing for that. (There is a maintenance penalty for
1727doing this, as mentioning C</o> constitutes a promise that you won't
18509dec 1728change the variables in the pattern. If you do change them, Perl won't
532c9e80
KW
1729even notice.)
1730
1731=item 2
1732
1733you want the pattern to use the initial values of the variables
1734regardless of whether they change or not. (But there are saner ways
1735of accomplishing this than using C</o>.)
1736
fa9b8686
DM
1737=item 3
1738
1739If the pattern contains embedded code, such as
1740
1741 use re 'eval';
1742 \$code = 'foo(?{ \$x })';
1743 /\$code/
1744
1745then perl will recompile each time, even though the pattern string hasn't
1746changed, to ensure that the current value of C<\$x> is seen each time.
1747Use C</o> if you want to avoid this.
1748
532c9e80 1749=back
a0d0e21e 1750
18509dec
KW
1751The bottom line is that using C</o> is almost never a good idea.
1752
e9d89077
DN
1753=item The empty pattern //
1754
5a964f20 1755If the PATTERN evaluates to the empty string, the last
d65afb4b 1756I<successfully> matched regular expression is used instead. In this
c543c01b 1757case, only the C<g> and C<c> flags on the empty pattern are honored;
d65afb4b
HS
1758the other flags are taken from the original pattern. If no match has
1759previously succeeded, this will (silently) act instead as a genuine
1760empty pattern (which will always match).
a0d0e21e 1761
89d205f2
YO
1762Note that it's possible to confuse Perl into thinking C<//> (the empty
1763regex) is really C<//> (the defined-or operator). Perl is usually pretty
1764good about this, but some pathological cases might trigger this, such as
1765C<\$a///> (is that C<(\$a) / (//)> or C<\$a // />?) and C<print \$fh //>
1766(C<print \$fh(//> or C<print(\$fh //>?). In all of these examples, Perl
1767will assume you meant defined-or. If you meant the empty regex, just
1768use parentheses or spaces to disambiguate, or even prefix the empty
c963b151
BD
1769regex with an C<m> (so C<//> becomes C<m//>).
1770
e9d89077
DN
1771=item Matching in list context
1772
19799a22 1773If the C</g> option is not used, C<m//> in list context returns a
a0d0e21e 1774list consisting of the subexpressions matched by the parentheses in the
1ca345ed 1775pattern, that is, (C<\$1>, C<\$2>, C<\$3>...). (Note that here C<\$1> etc. are
f7e33566
GS
1776also set, and that this differs from Perl 4's behavior.) When there are
1777no parentheses in the pattern, the return value is the list C<(1)> for
1778success. With or without parentheses, an empty list is returned upon
1779failure.
a0d0e21e
LW
1780
1781Examples:
1782
58bafa60 1783 open(TTY, "+</dev/tty")
c543c01b
TC
1784 || die "can't access /dev/tty: \$!";
1785
a0d0e21e
LW
1786 <TTY> =~ /^y/i && foo(); # do foo if desired
1787
1788 if (/Version: *([0-9.]*)/) { \$version = \$1; }
1789
1790 next if m#^/usr/spool/uucp#;
1791
1792 # poor man's grep
1793 \$arg = shift;
1794 while (<>) {
c543c01b 1795 print if /\$arg/o; # compile only once (no longer needed!)
a0d0e21e
LW
1796 }
1797
1798 if ((\$F1, \$F2, \$Etc) = (\$foo =~ /^(\S+)\s+(\S+)\s*(.*)/))
1799
1800This last example splits \$foo into the first two words and the
5f05dabc 1801remainder of the line, and assigns those three fields to \$F1, \$F2, and
c543c01b
TC
1802\$Etc. The conditional is true if any variables were assigned; that is,
1803if the pattern matched.
a0d0e21e 1804
19799a22 1805The C</g> modifier specifies global pattern matching--that is,
3dd93342 1806matching as many times as possible within the string. How it behaves
1807depends on the context. In list context, it returns a list of the
19799a22 1808substrings matched by any capturing parentheses in the regular
3dd93342 1809expression. If there are no parentheses, it returns a list of all
19799a22
GS
1810the matched strings, as if there were parentheses around the whole
1811pattern.
a0d0e21e 1812
7e86de3e 1813In scalar context, each execution of C<m//g> finds the next match,
19799a22 1814returning true if it matches, and false if there is no further match.
3dd93342 1815The position after the last match can be read or set using the C<pos()>
1816function; see L<perlfunc/pos>. A failed match normally resets the
7e86de3e 1817search position to the beginning of the string, but you can avoid that
1ca345ed 1818by adding the C</c> modifier (for example, C<m//gc>). Modifying the target
7e86de3e 1819string also resets the search position.
c90c0ff4 1820
e9d89077
DN
1821=item \G assertion
1822
c90c0ff4 1823You can intermix C<m//g> matches with C<m/\G.../g>, where C<\G> is a
3dd93342 1824zero-width assertion that matches the exact position where the
1825previous C<m//g>, if any, left off. Without the C</g> modifier, the
1826C<\G> assertion still anchors at C<pos()> as it was at the start of
1827the operation (see L<perlfunc/pos>), but the match is of course only
1828attempted once. Using C<\G> without C</g> on a target string that has
1829not previously had a C</g> match applied to it is the same as using
1830the C<\A> assertion to match the beginning of the string. Note also
1831that, currently, C<\G> is only properly supported when anchored at the
1832very beginning of the pattern.
c90c0ff4 1833
1834Examples:
a0d0e21e
LW
1835
1836 # list context
1837 (\$one,\$five,\$fifteen) = (`uptime` =~ /(\d+\.\d+)/g);
1838
1839 # scalar context
c543c01b
TC
1840 local \$/ = "";
1841 while (\$paragraph = <>) {
1842 while (\$paragraph =~ /\p{Ll}['")]*[.!?]+['")]*\s/g) {
19799a22 1843 \$sentences++;
a0d0e21e
LW
1844 }
1845 }
c543c01b
TC
1846 say \$sentences;
1847
1848Here's another way to check for sentences in a paragraph:
1849
1850 my \$sentence_rx = qr{
1851 (?: (?<= ^ ) | (?<= \s ) ) # after start-of-string or whitespace
1852 \p{Lu} # capital letter
1853 .*? # a bunch of anything
1854 (?<= \S ) # that ends in non-whitespace
1855 (?<! \b [DMS]r ) # but isn't a common abbreviation
1856 (?<! \b Mrs )
1857 (?<! \b Sra )
1858 (?<! \b St )
1859 [.?!] # followed by a sentence ender
1860 (?= \$ | \s ) # in front of end-of-string or whitespace
1861 }sx;
1862 local \$/ = "";
1863 while (my \$paragraph = <>) {
1864 say "NEW PARAGRAPH";
1865 my \$count = 0;
1866 while (\$paragraph =~ /(\$sentence_rx)/g) {
1867 printf "\tgot sentence %d: <%s>\n", ++\$count, \$1;
1868 }
1869 }
1870
1871Here's how to use C<m//gc> with C<\G>:
a0d0e21e 1872
137443ea 1873 \$_ = "ppooqppqq";
44a8e56a 1874 while (\$i++ < 2) {
1875 print "1: '";
c90c0ff4 1876 print \$1 while /(o)/gc; print "', pos=", pos, "\n";
44a8e56a 1877 print "2: '";
c90c0ff4 1878 print \$1 if /\G(q)/gc; print "', pos=", pos, "\n";
44a8e56a 1879 print "3: '";
c90c0ff4 1880 print \$1 while /(p)/gc; print "', pos=", pos, "\n";
44a8e56a 1881 }
5d43e42d 1882 print "Final: '\$1', pos=",pos,"\n" if /\G(.)/;
44a8e56a 1883
1884The last example should print:
1885
1886 1: 'oo', pos=4
137443ea 1887 2: 'q', pos=5
44a8e56a 1888 3: 'pp', pos=7
1889 1: '', pos=7
137443ea 1890 2: 'q', pos=8
1891 3: '', pos=8
5d43e42d
DC
1892 Final: 'q', pos=8
1893
1894Notice that the final match matched C<q> instead of C<p>, which a match
1895without the C<\G> anchor would have done. Also note that the final match
ac036724 1896did not update C<pos>. C<pos> is only updated on a C</g> match. If the
c543c01b
TC
1897final match did indeed match C<p>, it's a good bet that you're running a
1898very old (pre-5.6.0) version of Perl.
44a8e56a 1899
c90c0ff4 1900A useful idiom for C<lex>-like scanners is C</\G.../gc>. You can
e7ea3e70 1901combine several regexps like this to process a string part-by-part,
c90c0ff4 1902doing different actions depending on which regexp matched. Each
1903regexp tries to match where the previous one leaves off.
e7ea3e70 1904
3fe9a6f1 1905 \$_ = <<'EOL';
950b09ed 1906 \$url = URI::URL->new( "http://example.com/" ); die if \$url eq "xXx";
3fe9a6f1 1907 EOL
c543c01b
TC
1908
1909 LOOP: {
950b09ed 1910 print(" digits"), redo LOOP if /\G\d+\b[,.;]?\s*/gc;
c543c01b
TC
1911 print(" lowercase"), redo LOOP if /\G\p{Ll}+\b[,.;]?\s*/gc;
1912 print(" UPPERCASE"), redo LOOP if /\G\p{Lu}+\b[,.;]?\s*/gc;
1913 print(" Capitalized"), redo LOOP if /\G\p{Lu}\p{Ll}+\b[,.;]?\s*/gc;
1914 print(" MiXeD"), redo LOOP if /\G\pL+\b[,.;]?\s*/gc;
1915 print(" alphanumeric"), redo LOOP if /\G[\p{Alpha}\pN]+\b[,.;]?\s*/gc;
1916 print(" line-noise"), redo LOOP if /\G\W+/gc;
950b09ed 1917 print ". That's all!\n";
c543c01b 1918 }
e7ea3e70
IZ
1919
1920Here is the output (split into several lines):
1921
c543c01b
TC
1922 line-noise lowercase line-noise UPPERCASE line-noise UPPERCASE
1923 line-noise lowercase line-noise lowercase line-noise lowercase
1924 lowercase line-noise lowercase lowercase line-noise lowercase
1925 lowercase line-noise MiXeD line-noise. That's all!
44a8e56a 1926
c543c01b 1927=item m?PATTERN?msixpodualgc
725a61d7 1928X<?> X<operator, match-once>
87e95b7f 1929
c543c01b 1930=item ?PATTERN?msixpodualgc
55d389e7 1931
725a61d7
Z
1932This is just like the C<m/PATTERN/> search, except that it matches
1933only once between calls to the reset() operator. This is a useful
87e95b7f 1934optimization when you want to see only the first occurrence of
ceb131e8 1935something in each file of a set of files, for instance. Only C<m??>
87e95b7f
YO
1936patterns local to the current package are reset.
1937
1938 while (<>) {
ceb131e8 1939 if (m?^\$?) {
87e95b7f
YO
1940 # blank line between header and body
1941 }
1942 } continue {
725a61d7 1943 reset if eof; # clear m?? status for next file
87e95b7f
YO
1944 }
1945
c543c01b
TC
1946Another example switched the first "latin1" encoding it finds
1947to "utf8" in a pod file:
1948
1949 s//utf8/ if m? ^ =encoding \h+ \K latin1 ?x;
1950
1951The match-once behavior is controlled by the match delimiter being
725a61d7
Z
1952C<?>; with any other delimiter this is the normal C<m//> operator.
1953
1954For historical reasons, the leading C<m> in C<m?PATTERN?> is optional,
1955but the resulting C<?PATTERN?> syntax is deprecated, will warn on
c543c01b
TC
1956usage and might be removed from a future stable release of Perl (without
1957further notice!).
87e95b7f 1958
b6fa137b 1959=item s/PATTERN/REPLACEMENT/msixpodualgcer
87e95b7f 1960X<substitute> X<substitution> X<replace> X<regexp, replace>
4f4d7508 1961X<regexp, substitute> X</m> X</s> X</i> X</x> X</p> X</o> X</g> X</c> X</e> X</r>
87e95b7f
YO
1962
1963Searches a string for a pattern, and if found, replaces that pattern
1964with the replacement text and returns the number of substitutions
1965made. Otherwise it returns false (specifically, the empty string).
1966
c543c01b 1967If the C</r> (non-destructive) option is used then it runs the
679563bb
KW
1968substitution on a copy of the string and instead of returning the
1969number of substitutions, it returns the copy whether or not a
c543c01b
TC
1970substitution occurred. The original string is never changed when
1971C</r> is used. The copy will always be a plain string, even if the
1972input is an object or a tied variable.
4f4d7508 1973
87e95b7f 1974If no string is specified via the C<=~> or C<!~> operator, the C<\$_>
c543c01b
TC
1975variable is searched and modified. Unless the C</r> option is used,
1976the string specified must be a scalar variable, an array element, a
1977hash element, or an assignment to one of those; that is, some sort of
1978scalar lvalue.
87e95b7f
YO
1979
1980If the delimiter chosen is a single quote, no interpolation is
1981done on either the PATTERN or the REPLACEMENT. Otherwise, if the
1982PATTERN contains a \$ that looks like a variable rather than an
1983end-of-string test, the variable will be interpolated into the pattern
1984at run-time. If you want the pattern compiled only once the first time
1985the variable is interpolated, use the C</o> option. If the pattern
1986evaluates to the empty string, the last successfully executed regular
1987expression is used instead. See L<perlre> for further explanation on these.
87e95b7f
YO
1988
1989Options are as with m// with the addition of the following replacement
1990specific options:
1991
1992 e Evaluate the right side as an expression.
4f4d7508
DC
1993 ee Evaluate the right side as a string then eval the result.
1994 r Return substitution and leave the original string untouched.
87e95b7f 1995
ed02a3bf
DN
1996Any non-whitespace delimiter may replace the slashes. Add space after
1997the C<s> when using a character allowed in identifiers. If single quotes
1998are used, no interpretation is done on the replacement string (the C</e>
1999modifier overrides this, however). Unlike Perl 4, Perl 5 treats backticks
2000as normal delimiters; the replacement text is not evaluated as a command.
2001If the PATTERN is delimited by bracketing quotes, the REPLACEMENT has
1ca345ed 2002its own pair of quotes, which may or may not be bracketing quotes, for example,
87e95b7f
YO
2003C<s(foo)(bar)> or C<< s<foo>/bar/ >>. A C</e> will cause the
2004replacement portion to be treated as a full-fledged Perl expression
2005and evaluated right then and there. It is, however, syntax checked at
2006compile-time. A second C<e> modifier will cause the replacement portion
2007to be C<eval>ed before being run as a Perl expression.
2008
2009Examples:
2010
2011 s/\bgreen\b/mauve/g; # don't change wintergreen
2012
2013 \$path =~ s|/usr/bin|/usr/local/bin|;
2014
2015 s/Login: \$foo/Login: \$bar/; # run-time pattern
2016
2017 (\$foo = \$bar) =~ s/this/that/; # copy first, then change
4f4d7508
DC
2018 (\$foo = "\$bar") =~ s/this/that/; # convert to string, copy, then change
2019 \$foo = \$bar =~ s/this/that/r; # Same as above using /r
2020 \$foo = \$bar =~ s/this/that/r
2021 =~ s/that/the other/r; # Chained substitutes using /r
2022 @foo = map { s/this/that/r } @bar # /r is very useful in maps
87e95b7f
YO
2023
2024 \$count = (\$paragraph =~ s/Mister\b/Mr./g); # get change-count
2025
2026 \$_ = 'abc123xyz';
2027 s/\d+/\$&*2/e; # yields 'abc246xyz'
2028 s/\d+/sprintf("%5d",\$&)/e; # yields 'abc 246xyz'
2029 s/\w/\$& x 2/eg; # yields 'aabbcc 224466xxyyzz'
2030
2031 s/%(.)/\$percent{\$1}/g; # change percent escapes; no /e
2032 s/%(.)/\$percent{\$1} || \$&/ge; # expr now, so /e
2033 s/^=(\w+)/pod(\$1)/ge; # use function call
2034
4f4d7508
DC
2035 \$_ = 'abc123xyz';
2036 \$a = s/abc/def/r; # \$a is 'def123xyz' and
2037 # \$_ remains 'abc123xyz'.
2038
87e95b7f
YO
2039 # expand variables in \$_, but dynamics only, using
2040 # symbolic dereferencing
2041 s/\\$(\w+)/\${\$1}/g;
2042
2043 # Add one to the value of any numbers in the string
2044 s/(\d+)/1 + \$1/eg;
2045
c543c01b
TC
2046 # Titlecase words in the last 30 characters only
2047 substr(\$str, -30) =~ s/\b(\p{Alpha}+)\b/\u\L\$1/g;
2048
87e95b7f
YO
2049 # This will expand any embedded scalar variable
2050 # (including lexicals) in \$_ : First \$1 is interpolated
2051 # to the variable name, and then evaluated
2052 s/(\\$\w+)/\$1/eeg;
2053
2054 # Delete (most) C comments.
2055 \$program =~ s {
2056 /\* # Match the opening delimiter.
2057 .*? # Match a minimal number of characters.
2058 \*/ # Match the closing delimiter.
2059 } []gsx;
2060
2061 s/^\s*(.*?)\s*\$/\$1/; # trim whitespace in \$_, expensively
2062
2063 for (\$variable) { # trim whitespace in \$variable, cheap
2064 s/^\s+//;
2065 s/\s+\$//;
2066 }
2067
2068 s/([^ ]*) *([^ ]*)/\$2 \$1/; # reverse 1st two fields
2069
2070Note the use of \$ instead of \ in the last example. Unlike
2071B<sed>, we use the \<I<digit>> form in only the left hand side.
2072Anywhere else it's \$<I<digit>>.
2073
2074Occasionally, you can't use just a C</g> to get all the changes
2075to occur that you might want. Here are two common cases:
2076
2077 # put commas in the right places in an integer
2078 1 while s/(\d)(\d\d\d)(?!\d)/\$1,\$2/g;
2079
2080 # expand tabs to 8-column spacing
2081 1 while s/\t+/' ' x (length(\$&)*8 - length(\$`)%8)/e;
2082
2083=back
2084
2085=head2 Quote-Like Operators
2086X<operator, quote-like>
2087
01c6f5f4
RGS
2088=over 4
2089
a0d0e21e 2090=item q/STRING/
5d44bfff 2091X<q> X<quote, single> X<'> X<''>
a0d0e21e 2092
5d44bfff 2093=item 'STRING'
a0d0e21e 2094
19799a22 2095A single-quoted, literal string. A backslash represents a backslash
68dc0745 2096unless followed by the delimiter or another backslash, in which case
2097the delimiter or backslash is interpolated.
a0d0e21e
LW
2098
2099 \$foo = q!I said, "You said, 'She said it.'"!;
2100 \$bar = q('This is it.');
68dc0745 2101 \$baz = '\n'; # a two-character string
a0d0e21e
LW
2102
2103=item qq/STRING/
d74e8afc 2104X<qq> X<quote, double> X<"> X<"">
a0d0e21e
LW
2105
2106=item "STRING"
2107
2108A double-quoted, interpolated string.
2109
2110 \$_ .= qq
2111 (*** The previous line contains the naughty word "\$1".\n)
19799a22 2112 if /\b(tcl|java|python)\b/i; # :-)
68dc0745 2113 \$baz = "\n"; # a one-character string
a0d0e21e
LW
2114
2115=item qx/STRING/
d74e8afc 2116X<qx> X<`> X<``> X<backtick>
a0d0e21e
LW
2117
2118=item `STRING`
2119
43dd4d21
JH
2120A string which is (possibly) interpolated and then executed as a
2121system command with C</bin/sh> or its equivalent. Shell wildcards,
2122pipes, and redirections will be honored. The collected standard
2123output of the command is returned; standard error is unaffected. In
2124scalar context, it comes back as a single (potentially multi-line)
2125string, or undef if the command failed. In list context, returns a
2126list of lines (however you've defined lines with \$/ or
2127\$INPUT_RECORD_SEPARATOR), or an empty list if the command failed.
5a964f20
TC
2128
2129Because backticks do not affect standard error, use shell file descriptor
2130syntax (assuming the shell supports this) if you care to address this.
2131To capture a command's STDERR and STDOUT together:
a0d0e21e 2132
5a964f20
TC
2133 \$output = `cmd 2>&1`;
2134
2135To capture a command's STDOUT but discard its STDERR:
2136
2137 \$output = `cmd 2>/dev/null`;
2138
2139To capture a command's STDERR but discard its STDOUT (ordering is
2140important here):
2141
2142 \$output = `cmd 2>&1 1>/dev/null`;
2143
2144To exchange a command's STDOUT and STDERR in order to capture the STDERR
2145but leave its STDOUT to come out the old STDERR:
2146
2147 \$output = `cmd 3>&1 1>&2 2>&3 3>&-`;
2148
2149To read both a command's STDOUT and its STDERR separately, it's easiest
2359510d
SD
2150to redirect them separately to files, and then read from those files
2151when the program is done:
5a964f20 2152
2359510d 2153 system("program args 1>program.stdout 2>program.stderr");
5a964f20 2154
30398227
SP
2155The STDIN filehandle used by the command is inherited from Perl's STDIN.
2156For example:
2157
c543c01b
TC
2158 open(SPLAT, "stuff") || die "can't open stuff: \$!";
2159 open(STDIN, "<&SPLAT") || die "can't dupe SPLAT: \$!";
40bbb707 2160 print STDOUT `sort`;
30398227 2161
40bbb707 2162will print the sorted contents of the file named F<"stuff">.
30398227 2163
5a964f20
TC
2164Using single-quote as a delimiter protects the command from Perl's
2165double-quote interpolation, passing it on to the shell instead:
2166
2167 \$perl_info = qx(ps \$\$); # that's Perl's \$\$
2168 \$shell_info = qx'ps \$\$'; # that's the new shell's \$\$
2169
19799a22 2170How that string gets evaluated is entirely subject to the command
5a964f20
TC
2171interpreter on your system. On most platforms, you will have to protect
2172shell metacharacters if you want them treated literally. This is in
2173practice difficult to do, as it's unclear how to escape which characters.
2174See L<perlsec> for a clean and safe example of a manual fork() and exec()
2175to emulate backticks safely.
a0d0e21e 2176
bb32b41a
GS
2177On some platforms (notably DOS-like ones), the shell may not be
2178capable of dealing with multiline commands, so putting newlines in
2179the string may not get you what you want. You may be able to evaluate
2180multiple commands in a single line by separating them with the command
1ca345ed
TC
2181separator character, if your shell supports that (for example, C<;> on
2182many Unix shells and C<&> on the Windows NT C<cmd> shell).
bb32b41a 2183
0f897271
GS
2184Beginning with v5.6.0, Perl will attempt to flush all files opened for
2185output before starting the child process, but this may not be supported
2186on some platforms (see L<perlport>). To be safe, you may need to set
2187C<\$|> (\$AUTOFLUSH in English) or call the C<autoflush()> method of
2188C<IO::Handle> on any open handles.
2189
bb32b41a
GS
2190Beware that some command shells may place restrictions on the length
2191of the command line. You must ensure your strings don't exceed this
2192limit after any necessary interpolations. See the platform-specific
2193release notes for more details about your particular environment.
2194
5a964f20
TC
2195Using this operator can lead to programs that are difficult to port,
2196because the shell commands called vary between systems, and may in
2197fact not be present at all. As one example, the C<type> command under
2198the POSIX shell is very different from the C<type> command under DOS.
2199That doesn't mean you should go out of your way to avoid backticks
2200when they're the right way to get something done. Perl was made to be
2201a glue language, and one of the things it glues together is commands.
2202Just understand what you're getting yourself into.
bb32b41a 2203
da87341d 2204See L</"I/O Operators"> for more discussion.
a0d0e21e 2205
945c54fd 2206=item qw/STRING/
d74e8afc 2207X<qw> X<quote, list> X<quote, words>
945c54fd
JH
2208
2209Evaluates to a list of the words extracted out of STRING, using embedded
2210whitespace as the word delimiters. It can be understood as being roughly
2211equivalent to:
2212
c543c01b 2213 split(" ", q/STRING/);
945c54fd 2214
efb1e162
CW
2215the differences being that it generates a real list at compile time, and
2216in scalar context it returns the last element in the list. So
945c54fd
JH
2217this expression:
2218
2219 qw(foo bar baz)
2220
2221is semantically equivalent to the list:
2222
c543c01b 2223 "foo", "bar", "baz"
945c54fd
JH
2224
2225Some frequently seen examples:
2226
2227 use POSIX qw( setlocale localeconv )
2228 @EXPORT = qw( foo bar baz );
2229
2230A common mistake is to try to separate the words with comma or to
2231put comments into a multi-line C<qw>-string. For this reason, the
89d205f2 2232C<use warnings> pragma and the B<-w> switch (that is, the C<\$^W> variable)
945c54fd
JH
2233produces warnings if the STRING contains the "," or the "#" character.
2234
8ff32507 2235=item tr/SEARCHLIST/REPLACEMENTLIST/cdsr
d74e8afc 2236X<tr> X<y> X<transliterate> X</c> X</d> X</s>
a0d0e21e 2237
8ff32507 2238=item y/SEARCHLIST/REPLACEMENTLIST/cdsr
a0d0e21e 2239
2c268ad5 2240Transliterates all occurrences of the characters found in the search list
a0d0e21e
LW
2241with the corresponding character in the replacement list. It returns
2242the number of characters replaced or deleted. If no string is
c543c01b
TC
2243specified via the C<=~> or C<!~> operator, the \$_ string is transliterated.
2244
2245If the C</r> (non-destructive) option is present, a new copy of the string
2246is made and its characters transliterated, and this copy is returned no
2247matter whether it was modified or not: the original string is always
2248left unchanged. The new copy is always a plain string, even if the input
2249string is an object or a tied variable.
8ada0baa 2250
c543c01b
TC
2251Unless the C</r> option is used, the string specified with C<=~> must be a
2252scalar variable, an array element, a hash element, or an assignment to one
2253of those; in other words, an lvalue.
8ff32507 2254
89d205f2 2255A character range may be specified with a hyphen, so C<tr/A-J/0-9/>
2c268ad5 2256does the same replacement as C<tr/ACEGIBDFHJ/0246813579/>.
54310121 2257For B<sed> devotees, C<y> is provided as a synonym for C<tr>. If the
2258SEARCHLIST is delimited by bracketing quotes, the REPLACEMENTLIST has
c543c01b
TC
2259its own pair of quotes, which may or may not be bracketing quotes;
2260for example, C<tr[aeiouy][yuoiea]> or C<tr(+\-*/)/ABCD/>.
2261
2262Note that C<tr> does B<not> do regular expression character classes such as
2263C<\d> or C<\pL>. The C<tr> operator is not equivalent to the tr(1)
2264utility. If you want to map strings between lower/upper cases, see
2265L<perlfunc/lc> and L<perlfunc/uc>, and in general consider using the C<s>
2266operator if you need regular expressions. The C<\U>, C<\u>, C<\L>, and
2267C<\l> string-interpolation escapes on the right side of a substitution
2268operator will perform correct case-mappings, but C<tr[a-z][A-Z]> will not
2269(except sometimes on legacy 7-bit data).
cc255d5f 2270
8ada0baa
JH
2271Note also that the whole range idea is rather unportable between
2272character sets--and even within character sets they may cause results
2273you probably didn't expect. A sound principle is to use only ranges
2274that begin from and end at either alphabets of equal case (a-e, A-E),
2275or digits (0-4). Anything else is unsafe. If in doubt, spell out the
2276character sets in full.
2277
a0d0e21e
LW
2278Options:
2279
2280 c Complement the SEARCHLIST.
2281 d Delete found but unreplaced characters.
2282 s Squash duplicate replaced characters.
8ff32507
FC
2283 r Return the modified string and leave the original string
2284 untouched.
a0d0e21e 2285
19799a22
GS
2286If the C</c> modifier is specified, the SEARCHLIST character set
2287is complemented. If the C</d> modifier is specified, any characters
2288specified by SEARCHLIST not found in REPLACEMENTLIST are deleted.
2289(Note that this is slightly more flexible than the behavior of some
2290B<tr> programs, which delete anything they find in the SEARCHLIST,
2291period.) If the C</s> modifier is specified, sequences of characters
2292that were transliterated to the same character are squashed down
2293to a single instance of the character.
a0d0e21e
LW
2294
2295If the C</d> modifier is used, the REPLACEMENTLIST is always interpreted
2296exactly as specified. Otherwise, if the REPLACEMENTLIST is shorter
2297than the SEARCHLIST, the final character is replicated till it is long
5a964f20 2298enough. If the REPLACEMENTLIST is empty, the SEARCHLIST is replicated.
a0d0e21e
LW
2299This latter is useful for counting characters in a class or for
2300squashing character sequences in a class.
2301
2302Examples:
2303
c543c01b 2304 \$ARGV[1] =~ tr/A-Z/a-z/; # canonicalize to lower case ASCII
a0d0e21e
LW
2305
2306 \$cnt = tr/*/*/; # count the stars in \$_
2307
2308 \$cnt = \$sky =~ tr/*/*/; # count the stars in \$sky
2309
2310 \$cnt = tr/0-9//; # count the digits in \$_
2311
2312 tr/a-zA-Z//s; # bookkeeper -> bokeper
2313
2314 (\$HOST = \$host) =~ tr/a-z/A-Z/;
c543c01b 2315 \$HOST = \$host =~ tr/a-z/A-Z/r; # same thing
8ff32507 2316
c543c01b 2317 \$HOST = \$host =~ tr/a-z/A-Z/r # chained with s///r
8ff32507 2318 =~ s/:/ -p/r;
a0d0e21e
LW
2319
2320 tr/a-zA-Z/ /cs; # change non-alphas to single space
2321
8ff32507
FC
2322 @stripped = map tr/a-zA-Z/ /csr, @original;
2323 # /r with map
2324
a0d0e21e 2325 tr [\200-\377]
c543c01b 2326 [\000-\177]; # wickedly delete 8th bit
a0d0e21e 2327
19799a22
GS
2328If multiple transliterations are given for a character, only the
2329first one is used:
748a9306
LW
2330
2331 tr/AAA/XYZ/
2332
2c268ad5 2333will transliterate any A to X.
748a9306 2334
19799a22 2335Because the transliteration table is built at compile time, neither
a0d0e21e 2336the SEARCHLIST nor the REPLACEMENTLIST are subjected to double quote
19799a22
GS
2337interpolation. That means that if you want to use variables, you
2338must use an eval():
a0d0e21e
LW
2339
2340 eval "tr/\$oldlist/\$newlist/";
2341 die \$@ if \$@;
2342
2343 eval "tr/\$oldlist/\$newlist/, 1" or die \$@;
2344
7e3b091d 2345=item <<EOF
d74e8afc 2346X<here-doc> X<heredoc> X<here-document> X<<< << >>>
7e3b091d
DA
2347
2348A line-oriented form of quoting is based on the shell "here-document"
2349syntax. Following a C<< << >> you specify a string to terminate
2350the quoted material, and all lines following the current line down to
89d205f2
YO
2351the terminating string are the value of the item.
2352
2353The terminating string may be either an identifier (a word), or some
2354quoted text. An unquoted identifier works like double quotes.
2355There may not be a space between the C<< << >> and the identifier,
2356unless the identifier is explicitly quoted. (If you put a space it
2357will be treated as a null identifier, which is valid, and matches the
2358first empty line.) The terminating string must appear by itself
2359(unquoted and with no surrounding whitespace) on the terminating line.
2360
2361If the terminating string is quoted, the type of quotes used determine
2362the treatment of the text.
2363
2364=over 4
2365
2366=item Double Quotes
2367
2368Double quotes indicate that the text will be interpolated using exactly
2369the same rules as normal double quoted strings.
7e3b091d
DA
2370
2371 print <<EOF;
2372 The price is \$Price.
2373 EOF
2374
2375 print << "EOF"; # same as above
2376 The price is \$Price.
2377 EOF
2378
89d205f2
YO
2379
2380=item Single Quotes
2381
2382Single quotes indicate the text is to be treated literally with no
2383interpolation of its content. This is similar to single quoted
2384strings except that backslashes have no special meaning, with C<\\>
2385being treated as two backslashes and not one as they would in every
2386other quoting construct.
2387
c543c01b
TC
2388Just as in the shell, a backslashed bareword following the C<<< << >>>
2389means the same thing as a single-quoted string does:
2390
2391 \$cost = <<'VISTA'; # hasta la ...
2392 That'll be \$10 please, ma'am.
2393 VISTA
2394
2395 \$cost = <<\VISTA; # Same thing!
2396 That'll be \$10 please, ma'am.
2397 VISTA
2398
89d205f2
YO
2399This is the only form of quoting in perl where there is no need
2400to worry about escaping content, something that code generators
2401can and do make good use of.
2402
2403=item Backticks
2404
2405The content of the here doc is treated just as it would be if the
2406string were embedded in backticks. Thus the content is interpolated
2407as though it were double quoted and then executed via the shell, with
2408the results of the execution returned.
2409
2410 print << `EOC`; # execute command and get results
7e3b091d 2411 echo hi there
7e3b091d
DA
2412 EOC
2413
89d205f2
YO
2414=back
2415
2416It is possible to stack multiple here-docs in a row:
2417
7e3b091d
DA
2418 print <<"foo", <<"bar"; # you can stack them
2419 I said foo.
2420 foo
2421 I said bar.
2422 bar
2423
2424 myfunc(<< "THIS", 23, <<'THAT');
2425 Here's a line
2426 or two.
2427 THIS
2428 and here's another.
2429 THAT
2430
2431Just don't forget that you have to put a semicolon on the end
2432to finish the statement, as Perl doesn't know you're not going to
2433try to do this:
2434
2435 print <<ABC
2436 179231
2437 ABC
2438 + 20;
2439
872d7e53
TS
2440If you want to remove the line terminator from your here-docs,
2441use C<chomp()>.
2442
2443 chomp(\$string = <<'END');
2444 This is a string.
2445 END
2446
2447If you want your here-docs to be indented with the rest of the code,
2448you'll need to remove leading whitespace from each line manually:
7e3b091d
DA
2449
2450 (\$quote = <<'FINIS') =~ s/^\s+//gm;
89d205f2 2451 The Road goes ever on and on,
7e3b091d
DA
2452 down from the door where it began.
2453 FINIS
2454
2455If you use a here-doc within a delimited construct, such as in C<s///eg>,
2456the quoted material must come on the lines following the final delimiter.
2457So instead of
2458
2459 s/this/<<E . 'that'
2460 the other
2461 E
2462 . 'more '/eg;
2463
2464you have to write
2465
89d205f2
YO
2466 s/this/<<E . 'that'
2467 . 'more '/eg;
2468 the other
2469 E
7e3b091d
DA
2470
2471If the terminating identifier is on the last line of the program, you
2472must be sure there is a newline after it; otherwise, Perl will give the
2473warning B<Can't find string terminator "END" anywhere before EOF...>.
2474
c543c01b
TC
2475Additionally, quoting rules for the end-of-string identifier are
2476unrelated to Perl's quoting rules. C<q()>, C<qq()>, and the like are not
89d205f2
YO
2477supported in place of C<''> and C<"">, and the only interpolation is for
2478backslashing the quoting character:
7e3b091d
DA
2479
2480 print << "abc\"def";
2481 testing...
2482 abc"def
2483
2484Finally, quoted strings cannot span multiple lines. The general rule is
2485that the identifier must be a string literal. Stick with that, and you
2486should be safe.
2487
a0d0e21e
LW
2488=back
2489
75e14d17 2490=head2 Gory details of parsing quoted constructs
d74e8afc 2491X<quote, gory details>
75e14d17 2492
19799a22
GS
2493When presented with something that might have several different
2494interpretations, Perl uses the B<DWIM> (that's "Do What I Mean")
2495principle to pick the most probable interpretation. This strategy
2496is so successful that Perl programmers often do not suspect the
2497ambivalence of what they write. But from time to time, Perl's
2498notions differ substantially from what the author honestly meant.
2499
2500This section hopes to clarify how Perl handles quoted constructs.
2501Although the most common reason to learn this is to unravel labyrinthine
2502regular expressions, because the initial steps of parsing are the
2503same for all quoting operators, they are all discussed together.
2504
2505The most important Perl parsing rule is the first one discussed
2506below: when processing a quoted construct, Perl first finds the end
2507of that construct, then interprets its contents. If you understand
2508this rule, you may skip the rest of this section on the first
2509reading. The other rules are likely to contradict the user's
2510expectations much less frequently than this first one.
2511
2512Some passes discussed below are performed concurrently, but because
2513their results are the same, we consider them individually. For different
2514quoting constructs, Perl performs different numbers of passes, from
6deea57f 2515one to four, but these passes are always performed in the same order.
75e14d17 2516
13a2d996 2517=over 4
75e14d17
IZ
2518
2519=item Finding the end
2520
6deea57f
TS
2521The first pass is finding the end of the quoted construct, where
2522the information about the delimiters is used in parsing.
2523During this search, text between the starting and ending delimiters
2524is copied to a safe location. The text copied gets delimiter-independent.
2525
2526If the construct is a here-doc, the ending delimiter is a line
2527that has a terminating string as the content. Therefore C<<<EOF> is
2528terminated by C<EOF> immediately followed by C<"\n"> and starting
2529from the first column of the terminating line.
2530When searching for the terminating line of a here-doc, nothing
2531is skipped. In other words, lines after the here-doc syntax
2532are compared with the terminating string line by line.
2533
2534For the constructs except here-docs, single characters are used as starting
2535and ending delimiters. If the starting delimiter is an opening punctuation
2536(that is C<(>, C<[>, C<{>, or C<< < >>), the ending delimiter is the
2537corresponding closing punctuation (that is C<)>, C<]>, C<}>, or C<< > >>).
2538If the starting delimiter is an unpaired character like C</> or a closing
2539punctuation, the ending delimiter is same as the starting delimiter.
2540Therefore a C</> terminates a C<qq//> construct, while a C<]> terminates
2541C<qq[]> and C<qq]]> constructs.
2542
2543When searching for single-character delimiters, escaped delimiters
1ca345ed 2544and C<\\> are skipped. For example, while searching for terminating C</>,
6deea57f
TS
2545combinations of C<\\> and C<\/> are skipped. If the delimiters are
2546bracketing, nested pairs are also skipped. For example, while searching
2547for closing C<]> paired with the opening C<[>, combinations of C<\\>, C<\]>,
2548and C<\[> are all skipped, and nested C<[> and C<]> are skipped as well.
2549However, when backslashes are used as the delimiters (like C<qq\\> and
2550C<tr\\\>), nothing is skipped.
32581033
A
2551During the search for the end, backslashes that escape delimiters or
2552backslashes are removed (exactly speaking, they are not copied to the
2553safe location).
75e14d17 2554
19799a22
GS
2555For constructs with three-part delimiters (C<s///>, C<y///>, and
2556C<tr///>), the search is repeated once more.
6deea57f
TS
2557If the first delimiter is not an opening punctuation, three delimiters must
2558be same such as C<s!!!> and C<tr)))>, in which case the second delimiter
2559terminates the left part and starts the right part at once.
b6538e4f 2560If the left part is delimited by bracketing punctuation (that is C<()>,
6deea57f 2561C<[]>, C<{}>, or C<< <> >>), the right part needs another pair of
b6538e4f 2562delimiters such as C<s(){}> and C<tr[]//>. In these cases, whitespace
6deea57f 2563and comments are allowed between both parts, though the comment must follow
b6538e4f
TC
2564at least one whitespace character; otherwise a character expected as the
2565start of the comment may be regarded as the starting delimiter of the right part.
75e14d17 2566
19799a22
GS
2567During this search no attention is paid to the semantics of the construct.
2568Thus:
75e14d17
IZ
2569
2570 "\$hash{"\$foo/\$bar"}"
2571
2a94b7ce 2572or:
75e14d17 2573
89d205f2 2574 m/
2a94b7ce 2575 bar # NOT a comment, this slash / terminated m//!
75e14d17
IZ
2576 /x
2577
19799a22
GS
2578do not form legal quoted expressions. The quoted part ends on the
2579first C<"> and C</>, and the rest happens to be a syntax error.
2580Because the slash that terminated C<m//> was followed by a C<SPACE>,
2581the example above is not C<m//x>, but rather C<m//> with no C</x>
2582modifier. So the embedded C<#> is interpreted as a literal C<#>.
75e14d17 2583
89d205f2
YO
2584Also no attention is paid to C<\c\> (multichar control char syntax) during
2585this search. Thus the second C<\> in C<qq/\c\/> is interpreted as a part
2586of C<\/>, and the following C</> is not recognized as a delimiter.
0d594e51
TS
2587Instead, use C<\034> or C<\x1c> at the end of quoted constructs.
2588
75e14d17 2589=item Interpolation
d74e8afc 2590X<interpolation>
75e14d17 2591
19799a22 2592The next step is interpolation in the text obtained, which is now
89d205f2 2593delimiter-independent. There are multiple cases.
75e14d17 2594
13a2d996 2595=over 4
75e14d17 2596
89d205f2 2597=item C<<<'EOF'>
75e14d17
IZ
2598
2599No interpolation is performed.
6deea57f
TS
2600Note that the combination C<\\> is left intact, since escaped delimiters
2601are not available for here-docs.
75e14d17 2602
6deea57f 2603=item C<m''>, the pattern of C<s'''>
89d205f2 2604
6deea57f
TS
2605No interpolation is performed at this stage.
2606Any backslashed sequences including C<\\> are treated at the stage
2607to L</"parsing regular expressions">.
89d205f2 2608
6deea57f 2609=item C<''>, C<q//>, C<tr'''>, C<y'''>, the replacement of C<s'''>
75e14d17 2610
89d205f2 2611The only interpolation is removal of C<\> from pairs of C<\\>.
6deea57f
TS
2612Therefore C<-> in C<tr'''> and C<y'''> is treated literally
2613as a hyphen and no character range is available.
2614C<\1> in the replacement of C<s'''> does not work as C<\$1>.
89d205f2
YO
2615
2616=item C<tr///>, C<y///>
2617
6deea57f
TS
2618No variable interpolation occurs. String modifying combinations for
2619case and quoting such as C<\Q>, C<\U>, and C<\E> are not recognized.
2620The other escape sequences such as C<\200> and C<\t> and backslashed
2621characters such as C<\\> and C<\-> are converted to appropriate literals.
89d205f2
YO
2622The character C<-> is treated specially and therefore C<\-> is treated
2623as a literal C<->.
75e14d17 2624
89d205f2 2625=item C<"">, C<``>, C<qq//>, C<qx//>, C<< <file*glob> >>, C<<<"EOF">
75e14d17 2626
628253b8 2627C<\Q>, C<\U>, C<\u>, C<\L>, C<\l>, C<\F> (possibly paired with C<\E>) are
19799a22
GS
2628converted to corresponding Perl constructs. Thus, C<"\$foo\Qbaz\$bar">
2629is converted to C<\$foo . (quotemeta("baz" . \$bar))> internally.
6deea57f
TS
2630The other escape sequences such as C<\200> and C<\t> and backslashed
2631characters such as C<\\> and C<\-> are replaced with appropriate
2632expansions.
2a94b7ce 2633
19799a22
GS
2634Let it be stressed that I<whatever falls between C<\Q> and C<\E>>
2635is interpolated in the usual way. Something like C<"\Q\\E"> has
2636no C<\E> inside. instead, it has C<\Q>, C<\\>, and C<E>, so the
2637result is the same as for C<"\\\\E">. As a general rule, backslashes
2638between C<\Q> and C<\E> may lead to counterintuitive results. So,
2639C<"\Q\t\E"> is converted to C<quotemeta("\t")>, which is the same
2640as C<"\\\t"> (since TAB is not alphanumeric). Note also that:
2a94b7ce
IZ
2641
2642 \$str = '\t';
2643 return "\Q\$str";
2644
2645may be closer to the conjectural I<intention> of the writer of C<"\Q\t\E">.
2646
19799a22 2647Interpolated scalars and arrays are converted internally to the C<join> and
92d29cee 2648C<.> catenation operations. Thus, C<"\$foo XXX '@arr'"> becomes:
75e14d17 2649
19799a22 2650 \$foo . " XXX '" . (join \$", @arr) . "'";
75e14d17 2651
19799a22 2652All operations above are performed simultaneously, left to right.
75e14d17 2653
19799a22
GS
2654Because the result of C<"\Q STRING \E"> has all metacharacters
2655quoted, there is no way to insert a literal C<\$> or C<@> inside a
2656C<\Q\E> pair. If protected by C<\>, C<\$> will be quoted to became
2657C<"\\\\$">; if not, it is interpreted as the start of an interpolated
2658scalar.
75e14d17 2659
19799a22 2660Note also that the interpolation code needs to make a decision on
89d205f2 2661where the interpolated scalar ends. For instance, whether
35f2feb0 2662C<< "a \$b -> {c}" >> really means:
75e14d17
IZ
2663
2664 "a " . \$b . " -> {c}";
2665
2a94b7ce 2666or:
75e14d17
IZ
2667
2668 "a " . \$b -> {c};
2669
19799a22
GS
2670Most of the time, the longest possible text that does not include
2671spaces between components and which contains matching braces or
2672brackets. because the outcome may be determined by voting based
2673on heuristic estimators, the result is not strictly predictable.
2674Fortunately, it's usually correct for ambiguous cases.
75e14d17 2675
6deea57f 2676=item the replacement of C<s///>
75e14d17 2677
628253b8 2678Processing of C<\Q>, C<\U>, C<\u>, C<\L>, C<\l>, C<\F> and interpolation
6deea57f
TS
2679happens as with C<qq//> constructs.
2680
2681It is at this step that C<\1> is begrudgingly converted to C<\$1> in
2682the replacement text of C<s///>, in order to correct the incorrigible
2683I<sed> hackers who haven't picked up the saner idiom yet. A warning
2684is emitted if the C<use warnings> pragma or the B<-w> command-line flag
2685(that is, the C<\$^W> variable) was set.
2686
2687=item C<RE> in C<?RE?>, C</RE/>, C<m/RE/>, C<s/RE/foo/>,
2688
628253b8 2689Processing of C<\Q>, C<\U>, C<\u>, C<\L>, C<\l>, C<\F>, C<\E>,
cc74c5bd
TS
2690and interpolation happens (almost) as with C<qq//> constructs.
2691
5d03b57c
KW
2692Processing of C<\N{...}> is also done here, and compiled into an intermediate
2693form for the regex compiler. (This is because, as mentioned below, the regex
2694compilation may be done at execution time, and C<\N{...}> is a compile-time
2695construct.)
2696
cc74c5bd
TS
2697However any other combinations of C<\> followed by a character
2698are not substituted but only skipped, in order to parse them
2699as regular expressions at the following step.
6deea57f 2700As C<\c> is skipped at this step, C<@> of C<\c@> in RE is possibly
1749ea0d 2701treated as an array symbol (for example C<@foo>),
6deea57f 2702even though the same text in C<qq//> gives interpolation of C<\c@>.
6deea57f 2703
e128ab2c
DM
2704Code blocks such as C<(?{BLOCK})> are handled by temporarily passing control
2705back to the perl parser, in a similar way that an interpolated array
2706subscript expression such as C<"foo\$array[1+f("[xyz")]bar"> would be.
2707
6deea57f 2708Moreover, inside C<(?{BLOCK})>, C<(?# comment )>, and
19799a22
GS
2709a C<#>-comment in a C<//x>-regular expression, no processing is
2710performed whatsoever. This is the first step at which the presence
2711of the C<//x> modifier is relevant.
2712
1749ea0d
TS
2713Interpolation in patterns has several quirks: C<\$|>, C<\$(>, C<\$)>, C<@+>
2714and C<@-> are not interpolated, and constructs C<\$var[SOMETHING]> are
2715voted (by several different estimators) to be either an array element
2716or C<\$var> followed by an RE alternative. This is where the notation
19799a22
GS
2717C<\${arr[\$bar]}> comes handy: C</\${arr[0-9]}/> is interpreted as
2718array element C<-9>, not as a regular expression from the variable
2719C<\$arr> followed by a digit, which would be the interpretation of
2720C</\$arr[0-9]/>. Since voting among different estimators may occur,
2721the result is not predictable.
2722
19799a22
GS
2723The lack of processing of C<\\> creates specific restrictions on
2724the post-processed text. If the delimiter is C</>, one cannot get
2725the combination C<\/> into the result of this step. C</> will
2726finish the regular expression, C<\/> will be stripped to C</> on
2727the previous step, and C<\\/> will be left as is. Because C</> is
2728equivalent to C<\/> inside a regular expression, this does not
2729matter unless the delimiter happens to be character special to the
2730RE engine, such as in C<s*foo*bar*>, C<m[foo]>, or C<?foo?>; or an
2731alphanumeric char, as in:
2a94b7ce
IZ
2732
2733 m m ^ a \s* b mmx;
2734
19799a22 2735In the RE above, which is intentionally obfuscated for illustration, the
6deea57f 2736delimiter is C<m>, the modifier is C<mx>, and after delimiter-removal the
89d205f2 2737RE is the same as for C<m/ ^ a \s* b /mx>. There's more than one
19799a22
GS
2738reason you're encouraged to restrict your delimiters to non-alphanumeric,
2739non-whitespace choices.
75e14d17
IZ
2740
2741=back
2742
19799a22 2743This step is the last one for all constructs except regular expressions,
75e14d17
IZ
2744which are processed further.
2745
6deea57f
TS
2746=item parsing regular expressions
2747X<regexp, parse>
75e14d17 2748
19799a22 2749Previous steps were performed during the compilation of Perl code,
ac036724 2750but this one happens at run time, although it may be optimized to
19799a22 2751be calculated at compile time if appropriate. After preprocessing
6deea57f 2752described above, and possibly after evaluation if concatenation,
19799a22
GS
2753joining, casing translation, or metaquoting are involved, the
2754resulting I<string> is passed to the RE engine for compilation.
2755
2756Whatever happens in the RE engine might be better discussed in L<perlre>,
2757but for the sake of continuity, we shall do so here.
2758
2759This is another step where the presence of the C<//x> modifier is
2760relevant. The RE engine scans the string from left to right and
2761converts it to a finite automaton.
2762
2763Backslashed characters are either replaced with corresponding
2764literal strings (as with C<\{>), or else they generate special nodes
2765in the finite automaton (as with C<\b>). Characters special to the
2766RE engine (such as C<|>) generate corresponding nodes or groups of
2767nodes. C<(?#...)> comments are ignored. All the rest is either
2768converted to literal strings to match, or else is ignored (as is
2769whitespace and C<#>-style comments if C<//x> is present).
2770
2771Parsing of the bracketed character class construct, C<[...]>, is
2772rather different than the rule used for the rest of the pattern.
2773The terminator of this construct is found using the same rules as
2774for finding the terminator of a C<{}>-delimited construct, the only
2775exception being that C<]> immediately following C<[> is treated as
e128ab2c
DM
2776though preceded by a backslash.
2777
2778The terminator of runtime C<(?{...})> is found by temporarily switching
2779control to the perl parser, which should stop at the point where the
2780logically balancing terminating C<}> is found.
19799a22
GS
2781
2782It is possible to inspect both the string given to RE engine and the
2783resulting finite automaton. See the arguments C<debug>/C<debugcolor>
2784in the C<use L<re>> pragma, as well as Perl's B<-Dr> command-line
4a4eefd0 2785switch documented in L<perlrun/"Command Switches">.
75e14d17
IZ
2786
2787=item Optimization of regular expressions
d74e8afc 2788X<regexp, optimization>
75e14d17 2789
7522fed5 2790This step is listed for completeness only. Since it does not change
75e14d17 2791semantics, details of this step are not documented and are subject
19799a22
GS
2792to change without notice. This step is performed over the finite
2793automaton that was generated during the previous pass.
2a94b7ce 2794
19799a22
GS
2795It is at this stage that C<split()> silently optimizes C</^/> to
2796mean C</^/m>.
75e14d17
IZ
2797
2798=back
2799
a0d0e21e 2800=head2 I/O Operators
d74e8afc
ITB
2801X<operator, i/o> X<operator, io> X<io> X<while> X<filehandle>
2802X<< <> >> X<@ARGV>
a0d0e21e 2803
54310121 2804There are several I/O operators you should know about.
fbad3eb5 2805
7b8d334a 2806A string enclosed by backticks (grave accents) first undergoes
19799a22
GS
2807double-quote interpolation. It is then interpreted as an external
2808command, and the output of that command is the value of the
e9c56f9b
JH
2809backtick string, like in a shell. In scalar context, a single string
2810consisting of all output is returned. In list context, a list of
2811values is returned, one per line of output. (You can set C<\$/> to use
2812a different line terminator.) The command is executed each time the
2813pseudo-literal is evaluated. The status value of the command is
2814returned in C<\$?> (see L<perlvar> for the interpretation of C<\$?>).
2815Unlike in B<csh>, no translation is done on the return data--newlines
2816remain newlines. Unlike in any of the shells, single quotes do not
2817hide variable names in the command from interpretation. To pass a
2818literal dollar-sign through to the shell you need to hide it with a
2819backslash. The generalized form of backticks is C<qx//>. (Because
2820backticks always undergo shell expansion as well, see L<perlsec> for
2821security concerns.)
d74e8afc 2822X<qx> X<`> X<``> X<backtick> X<glob>
19799a22
GS
2823
2824In scalar context, evaluating a filehandle in angle brackets yields
2825the next line from that file (the newline, if any, included), or
2826C<undef> at end-of-file or on error. When C<\$/> is set to C<undef>
2827(sometimes known as file-slurp mode) and the file is empty, it
2828returns C<''> the first time, followed by C<undef> subsequently.
2829
2830Ordinarily you must assign the returned value to a variable, but
2831there is one situation where an automatic assignment happens. If
2832and only if the input symbol is the only thing inside the conditional
2833of a C<while> statement (even if disguised as a C<for(;;)> loop),
2834the value is automatically assigned to the global variable \$_,
2835destroying whatever was there previously. (This may seem like an
2836odd thing to you, but you'll use the construct in almost every Perl
17b829fa 2837script you write.) The \$_ variable is not implicitly localized.
19799a22
GS
2838You'll have to put a C<local \$_;> before the loop if you want that
2839to happen.
2840
2841The following lines are equivalent:
a0d0e21e 2842
748a9306 2843 while (defined(\$_ = <STDIN>)) { print; }
7b8d334a 2844 while (\$_ = <STDIN>) { print; }
a0d0e21e
LW
2845 while (<STDIN>) { print; }
2846 for (;<STDIN>;) { print; }
748a9306 2847 print while defined(\$_ = <STDIN>);
7b8d334a 2848 print while (\$_ = <STDIN>);
a0d0e21e
LW
2849 print while <STDIN>;
2850
1ca345ed
TC
2851This also behaves similarly, but assigns to a lexical variable
2852instead of to C<\$_>:
7b8d334a 2853
89d205f2 2854 while (my \$line = <STDIN>) { print \$line }
7b8d334a 2855
19799a22
GS
2856In these loop constructs, the assigned value (whether assignment
2857is automatic or explicit) is then tested to see whether it is
1ca345ed
TC
2858defined. The defined test avoids problems where the line has a string
2859value that would be treated as false by Perl; for example a "" or
19799a22
GS
2860a "0" with no trailing newline. If you really mean for such values
2861to terminate the loop, they should be tested for explicitly:
7b8d334a
GS
2862
2863 while ((\$_ = <STDIN>) ne '0') { ... }
2864 while (<STDIN>) { last unless \$_; ... }
2865
1ca345ed 2866In other boolean contexts, C<< <FILEHANDLE> >> without an
5ef4d93e 2867explicit C<defined> test or comparison elicits a warning if the
9f1b1f2d 2868C<use warnings> pragma or the B<-w>
19799a22 2869command-line switch (the C<\$^W> variable) is in effect.
7b8d334a 2870
5f05dabc 2871The filehandles STDIN, STDOUT, and STDERR are predefined. (The
19799a22
GS
2872filehandles C<stdin>, C<stdout>, and C<stderr> will also work except
2873in packages, where they would be interpreted as local identifiers
2874rather than global.) Additional filehandles may be created with
2875the open() function, amongst others. See L<perlopentut> and
2876L<perlfunc/open> for details on this.
d74e8afc 2877X<stdin> X<stdout> X<sterr>
a0d0e21e 2878
35f2feb0 2879If a <FILEHANDLE> is used in a context that is looking for
19799a22
GS
2880a list, a list comprising all input lines is returned, one line per
2881list element. It's easy to grow to a rather large data space this
2882way, so use with care.
a0d0e21e 2883
35f2feb0 2884<FILEHANDLE> may also be spelled C<readline(*FILEHANDLE)>.
19799a22 2885See L<perlfunc/readline>.
fbad3eb5 2886
35f2feb0 2887The null filehandle <> is special: it can be used to emulate the
1ca345ed
TC
2888behavior of B<sed> and B<awk>, and any other Unix filter program
2889that takes a list of filenames, doing the same to each line
2890of input from all of them. Input from <> comes either from
a0d0e21e 2891standard input, or from each file listed on the command line. Here's
35f2feb0 2892how it works: the first time <> is evaluated, the @ARGV array is
5a964f20 2893checked, and if it is empty, C<\$ARGV[0]> is set to "-", which when opened
a0d0e21e
LW
2894gives you standard input. The @ARGV array is then processed as a list
2895of filenames. The loop
2896
2897 while (<>) {
2898 ... # code for each line
2899 }
2900
2901is equivalent to the following Perl-like pseudo code:
2902
3e3baf6d 2903 unshift(@ARGV, '-') unless @ARGV;
a0d0e21e
LW
2904 while (\$ARGV = shift) {
2905 open(ARGV, \$ARGV);
2906 while (<ARGV>) {
2907 ... # code for each line
2908 }
2909 }
2910
19799a22
GS
2911except that it isn't so cumbersome to say, and will actually work.
2912It really does shift the @ARGV array and put the current filename
2913into the \$ARGV variable. It also uses filehandle I<ARGV>
ac036724 2914internally. <> is just a synonym for <ARGV>, which
19799a22 2915is magical. (The pseudo code above doesn't work because it treats
35f2feb0 2916<ARGV> as non-magical.)
a0d0e21e 2917
48ab5743
ML
2918Since the null filehandle uses the two argument form of L<perlfunc/open>
2919it interprets special characters, so if you have a script like this:
2920
2921 while (<>) {
2922 print;
2923 }
2924
2925and call it with C<perl dangerous.pl 'rm -rfv *|'>, it actually opens a
2926pipe, executes the C<rm> command and reads C<rm>'s output from that pipe.
2927If you want all items in C<@ARGV> to be interpreted as file names, you
2928can use the module C<ARGV::readonly> from CPAN.
2929
35f2feb0 2930You can modify @ARGV before the first <> as long as the array ends up
a0d0e21e 2931containing the list of filenames you really want. Line numbers (C<\$.>)
19799a22
GS
2932continue as though the input were one big happy file. See the example
2933in L<perlfunc/eof> for how to reset line numbers on each file.
5a964f20 2934
89d205f2 2935If you want to set @ARGV to your own list of files, go right ahead.
5a964f20
TC
2936This sets @ARGV to all plain text files if no @ARGV was given:
2937
2938 @ARGV = grep { -f && -T } glob('*') unless @ARGV;
a0d0e21e 2939
5a964f20
TC
2940You can even set them to pipe commands. For example, this automatically
2941filters compressed arguments through B<gzip>:
2942
2943 @ARGV = map { /\.(gz|Z)\$/ ? "gzip -dc < \$_ |" : \$_ } @ARGV;
2944
2945If you want to pass switches into your script, you can use one of the
a0d0e21e
LW
2946Getopts modules or put a loop on the front like this:
2947
2948 while (\$_ = \$ARGV[0], /^-/) {
2949 shift;
2950 last if /^--\$/;
2951 if (/^-D(.*)/) { \$debug = \$1 }
2952 if (/^-v/) { \$verbose++ }
5a964f20 2953 # ... # other switches
a0d0e21e 2954 }
5a964f20 2955
a0d0e21e 2956 while (<>) {
5a964f20 2957 # ... # code for each line
a0d0e21e
LW
2958 }
2959
89d205f2
YO
2960The <> symbol will return C<undef> for end-of-file only once.
2961If you call it again after this, it will assume you are processing another
19799a22 2962@ARGV list, and if you haven't set @ARGV, will read input from STDIN.
a0d0e21e 2963
1ca345ed 2964If what the angle brackets contain is a simple scalar variable (for example,
35f2feb0 2965<\$foo>), then that variable contains the name of the
19799a22
GS
2966filehandle to input from, or its typeglob, or a reference to the
2967same. For example:
cb1a09d0
AD
2968
2969 \$fh = \*STDIN;
2970 \$line = <\$fh>;
a0d0e21e 2971
5a964f20
TC
2972If what's within the angle brackets is neither a filehandle nor a simple
2973scalar variable containing a filehandle name, typeglob, or typeglob
2974reference, it is interpreted as a filename pattern to be globbed, and
2975either a list of filenames or the next filename in the list is returned,
19799a22 2976depending on context. This distinction is determined on syntactic
35f2feb0
GS
2977grounds alone. That means C<< <\$x> >> is always a readline() from
2978an indirect handle, but C<< <\$hash{key}> >> is always a glob().
5a964f20 2979That's because \$x is a simple scalar variable, but C<\$hash{key}> is
ef191992
YST
2980not--it's a hash element. Even C<< <\$x > >> (note the extra space)
2981is treated as C<glob("\$x ")>, not C<readline(\$x)>.
5a964f20
TC
2982
2983One level of double-quote interpretation is done first, but you can't
35f2feb0 2984say C<< <\$foo> >> because that's an indirect filehandle as explained
5a964f20
TC
2985in the previous paragraph. (In older versions of Perl, programmers
2986would insert curly brackets to force interpretation as a filename glob:
35f2feb0 2987C<< <\${foo}> >>. These days, it's considered cleaner to call the
5a964f20 2988internal function directly as C<glob(\$foo)>, which is probably the right
19799a22 2989way to have done it in the first place.) For example:
a0d0e21e
LW
2990
2991 while (<*.c>) {
2992 chmod 0644, \$_;
2993 }
2994
3a4b19e4 2995is roughly equivalent to:
a0d0e21e
LW
2996
2997 open(FOO, "echo *.c | tr -s ' \t\r\f' '\\012\\012\\012\\012'|");
2998 while (<FOO>) {
5b3eff12 2999 chomp;
a0d0e21e
LW
3000 chmod 0644, \$_;
3001 }
3002
3a4b19e4
GS
3003except that the globbing is actually done internally using the standard
3004C<File::Glob> extension. Of course, the shortest way to do the above is:
a0d0e21e
LW
3005
3006 chmod 0644, <*.c>;
3007
19799a22
GS
3008A (file)glob evaluates its (embedded) argument only when it is
3009starting a new list. All values must be read before it will start
3010over. In list context, this isn't important because you automatically
3011get them all anyway. However, in scalar context the operator returns
069e01df 3012the next value each time it's called, or C<undef> when the list has
19799a22
GS
3013run out. As with filehandle reads, an automatic C<defined> is
3014generated when the glob occurs in the test part of a C<while>,
1ca345ed
TC
3015because legal glob returns (for example,
3016a file called F<0>) would otherwise
19799a22
GS
3017terminate the loop. Again, C<undef> is returned only once. So if
3018you're expecting a single value from a glob, it is much better to
3019say
4633a7c4
LW
3020
3021 (\$file) = <blurch*>;
3022
3023than
3024
3025 \$file = <blurch*>;
3026
3027because the latter will alternate between returning a filename and
19799a22 3028returning false.
4633a7c4 3029
b159ebd3 3030If you're trying to do variable interpolation, it's definitely better
4633a7c4 3031to use the glob() function, because the older notation can cause people
e37d713d 3032to become confused with the indirect filehandle notation.
4633a7c4
LW
3033
3034 @files = glob("\$dir/*.[ch]");
3035 @files = glob(\$files[\$i]);
3036
a0d0e21e 3037=head2 Constant Folding
d74e8afc 3038X<constant folding> X<folding>
a0d0e21e
LW
3039
3040Like C, Perl does a certain amount of expression evaluation at
19799a22 3041compile time whenever it determines that all arguments to an
a0d0e21e
LW
3042operator are static and have no side effects. In particular, string
3043concatenation happens at compile time between literals that don't do
19799a22 3044variable substitution. Backslash interpolation also happens at
a0d0e21e
LW
3045compile time. You can say
3046
1ca345ed
TC
3047 'Now is the time for all'
3048 . "\n"
3049 . 'good men to come to.'
a0d0e21e 3050
54310121 3051and this all reduces to one string internally. Likewise, if
a0d0e21e
LW
3052you say
3053
3054 foreach \$file (@filenames) {
5a964f20 3055 if (-s \$file > 5 + 100 * 2**16) { }
54310121 3056 }
a0d0e21e 3057
1ca345ed 3058the compiler precomputes the number which that expression
19799a22 3059represents so that the interpreter won't have to.
a0d0e21e 3060
fd1abbef 3061=head2 No-ops
d74e8afc 3062X<no-op> X<nop>
fd1abbef
DN
3063
3064Perl doesn't officially have a no-op operator, but the bare constants
1ca345ed 3065C<0> and C<1> are special-cased not to produce a warning in void
fd1abbef
DN
3066context, so you can for example safely do
3067
3068 1 while foo();
3069
2c268ad5 3070=head2 Bitwise String Operators
d74e8afc 3071X<operator, bitwise, string>
2c268ad5
TP
3072
3073Bitstrings of any size may be manipulated by the bitwise operators
3074(C<~ | & ^>).
3075
19799a22
GS
3076If the operands to a binary bitwise op are strings of different
3077sizes, B<|> and B<^> ops act as though the shorter operand had
3078additional zero bits on the right, while the B<&> op acts as though
3079the longer operand were truncated to the length of the shorter.
3080The granularity for such extension or truncation is one or more
3081bytes.
2c268ad5 3082
89d205f2 3083 # ASCII-based examples
2c268ad5
TP
3084 print "j p \n" ^ " a h"; # prints "JAPH\n"
3085 print "JA" | " ph\n"; # prints "japh\n"
3086 print "japh\nJunk" & '_____'; # prints "JAPH\n";
3087 print 'p N\$' ^ " E<H\n"; # prints "Perl\n";
3088
19799a22 3089If you are intending to manipulate bitstrings, be certain that
2c268ad5 3090you're supplying bitstrings: If an operand is a number, that will imply
19799a22 3091a B<numeric> bitwise operation. You may explicitly show which type of
2c268ad5
TP
3092operation you intend by using C<""> or C<0+>, as in the examples below.
3093
4358a253
SS
3094 \$foo = 150 | 105; # yields 255 (0x96 | 0x69 is 0xFF)
3095 \$foo = '150' | 105; # yields 255
2c268ad5
TP
3096 \$foo = 150 | '105'; # yields 255
3097 \$foo = '150' | '105'; # yields string '155' (under ASCII)
3098
3099 \$baz = 0+\$foo & 0+\$bar; # both ops explicitly numeric
3100 \$biz = "\$foo" ^ "\$bar"; # both ops explicitly stringy
a0d0e21e 3101
1ae175c8
GS
3102See L<perlfunc/vec> for information on how to manipulate individual bits
3103in a bit vector.
3104
55497cff 3105=head2 Integer Arithmetic
d74e8afc 3106X<integer>
a0d0e21e 3107
19799a22 3108By default, Perl assumes that it must do most of its arithmetic in
a0d0e21e
LW
3109floating point. But by saying
3110
3111 use integer;
3112
3eab78e3
CW
3113you may tell the compiler to use integer operations
3114(see L<integer> for a detailed explanation) from here to the end of
3115the enclosing BLOCK. An inner BLOCK may countermand this by saying
a0d0e21e
LW
3116
3117 no integer;
3118
19799a22 3119which lasts until the end of that BLOCK. Note that this doesn't
3eab78e3
CW
3120mean everything is an integer, merely that Perl will use integer
3121operations for arithmetic, comparison, and bitwise operators. For
3122example, even under C<use integer>, if you take the C<sqrt(2)>, you'll
3123still get C<1.4142135623731> or so.
19799a22
GS
3124
3125Used on numbers, the bitwise operators ("&", "|", "^", "~", "<<",
89d205f2 3126and ">>") always produce integral results. (But see also
13a2d996 3127L<Bitwise String Operators>.) However, C<use integer> still has meaning for
19799a22
GS
3128them. By default, their results are interpreted as unsigned integers, but
3129if C<use integer> is in effect, their results are interpreted
3130as signed integers. For example, C<~0> usually evaluates to a large
0be96356 3131integral value. However, C<use integer; ~0> is C<-1> on two's-complement
19799a22 3132machines.
68dc0745 3133
3134=head2 Floating-point Arithmetic
06ce2fa3 3135
d74e8afc 3136X<floating-point> X<floating point> X<float> X<real>
68dc0745 3137
3138While C<use integer> provides integer-only arithmetic, there is no
19799a22
GS
3139analogous mechanism to provide automatic rounding or truncation to a
3140certain number of decimal places. For rounding to a certain number
3141of digits, sprintf() or printf() is usually the easiest route.
3142See L<perlfaq4>.
68dc0745 3143
5a964f20
TC
3144Floating-point numbers are only approximations to what a mathematician
3145would call real numbers. There are infinitely more reals than floats,
3146so some corners must be cut. For example:
3147
3148 printf "%.20g\n", 123456789123456789;
3149 # produces 123456789123456784
3150
8548cb57
RGS
3151Testing for exact floating-point equality or inequality is not a
3152good idea. Here's a (relatively expensive) work-around to compare
5a964f20
TC
3153whether two floating-point numbers are equal to a particular number of
3154decimal places. See Knuth, volume II, for a more robust treatment of
3155this topic.
3156
3157 sub fp_equal {
3158 my (\$X, \$Y, \$POINTS) = @_;
3159 my (\$tX, \$tY);
3160 \$tX = sprintf("%.\${POINTS}g", \$X);
3161 \$tY = sprintf("%.\${POINTS}g", \$Y);
3162 return \$tX eq \$tY;
3163 }
3164
68dc0745 3165The POSIX module (part of the standard perl distribution) implements
19799a22
GS
3166ceil(), floor(), and other mathematical and trigonometric functions.
3167The Math::Complex module (part of the standard perl distribution)
3168defines mathematical functions that work on both the reals and the
3169imaginary numbers. Math::Complex not as efficient as POSIX, but
68dc0745 3170POSIX can't work with complex numbers.
3171
3172Rounding in financial applications can have serious implications, and
3173the rounding method used should be specified precisely. In these
3174cases, it probably pays not to trust whichever system rounding is
3175being used by Perl, but to instead implement the rounding function you
3176need yourself.
5a964f20
TC
3177
3178=head2 Bigger Numbers
d74e8afc 3179X<number, arbitrary precision>
5a964f20 3180
c543c01b 3181The standard C<Math::BigInt>, C<Math::BigRat>, and C<Math::BigFloat> modules,
fb1a95c6 3182along with the C<bignum>, C<bigint>, and C<bigrat> pragmas, provide
19799a22 3183variable-precision arithmetic and overloaded operators, although
cd5c4fce 3184they're currently pretty slow. At the cost of some space and
19799a22
GS
3185considerable speed, they avoid the normal pitfalls associated with
3186limited-precision representations.
5a964f20 3187
c543c01b
TC
3188 use 5.010;
3189 use bigint; # easy interface to Math::BigInt
3190 \$x = 123456789123456789;
3191 say \$x * \$x;
3192 +15241578780673678515622620750190521
3193
3194Or with rationals:
3195
3196 use 5.010;
3197 use bigrat;
3198 \$a = 3/22;
3199 \$b = 4/6;
3200 say "a/b is ", \$a/\$b;
3201 say "a*b is ", \$a*\$b;
3202 a/b is 9/44
3203 a*b is 1/11
3204
3205Several modules let you calculate with (bound only by memory and CPU time)
3206unlimited or fixed precision. There are also some non-standard modules that
3207provide faster implementations via external C libraries.
cd5c4fce
T
3208
3209Here is a short, but incomplete summary:
3210
950b09ed
KW
3211 Math::Fraction big, unlimited fractions like 9973 / 12967
3212 Math::String treat string sequences like numbers
3213 Math::FixedPrecision calculate with a fixed precision
3214 Math::Currency for currency calculations
3215 Bit::Vector manipulate bit vectors fast (uses C)
3216 Math::BigIntFast Bit::Vector wrapper for big numbers
3217 Math::Pari provides access to the Pari C library
3218 Math::BigInteger uses an external C library
3219 Math::Cephes uses external Cephes C library (no big numbers)
3220 Math::Cephes::Fraction fractions via the Cephes library
3221 Math::GMP another one using an external C library
cd5c4fce
T
3222
3223Choose wisely.
16070b82
GS
3224
3225=cut