This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
perldata: Change pod to reflect reality
[perl5.git] / pod / perldata.pod
CommitLineData
a0d0e21e
LW
1=head1 NAME
2
cb1a09d0 3perldata - Perl data types
a0d0e21e
LW
4
5=head1 DESCRIPTION
6
7=head2 Variable names
d74e8afc 8X<variable, name> X<variable name> X<data type> X<type>
a0d0e21e 9
d55a8828 10Perl has three built-in data types: scalars, arrays of scalars, and
692ef166
SF
11associative arrays of scalars, known as "hashes". A scalar is a
12single string (of any size, limited only by the available memory),
13number, or a reference to something (which will be discussed
14in L<perlref>). Normal arrays are ordered lists of scalars indexed
15by number, starting with 0. Hashes are unordered collections of scalar
16values indexed by their associated string key.
a0d0e21e 17
d55a8828 18Values are usually referred to by name, or through a named reference.
b88cefa9
PP
19The first character of the name tells you to what sort of data
20structure it refers. The rest of the name tells you the particular
d55a8828
TC
21value to which it refers. Usually this name is a single I<identifier>,
22that is, a string beginning with a letter or underscore, and
23containing letters, underscores, and digits. In some cases, it may
24be a chain of identifiers, separated by C<::> (or by the slightly
25archaic C<'>); all but the last are interpreted as names of packages,
26to locate the namespace in which to look up the final identifier
32833930
BF
27(see L<perlmod/Packages> for details). For a more in-depth discussion
28on identifiers, see L<Identifier parsing>. It's possible to
29substitute for a simple identifier, an expression that produces a reference
d55a8828
TC
30to the value at runtime. This is described in more detail below
31and in L<perlref>.
d74e8afc 32X<identifier>
d55a8828
TC
33
34Perl also has its own built-in variables whose names don't follow
35these rules. They have strange names so they don't accidentally
36collide with one of your normal variables. Strings that match
37parenthesized parts of a regular expression are saved under names
38containing only digits after the C<$> (see L<perlop> and L<perlre>).
39In addition, several special variables that provide windows into
40the inner working of Perl have names containing punctuation characters
41and control characters. These are documented in L<perlvar>.
d74e8afc 42X<variable, built-in>
d55a8828
TC
43
44Scalar values are always named with '$', even when referring to a
45scalar that is part of an array or a hash. The '$' symbol works
46semantically like the English word "the" in that it indicates a
47single value is expected.
d74e8afc 48X<scalar>
a0d0e21e
LW
49
50 $days # the simple scalar value "days"
51 $days[28] # the 29th element of array @days
52 $days{'Feb'} # the 'Feb' value from hash %days
53 $#days # the last index of array @days
54
d55a8828 55Entire arrays (and slices of arrays and hashes) are denoted by '@',
3921068c 56which works much as the word "these" or "those" does in English,
d55a8828 57in that it indicates multiple values are expected.
d74e8afc 58X<array>
a0d0e21e
LW
59
60 @days # ($days[0], $days[1],... $days[n])
d55a8828 61 @days[3,4,5] # same as ($days[3],$days[4],$days[5])
a0d0e21e
LW
62 @days{'a','c'} # same as ($days{'a'},$days{'c'})
63
d55a8828 64Entire hashes are denoted by '%':
d74e8afc 65X<hash>
a0d0e21e
LW
66
67 %days # (key1, val1, key2, val2 ...)
68
d55a8828
TC
69In addition, subroutines are named with an initial '&', though this
70is optional when unambiguous, just as the word "do" is often redundant
71in English. Symbol table entries can be named with an initial '*',
72but you don't really care about that yet (if ever :-).
73
74Every variable type has its own namespace, as do several
75non-variable identifiers. This means that you can, without fear
76of conflict, use the same name for a scalar variable, an array, or
77a hash--or, for that matter, for a filehandle, a directory handle, a
78subroutine name, a format name, or a label. This means that $foo
79and @foo are two different variables. It also means that C<$foo[1]>
80is a part of @foo, not a part of $foo. This may seem a bit weird,
81but that's okay, because it is weird.
d74e8afc 82X<namespace>
d55a8828
TC
83
84Because variable references always start with '$', '@', or '%', the
85"reserved" words aren't in fact reserved with respect to variable
86names. They I<are> reserved with respect to labels and filehandles,
87however, which don't have an initial special character. You can't
88have a filehandle named "log", for instance. Hint: you could say
89C<open(LOG,'logfile')> rather than C<open(log,'logfile')>. Using
90uppercase filehandles also improves readability and protects you
91from conflict with future reserved words. Case I<is> significant--"FOO",
92"Foo", and "foo" are all different names. Names that start with a
93letter or underscore may also contain digits and underscores.
d74e8afc
ITB
94X<identifier, case sensitivity>
95X<case>
a0d0e21e
LW
96
97It is possible to replace such an alphanumeric name with an expression
d55a8828 98that returns a reference to the appropriate type. For a description
a0d0e21e
LW
99of this, see L<perlref>.
100
5f05dabc 101Names that start with a digit may contain only more digits. Names
9539f610
RGS
102that do not start with a letter, underscore, digit or a caret (i.e.
103a control character) are limited to one character, e.g., C<$%> or
104C<$$>. (Most of these one character names have a predefined
105significance to Perl. For instance, C<$$> is the current process
106id.)
a0d0e21e 107
32833930
BF
108=head2 Identifier parsing
109X<identifiers>
110
111Up until Perl 5.18, the actual rules of what a valid identifier
112was were a bit fuzzy. However, in general, anything defined here should
113work on previous versions of Perl, while the opposite -- edge cases
114that work in previous versions, but aren't defined here -- probably
115won't work on newer versions.
116As an important side note, please note that the following only applies
117to bareword identifiers as found in Perl source code, not identifiers
118introduced through symbolic references, which have much fewer
119restrictions.
120If working under the effect of the C<use utf8;> pragma, the following
121rules apply:
122
9c1129c7
KW
123 / (?[ ( \p{Word} & \p{XID_Start} ) + [_] ])
124 (?[ ( \p{Word} & \p{XID_Continue} ) ]) * /x
125
126That is, a "start" character followed by any number of "continue"
127characters. Perl requires every character in an identifier to also
128match C<\w> (this prevents some problematic cases); and Perl
129additionally accepts identfier names beginning with an underscore.
32833930
BF
130
131If not under C<use utf8>, the source is treated as ASCII + 128 extra
132controls, and identifiers should match
133
134 / (?aa) (?!\d) \w+ /x
135
136That is, any word character in the ASCII range, as long as the first
137character is not a digit.
138
139There are two package separators in Perl: A double colon (C<::>) and a single
140quote (C<'>). Normal identifiers can start or end with a double colon, and
141can contain several parts delimited by double colons.
142Single quotes have similar rules, but with the exception that they are not
143legal at the end of an identifier: That is, C<$'foo> and C<$foo'bar> are
1d268002 144legal, but C<$foo'bar'> is not.
32833930 145
1d268002 146Additionally, if the identifier is preceded by a sigil --
32833930
BF
147that is, if the identifier is part of a variable name -- it
148may optionally be enclosed in braces.
149
150While you can mix double colons with singles quotes, the quotes must come
151after the colons: C<$::::'foo> and C<$foo::'bar> are legal, but C<$::'::foo>
152and C<$foo'::bar> are not.
153
154Put together, a grammar to match a basic identifier becomes
155
156 /
157 (?(DEFINE)
158 (?<variable>
159 (?&sigil)
160 (?:
161 (?&normal_identifier)
162 | \{ \s* (?&normal_identifier) \s* \}
163 )
164 )
165 (?<normal_identifier>
166 (?: :: )* '?
167 (?&basic_identifier)
168 (?: (?= (?: :: )+ '? | (?: :: )* ' ) (?&normal_identifier) )?
169 (?: :: )*
170 )
171 (?<basic_identifier>
172 # is use utf8 on?
173 (?(?{ (caller(0))[8] & $utf8::hint_bits })
4c106081 174 (?&Perl_XIDS) (?&Perl_XIDC)*
32833930
BF
175 | (?aa) (?!\d) \w+
176 )
177 )
178 (?<sigil> [&*\$\@\%])
179 (?<Perl_XIDS> (?[ ( \p{Word} & \p{XID_Start} ) + [_] ]) )
4c106081 180 (?<Perl_XIDC> (?[ \p{Word} & \p{XID_Continue} ]) )
32833930
BF
181 )
182 /x
183
184Meanwhile, special identifiers don't follow the above rules; For the most
185part, all of the identifiers in this category have a special meaning given
186by Perl. Because they have special parsing rules, these generally can't be
187fully-qualified. They come in four forms:
188
189=over
190
42327f06
KW
191=item *
192
193A sigil, followed solely by digits matching C<\p{POSIX_Digit}>, like
194C<$0>, C<$1>, or C<$10000>.
195
196=item *
197
87518e92
KW
198A sigil, followed by a caret and any one of the characters
199C<[][A-Z^_?\]>, like C<$^V> or C<$^]>, or a sigil followed by a literal non-space,
42327f06
KW
200non-C<NUL> control character matching the C<\p{POSIX_Cntrl}> property.
201Due to a historical oddity, if not running under C<use utf8>, the 128
202characters in the C<[0x80-0xff]> range are considered to be controls,
203and may also be used in length-one variables. However, the use of
204non-graphical characters is deprecated as of v5.22, and support for them
205will be removed in a future version of perl. ASCII space characters and
206C<NUL> already aren't allowed, so this means that a single-character
207variable name with that name being any other C0 control C<[0x01-0x1F]>,
208or C<DEL> will generate a deprecated warning. Already, under C<"use
209utf8">, non-ASCII characters must match C<Perl_XIDS>. As of v5.22, when
210not under C<"use utf8"> C1 controls C<[0x80-0x9F]>, NO BREAK SPACE, and
211SOFT HYPHEN (C<SHY>)) generate a deprecated warning.
212
213=item *
214
87518e92
KW
215Similar to the above, a sigil, followed by bareword text in braces,
216where the first character is either a caret followed by any one of
217the characters C<[][A-Z^_?\]>, like C<${^GLOBAL_PHASE}>, or a non-C<NUL>,
218non-space literal
42327f06
KW
219control like C<${\7LOBAL_PHASE}>. Like the above, when not under
220C<"use utf8">, the characters in C<[0x80-0xFF]> are considered controls, but as
221of v5.22, the use of any that are non-graphical are deprecated, and as
222of v5.20 the use of any ASCII-range literal control is deprecated.
223Support for these will be removed in a future version of perl.
224
225=item *
226
227A sigil followed by a single character matching the C<\p{POSIX_Punct}>
228property, like C<$!> or C<%+>, except the character C<"{"> doesn't work.
32833930
BF
229
230=back
231
b29f65fc 232Note that as of Perl 5.20, literal control characters in variable names
42327f06
KW
233are deprecated; and as of Perl 5.22, any other non-graphic characters
234are also deprecated.
b29f65fc 235
a0d0e21e 236=head2 Context
d74e8afc 237X<context> X<scalar context> X<list context>
a0d0e21e
LW
238
239The interpretation of operations and values in Perl sometimes depends
240on the requirements of the context around the operation or value.
d55a8828 241There are two major contexts: list and scalar. Certain operations
a0d0e21e 242return list values in contexts wanting a list, and scalar values
d55a8828
TC
243otherwise. If this is true of an operation it will be mentioned in
244the documentation for that operation. In other words, Perl overloads
a0d0e21e 245certain operations based on whether the expected return value is
d55a8828
TC
246singular or plural. Some words in English work this way, like "fish"
247and "sheep".
a0d0e21e
LW
248
249In a reciprocal fashion, an operation provides either a scalar or a
250list context to each of its arguments. For example, if you say
251
252 int( <STDIN> )
253
c47ff5f1 254the integer operation provides scalar context for the <>
a0d0e21e
LW
255operator, which responds by reading one line from STDIN and passing it
256back to the integer operation, which will then find the integer value
257of that line and return that. If, on the other hand, you say
258
259 sort( <STDIN> )
260
c47ff5f1 261then the sort operation provides list context for <>, which
a0d0e21e
LW
262will proceed to read every line available up to the end of file, and
263pass that list of lines back to the sort routine, which will then
264sort those lines and return them as a list to whatever the context
265of the sort was.
266
d55a8828
TC
267Assignment is a little bit special in that it uses its left argument
268to determine the context for the right argument. Assignment to a
269scalar evaluates the right-hand side in scalar context, while
270assignment to an array or hash evaluates the righthand side in list
271context. Assignment to a list (or slice, which is just a list
3921068c 272anyway) also evaluates the right-hand side in list context.
d55a8828 273
9f1b1f2d
GS
274When you use the C<use warnings> pragma or Perl's B<-w> command-line
275option, you may see warnings
d55a8828
TC
276about useless uses of constants or functions in "void context".
277Void context just means the value has been discarded, such as a
278statement containing only C<"fred";> or C<getpwuid(0);>. It still
279counts as scalar context for functions that care whether or not
280they're being called in list context.
281
282User-defined subroutines may choose to care whether they are being
283called in a void, scalar, or list context. Most subroutines do not
284need to bother, though. That's because both scalars and lists are
285automatically interpolated into lists. See L<perlfunc/wantarray>
286for how you would dynamically discern your function's calling
287context.
a0d0e21e
LW
288
289=head2 Scalar values
d74e8afc 290X<scalar> X<number> X<string> X<reference>
a0d0e21e 291
d55a8828
TC
292All data in Perl is a scalar, an array of scalars, or a hash of
293scalars. A scalar may contain one single value in any of three
294different flavors: a number, a string, or a reference. In general,
295conversion from one form to another is transparent. Although a
296scalar may not directly hold multiple values, it may contain a
297reference to an array or hash which in turn contains multiple values.
298
299Scalars aren't necessarily one thing or another. There's no place
300to declare a scalar variable to be of type "string", type "number",
301type "reference", or anything else. Because of the automatic
302conversion of scalars, operations that return scalars don't need
303to care (and in fact, cannot care) whether their caller is looking
304for a string, a number, or a reference. Perl is a contextually
305polymorphic language whose scalars can be strings, numbers, or
306references (which includes objects). Although strings and numbers
307are considered pretty much the same thing for nearly all purposes,
308references are strongly-typed, uncastable pointers with builtin
309reference-counting and destructor invocation.
a0d0e21e 310
32860eee
FC
311A scalar value is interpreted as FALSE in the Boolean sense
312if it is undefined, the null string or the number 0 (or its
313string equivalent, "0"), and TRUE if it is anything else. The
d55a8828
TC
314Boolean context is just a special kind of scalar context where no
315conversion to a string or a number is ever performed.
d74e8afc 316X<boolean> X<bool> X<true> X<false> X<truth>
d55a8828
TC
317
318There are actually two varieties of null strings (sometimes referred
319to as "empty" strings), a defined one and an undefined one. The
320defined version is just a string of length zero, such as C<"">.
321The undefined version is the value that indicates that there is
322no real value for something, such as when there was an error, or
323at end of file, or when you refer to an uninitialized variable or
324element of an array or hash. Although in early versions of Perl,
325an undefined scalar could become defined when first used in a
326place expecting a defined value, this no longer happens except for
327rare cases of autovivification as explained in L<perlref>. You can
328use the defined() operator to determine whether a scalar value is
329defined (this has no meaning on arrays or hashes), and the undef()
330operator to produce an undefined value.
d74e8afc 331X<defined> X<undefined> X<undef> X<null> X<string, null>
d55a8828
TC
332
333To find out whether a given string is a valid non-zero number, it's
334sometimes enough to test it against both numeric 0 and also lexical
692ef166
SF
335"0" (although this will cause noises if warnings are on). That's
336because strings that aren't numbers count as 0, just as they do in B<awk>:
4633a7c4
LW
337
338 if ($str == 0 && $str ne "0") {
339 warn "That doesn't look like a number";
54310121 340 }
4633a7c4 341
d55a8828
TC
342That method may be best because otherwise you won't treat IEEE
343notations like C<NaN> or C<Infinity> properly. At other times, you
344might prefer to determine whether string data can be used numerically
345by calling the POSIX::strtod() function or by inspecting your string
346with a regular expression (as documented in L<perlre>).
cb1a09d0
AD
347
348 warn "has nondigits" if /\D/;
5a964f20
TC
349 warn "not a natural number" unless /^\d+$/; # rejects -3
350 warn "not an integer" unless /^-?\d+$/; # rejects +3
351 warn "not an integer" unless /^[+-]?\d+$/;
352 warn "not a decimal number" unless /^-?\d+\.?\d*$/; # rejects .2
353 warn "not a decimal number" unless /^-?(?:\d+(?:\.\d*)?|\.\d+)$/;
54310121 354 warn "not a C float"
cb1a09d0
AD
355 unless /^([+-]?)(?=\d|\.\d)\d*(\.\d*)?([Ee]([+-]?\d+))?$/;
356
d55a8828 357The length of an array is a scalar value. You may find the length
fc518ee5 358of array @days by evaluating C<$#days>, as in B<csh>. However, this
359isn't the length of the array; it's the subscript of the last element,
360which is a different value since there is ordinarily a 0th element.
d55a8828
TC
361Assigning to C<$#days> actually changes the length of the array.
362Shortening an array this way destroys intervening values. Lengthening
363an array that was previously shortened does not recover values
0568eccd 364that were in those elements.
d74e8afc 365X<$#> X<array, length>
d55a8828 366
210b36aa 367You can also gain some minuscule measure of efficiency by pre-extending
d55a8828
TC
368an array that is going to get big. You can also extend an array
369by assigning to an element that is off the end of the array. You
19799a22 370can truncate an array down to nothing by assigning the null list
d55a8828 371() to it. The following are equivalent:
a0d0e21e 372
84f709e7 373 @whatever = ();
3e3baf6d 374 $#whatever = -1;
a0d0e21e 375
d55a8828
TC
376If you evaluate an array in scalar context, it returns the length
377of the array. (Note that this is not true of lists, which return
378the last value, like the C comma operator, nor of built-in functions,
379which return whatever they feel like returning.) The following is
380always true:
d74e8afc 381X<array, length>
a0d0e21e 382
a0d0e21e
LW
383 scalar(@whatever) == $#whatever + 1;
384
d55a8828
TC
385Some programmers choose to use an explicit conversion so as to
386leave nothing to doubt:
4633a7c4
LW
387
388 $element_count = scalar(@whatever);
389
d55a8828
TC
390If you evaluate a hash in scalar context, it returns false if the
391hash is empty. If there are any key/value pairs, it returns true;
392more precisely, the value returned is a string consisting of the
393number of used buckets and the number of allocated buckets, separated
394by a slash. This is pretty much useful only to find out whether
395Perl's internal hashing algorithm is performing poorly on your data
396set. For example, you stick 10,000 things in a hash, but evaluating
397%HASH in scalar context reveals C<"1/16">, which means only one out
398of sixteen buckets has been touched, and presumably contains all
126c71c8 39910,000 of your items. This isn't supposed to happen. If a tied hash
6e4c4703
DM
400is evaluated in scalar context, the C<SCALAR> method is called (with a
401fallback to C<FIRSTKEY>).
d74e8afc 402X<hash, scalar context> X<hash, bucket> X<bucket>
a0d0e21e 403
5a964f20 404You can preallocate space for a hash by assigning to the keys() function.
65841adf 405This rounds up the allocated buckets to the next power of two:
5a964f20
TC
406
407 keys(%users) = 1000; # allocate 1024 buckets
408
a0d0e21e 409=head2 Scalar value constructors
d74e8afc 410X<scalar, literal> X<scalar, constant>
a0d0e21e 411
d55a8828 412Numeric literals are specified in any of the following floating point or
a0d0e21e
LW
413integer formats:
414
a94ee3a1
KW
415 12345
416 12345.67
417 .23E-10 # a very small number
418 3.14_15_92 # a very important number
419 4_294_967_296 # underscore for legibility
420 0xff # hex
421 0xdead_beef # more hex
422 0377 # octal (only numbers, begins with 0)
423 0b011011 # binary
424 0x1.999ap-4 # hexadecimal floating point (the 'p' is required)
a0d0e21e 425
d4ced10d 426You are allowed to use underscores (underbars) in numeric literals
d823de2e
KW
427between digits for legibility (but not multiple underscores in a row:
428C<23__500> is not legal; C<23_500> is).
429You could, for example, group binary
d4ced10d
JH
430digits by threes (as for a Unix-style mode argument such as 0b110_100_100)
431or by fours (to represent nibbles, as in 0b1010_0110) or in other groups.
d74e8afc 432X<number, literal>
1d277562 433
55497cff 434String literals are usually delimited by either single or double
d55a8828
TC
435quotes. They work much like quotes in the standard Unix shells:
436double-quoted string literals are subject to backslash and variable
19799a22
GS
437substitution; single-quoted strings are not (except for C<\'> and
438C<\\>). The usual C-style backslash rules apply for making
d55a8828 439characters such as newline, tab, etc., as well as some more exotic
4a4eefd0 440forms. See L<perlop/"Quote and Quote-like Operators"> for a list.
d74e8afc 441X<string, literal>
d55a8828
TC
442
443Hexadecimal, octal, or binary, representations in string literals
444(e.g. '0xff') are not automatically converted to their integer
445representation. The hex() and oct() functions make these conversions
446for you. See L<perlfunc/hex> and L<perlfunc/oct> for more details.
68dc0745 447
61e61fbc
JH
448Hexadecimal floating point can start just like a hexadecimal literal,
449and it can be followed by an optional fractional hexadecimal part,
450but it must be followed by C<p>, an optional sign, and a power of two.
451The format is useful for accurately presenting floating point values,
452avoiding conversions to or from decimal floating point, and therefore
453avoiding possible loss in precision. Notice that while most current
60aaad76
JH
454platforms use the 64-bit IEEE 754 floating point, not all do. Another
455potential source of (low-order) differences are the floating point
456rounding modes, which can differ between CPUs, operating systems,
457and compilers, and which Perl doesn't control.
61e61fbc 458
5f05dabc 459You can also embed newlines directly in your strings, i.e., they can end
a0d0e21e
LW
460on a different line than they begin. This is nice, but if you forget
461your trailing quote, the error will not be reported until Perl finds
462another line containing the quote character, which may be much further
463on in the script. Variable substitution inside strings is limited to
d55a8828 464scalar variables, arrays, and array or hash slices. (In other words,
b88cefa9 465names beginning with $ or @, followed by an optional bracketed
a0d0e21e 466expression as a subscript.) The following code segment prints out "The
184e9718 467price is $Z<>100."
d74e8afc 468X<interpolation>
a0d0e21e 469
692ef166
SF
470 $Price = '$100'; # not interpolated
471 print "The price is $Price.\n"; # interpolated
472
473There is no double interpolation in Perl, so the C<$100> is left as is.
a0d0e21e 474
7e4353e9
RGS
475By default floating point numbers substituted inside strings use the
476dot (".") as the decimal separator. If C<use locale> is in effect,
477and POSIX::setlocale() has been called, the character used for the
478decimal separator is affected by the LC_NUMERIC locale.
479See L<perllocale> and L<POSIX>.
480
d55a8828 481As in some shells, you can enclose the variable name in braces to
f1cbbd6e
GS
482disambiguate it from following alphanumerics (and underscores).
483You must also do
d55a8828
TC
484this when interpolating a variable into a string to separate the
485variable name from a following double-colon or an apostrophe, since
486these would be otherwise treated as a package separator:
d74e8afc 487X<interpolation>
d55a8828 488
84f709e7 489 $who = "Larry";
d55a8828
TC
490 print PASSWD "${who}::0:0:Superuser:/:/bin/perl\n";
491 print "We use ${who}speak when ${who}'s here.\n";
492
493Without the braces, Perl would have looked for a $whospeak, a
494C<$who::0>, and a C<$who's> variable. The last two would be the
495$0 and the $s variables in the (presumably) non-existent package
496C<who>.
497
34a2706e 498In fact, a simple identifier within such curlies is forced to be
b4e2e1dd 499a string, and likewise within a hash subscript. Neither need
d55a8828
TC
500quoting. Our earlier example, C<$days{'Feb'}> can be written as
501C<$days{Feb}> and the quotes will be assumed automatically. But
719b43e8
RGS
502anything more complicated in the subscript will be interpreted as an
503expression. This means for example that C<$version{2.0}++> is
504equivalent to C<$version{2}++>, not to C<$version{'2.0'}++>.
d55a8828 505
f17ecf24
JH
506=head3 Special floating point: infinity (Inf) and not-a-number (NaN)
507
508Floating point values include the special values C<Inf> and C<NaN>,
509for infinity and not-a-number. The infinity can be also negative.
510
511The infinity is the result of certain math operations that overflow
512the floating point range, like 9**9**9. The not-a-number is the
513result when the result is undefined or unrepresentable. Though note
514that you cannot get C<NaN> from some common "undefined" or
515"out-of-range" operations like dividing by zero, or square root of
516a negative number, since Perl generates fatal errors for those.
517
518The infinity and not-a-number have their own special arithmetic rules.
519The general rule is that they are "contagious": C<Inf> plus one is
520C<Inf>, and C<NaN> plus one is C<NaN>. Where things get interesting
521is when you combine infinities and not-a-numbers: C<Inf> minus C<Inf>
522and C<Inf> divided by C<INf> are C<NaN> (while C<Inf> plus C<Inf> is
523C<Inf> and C<Inf> times C<Inf> is C<Inf>). C<NaN> is also curious
524in that it does not equal any number, I<including> itself:
525C<NaN> != C<NaN>.
526
527Perl doesn't understand C<Inf> and C<NaN> as numeric literals, but
528you can have them as strings, and Perl will convert them as needed:
529"Inf" + 1. (You can, however, import them from the POSIX extension;
530C<use POSIX qw(Inf NaN);> and then use them as literals.)
531
532Note that on input (string to number) Perl accepts C<Inf> and C<NaN>
533in many forms. Case is ignored, and the Win32-specific forms like
534C<1.#INF> are understood, but on output the values are normalized to
535C<Inf> and C<NaN>.
536
692ef166 537=head3 Version Strings
d74e8afc 538X<version string> X<vstring> X<v-string>
692ef166 539
191d61a7 540A literal of the form C<v1.20.300.4000> is parsed as a string composed
6b2463a0
JH
541of characters with the specified ordinals. This form, known as
542v-strings, provides an alternative, more readable way to construct
543strings, rather than use the somewhat less readable interpolation form
544C<"\x{1}\x{14}\x{12c}\x{fa0}">. This is useful for representing
545Unicode strings, and for comparing version "numbers" using the string
546comparison operators, C<cmp>, C<gt>, C<lt> etc. If there are two or
547more dots in the literal, the leading C<v> may be omitted.
b9c62f5b 548
2575c402 549 print v9786; # prints SMILEY, "\x{263a}"
b9c62f5b
GS
550 print v102.111.111; # prints "foo"
551 print 102.111.111; # same
552
553Such literals are accepted by both C<require> and C<use> for
a32521b7
JD
554doing a version check. Note that using the v-strings for IPv4
555addresses is not portable unless you also use the
556inet_aton()/inet_ntoa() routines of the Socket package.
191d61a7 557
d32a65d2 558Note that since Perl 5.8.1 the single-number v-strings (like C<v65>)
8fa72689 559are not v-strings before the C<< => >> operator (which is usually used
3921068c 560to separate a hash key from a hash value); instead they are interpreted
15ecd4ae
JH
561as literal strings ('v65'). They were v-strings from Perl 5.6.0 to
562Perl 5.8.0, but that caused more confusion and breakage than good.
563Multi-number v-strings like C<v65.66> and C<65.66.67> continue to
564be v-strings always.
d32a65d2 565
692ef166 566=head3 Special Literals
d74e8afc
ITB
567X<special literal> X<__END__> X<__DATA__> X<END> X<DATA>
568X<end> X<data> X<^D> X<^Z>
692ef166 569
d55a8828 570The special literals __FILE__, __LINE__, and __PACKAGE__
68dc0745 571represent the current filename, line number, and package name at that
84ed0108
FC
572point in your program. __SUB__ gives a reference to the current
573subroutine. They may be used only as separate tokens; they
68dc0745 574will not be interpolated into strings. If there is no current package
3e92a254 575(due to an empty C<package;> directive), __PACKAGE__ is the undefined
8fdd8881 576value. (But the empty C<package;> is no longer supported, as of version
84ed0108
FC
5775.10.) Outside of a subroutine, __SUB__ is the undefined value. __SUB__
578is only available in 5.16 or higher, and only with a C<use v5.16> or
579C<use feature "current_sub"> declaration.
580X<__FILE__> X<__LINE__> X<__PACKAGE__> X<__SUB__>
581X<line> X<file> X<package>
3e92a254
GS
582
583The two control characters ^D and ^Z, and the tokens __END__ and __DATA__
584may be used to indicate the logical end of the script before the actual
585end of file. Any following text is ignored.
586
1bab44f9 587Text after __DATA__ may be read via the filehandle C<PACKNAME::DATA>,
3e92a254
GS
588where C<PACKNAME> is the package that was current when the __DATA__
589token was encountered. The filehandle is left open pointing to the
4d383607 590line after __DATA__. The program should C<close DATA> when it is done
9c205800
FC
591reading from it. (Leaving it open leaks filehandles if the module is
592reloaded for any reason, so it's a safer practice to close it.) For
4d383607
JK
593compatibility with older scripts written before __DATA__ was
594introduced, __END__ behaves like __DATA__ in the top level script (but
595not in files loaded with C<require> or C<do>) and leaves the remaining
596contents of the file accessible via C<main::DATA>.
3e92a254
GS
597
598See L<SelfLoader> for more description of __DATA__, and
d55a8828
TC
599an example of its use. Note that you cannot read from the DATA
600filehandle in a BEGIN block: the BEGIN block is executed as soon
601as it is seen (during compilation), at which point the corresponding
a00c1fe5 602__DATA__ (or __END__) token has not yet been seen.
a0d0e21e 603
692ef166 604=head3 Barewords
d74e8afc 605X<bareword>
692ef166 606
748a9306 607A word that has no other interpretation in the grammar will
a0d0e21e
LW
608be treated as if it were a quoted string. These are known as
609"barewords". As with filehandles and labels, a bareword that consists
610entirely of lowercase letters risks conflict with future reserved
9f1b1f2d 611words, and if you use the C<use warnings> pragma or the B<-w> switch,
05b4f1ec
FW
612Perl will warn you about any such words. Perl limits barewords (like
613identifiers) to about 250 characters. Future versions of Perl are likely
614to eliminate these arbitrary limitations.
615
616Some people may wish to outlaw barewords entirely. If you
a0d0e21e
LW
617say
618
619 use strict 'subs';
620
621then any bareword that would NOT be interpreted as a subroutine call
622produces a compile-time error instead. The restriction lasts to the
54310121 623end of the enclosing block. An inner block may countermand this
a0d0e21e
LW
624by saying C<no strict 'subs'>.
625
e2b457c0 626=head3 Array Interpolation
d74e8afc 627X<array, interpolation> X<interpolation, array> X<$">
692ef166 628
d55a8828
TC
629Arrays and slices are interpolated into double-quoted strings
630by joining the elements with the delimiter specified in the C<$">
692ef166
SF
631variable (C<$LIST_SEPARATOR> if "use English;" is specified),
632space by default. The following are equivalent:
a0d0e21e 633
84f709e7 634 $temp = join($", @ARGV);
a0d0e21e
LW
635 system "echo $temp";
636
637 system "echo @ARGV";
638
639Within search patterns (which also undergo double-quotish substitution)
d55a8828 640there is an unfortunate ambiguity: Is C</$foo[bar]/> to be interpreted as
a0d0e21e
LW
641C</${foo}[bar]/> (where C<[bar]> is a character class for the regular
642expression) or as C</${foo[bar]}/> (where C<[bar]> is the subscript to array
643@foo)? If @foo doesn't otherwise exist, then it's obviously a
644character class. If @foo exists, Perl takes a good guess about C<[bar]>,
645and is almost always right. If it does guess wrong, or if you're just
646plain paranoid, you can force the correct interpretation with curly
d55a8828 647braces as above.
a0d0e21e 648
7e3b091d 649If you're looking for the information on how to use here-documents,
210b36aa
AMS
650which used to be here, that's been moved to
651L<perlop/Quote and Quote-like Operators>.
be16fac9 652
a0d0e21e 653=head2 List value constructors
d74e8afc 654X<list>
a0d0e21e
LW
655
656List values are denoted by separating individual values by commas
657(and enclosing the list in parentheses where precedence requires it):
658
659 (LIST)
660
d55a8828
TC
661In a context not requiring a list value, the value of what appears
662to be a list literal is simply the value of the final element, as
663with the C comma operator. For example,
a0d0e21e 664
84f709e7 665 @foo = ('cc', '-E', $bar);
a0d0e21e 666
d55a8828 667assigns the entire list value to array @foo, but
a0d0e21e 668
84f709e7 669 $foo = ('cc', '-E', $bar);
a0d0e21e 670
d55a8828
TC
671assigns the value of variable $bar to the scalar variable $foo.
672Note that the value of an actual array in scalar context is the
673length of the array; the following assigns the value 3 to $foo:
a0d0e21e 674
84f709e7 675 @foo = ('cc', '-E', $bar);
7e3b091d 676 $foo = @foo; # $foo gets 3
a0d0e21e 677
54310121 678You may have an optional comma before the closing parenthesis of a
a0d0e21e
LW
679list literal, so that you can say:
680
84f709e7 681 @foo = (
7e3b091d
DA
682 1,
683 2,
684 3,
a0d0e21e
LW
685 );
686
d55a8828
TC
687To use a here-document to assign an array, one line per element,
688you might use an approach like this:
689
84f709e7 690 @sauces = <<End_Lines =~ m/(\S.*\S)/g;
7e3b091d
DA
691 normal tomato
692 spicy tomato
693 green chile
694 pesto
695 white wine
d55a8828
TC
696 End_Lines
697
a0d0e21e 698LISTs do automatic interpolation of sublists. That is, when a LIST is
d55a8828 699evaluated, each element of the list is evaluated in list context, and
a0d0e21e 700the resulting list value is interpolated into LIST just as if each
5a964f20 701individual element were a member of LIST. Thus arrays and hashes lose their
a0d0e21e
LW
702identity in a LIST--the list
703
5a964f20 704 (@foo,@bar,&SomeSub,%glarch)
a0d0e21e
LW
705
706contains all the elements of @foo followed by all the elements of @bar,
5a964f20 707followed by all the elements returned by the subroutine named SomeSub
d55a8828 708called in list context, followed by the key/value pairs of %glarch.
a0d0e21e
LW
709To make a list reference that does I<NOT> interpolate, see L<perlref>.
710
19799a22 711The null list is represented by (). Interpolating it in a list
a0d0e21e
LW
712has no effect. Thus ((),(),()) is equivalent to (). Similarly,
713interpolating an array with no elements is the same as if no
714array had been interpolated at that point.
715
c2689353 716This interpolation combines with the facts that the opening
ab1f959b 717and closing parentheses are optional (except when necessary for
c2689353 718precedence) and lists may end with an optional comma to mean that
8fdd8881 719multiple commas within lists are legal syntax. The list C<1,,3> is a
c2689353
NC
720concatenation of two lists, C<1,> and C<3>, the first of which ends
721with that optional comma. C<1,,3> is C<(1,),(3)> is C<1,3> (And
722similarly for C<1,,,3> is C<(1,),(,),3> is C<1,3> and so on.) Not that
723we'd advise you to use this obfuscation.
724
a0d0e21e 725A list value may also be subscripted like a normal array. You must
54310121 726put the list in parentheses to avoid ambiguity. For example:
a0d0e21e
LW
727
728 # Stat returns list value.
84f709e7 729 $time = (stat($file))[8];
a0d0e21e 730
4633a7c4 731 # SYNTAX ERROR HERE.
84f709e7 732 $time = stat($file)[8]; # OOPS, FORGOT PARENTHESES
4633a7c4 733
a0d0e21e 734 # Find a hex digit.
84f709e7 735 $hexdigit = ('a','b','c','d','e','f')[$digit-10];
a0d0e21e
LW
736
737 # A "reverse comma operator".
738 return (pop(@foo),pop(@foo))[0];
739
d55a8828
TC
740Lists may be assigned to only when each element of the list
741is itself legal to assign to:
a0d0e21e 742
84f709e7 743 ($a, $b, $c) = (1, 2, 3);
a0d0e21e 744
84f709e7 745 ($map{'red'}, $map{'blue'}, $map{'green'}) = (0x00f, 0x0f0, 0xf00);
a0d0e21e 746
d55a8828
TC
747An exception to this is that you may assign to C<undef> in a list.
748This is useful for throwing away some of the return values of a
749function:
750
84f709e7 751 ($dev, $ino, undef, undef, $uid, $gid) = stat($file);
d55a8828 752
e1817ab9
FC
753As of Perl 5.22, you can also use C<(undef)x2> instead of C<undef, undef>.
754(You can also do C<($x) x 2>, which is less useful, because it assigns to
755the same variable twice, clobbering the first value assigned.)
756
d55a8828 757List assignment in scalar context returns the number of elements
4633a7c4
LW
758produced by the expression on the right side of the assignment:
759
7e3b091d
DA
760 $x = (($foo,$bar) = (3,2,1)); # set $x to 3, not 2
761 $x = (($foo,$bar) = f()); # set $x to f()'s return count
4633a7c4 762
d55a8828 763This is handy when you want to do a list assignment in a Boolean
19799a22 764context, because most list functions return a null list when finished,
4633a7c4
LW
765which when assigned produces a 0, which is interpreted as FALSE.
766
ab1f959b
PN
767It's also the source of a useful idiom for executing a function or
768performing an operation in list context and then counting the number of
769return values, by assigning to an empty list and then using that
8fdd8881 770assignment in scalar context. For example, this code:
ab1f959b 771
84f709e7 772 $count = () = $string =~ /\d+/g;
ab1f959b
PN
773
774will place into $count the number of digit groups found in $string.
775This happens because the pattern match is in list context (since it
776is being assigned to the empty list), and will therefore return a list
8fdd8881 777of all matching parts of the string. The list assignment in scalar
ab1f959b 778context will translate that into the number of elements (here, the
8fdd8881 779number of times the pattern matched) and assign that to $count. Note
ab1f959b
PN
780that simply using
781
84f709e7 782 $count = $string =~ /\d+/g;
ab1f959b
PN
783
784would not have worked, since a pattern match in scalar context will
785only return true or false, rather than a count of matches.
786
787The final element of a list assignment may be an array or a hash:
a0d0e21e 788
84f709e7 789 ($a, $b, @rest) = split;
5a964f20 790 my($a, $b, %rest) = @_;
a0d0e21e 791
4633a7c4 792You can actually put an array or hash anywhere in the list, but the first one
d55a8828
TC
793in the list will soak up all the values, and anything after it will become
794undefined. This may be useful in a my() or local().
a0d0e21e 795
d55a8828
TC
796A hash can be initialized using a literal list holding pairs of
797items to be interpreted as a key and a value:
a0d0e21e
LW
798
799 # same as map assignment above
84f709e7 800 %map = ('red',0x00f,'blue',0x0f0,'green',0xf00);
a0d0e21e 801
d55a8828 802While literal lists and named arrays are often interchangeable, that's
4633a7c4
LW
803not the case for hashes. Just because you can subscript a list value like
804a normal array does not mean that you can subscript a list value as a
805hash. Likewise, hashes included as parts of other lists (including
806parameters lists and return lists from functions) always flatten out into
807key/value pairs. That's why it's good to use references sometimes.
a0d0e21e 808
c47ff5f1
GS
809It is often more readable to use the C<< => >> operator between key/value
810pairs. The C<< => >> operator is mostly just a more visually distinctive
b88cefa9 811synonym for a comma, but it also arranges for its left-hand operand to be
ac036724 812interpreted as a string if it's a bareword that would be a legal simple
8fdd8881
FC
813identifier. C<< => >> doesn't quote compound identifiers, that contain
814double colons. This makes it nice for initializing hashes:
a0d0e21e 815
84f709e7 816 %map = (
7e3b091d
DA
817 red => 0x00f,
818 blue => 0x0f0,
819 green => 0xf00,
4633a7c4
LW
820 );
821
822or for initializing hash references to be used as records:
823
84f709e7 824 $rec = {
7e3b091d
DA
825 witch => 'Mable the Merciless',
826 cat => 'Fluffy the Ferocious',
827 date => '10/31/1776',
4633a7c4
LW
828 };
829
830or for using call-by-named-parameter to complicated functions:
831
84f709e7 832 $field = $query->radio_group(
7e3b091d 833 name => 'group_name',
4633a7c4
LW
834 values => ['eenie','meenie','minie'],
835 default => 'meenie',
836 linebreak => 'true',
84f709e7 837 labels => \%labels
4633a7c4 838 );
cb1a09d0
AD
839
840Note that just because a hash is initialized in that order doesn't
841mean that it comes out in that order. See L<perlfunc/sort> for examples
842of how to arrange for an output ordering.
843
c9e3649f
LM
844If a key appears more than once in the initializer list of a hash, the last
845occurrence wins:
846
847 %circle = (
848 center => [5, 10],
849 center => [27, 9],
850 radius => 100,
851 color => [0xDF, 0xFF, 0x00],
852 radius => 54,
853 );
854
855 # same as
856 %circle = (
857 center => [27, 9],
858 color => [0xDF, 0xFF, 0x00],
859 radius => 54,
860 );
861
862This can be used to provide overridable configuration defaults:
863
864 # values in %args take priority over %config_defaults
865 %config = (%config_defaults, %args);
866
692ef166
SF
867=head2 Subscripts
868
aa80e1dc
FC
869An array can be accessed one scalar at a
870time by specifying a dollar sign (C<$>), then the
692ef166
SF
871name of the array (without the leading C<@>), then the subscript inside
872square brackets. For example:
873
874 @myarray = (5, 50, 500, 5000);
2adc35dd 875 print "The Third Element is", $myarray[2], "\n";
692ef166 876
8fdd8881 877The array indices start with 0. A negative subscript retrieves its
692ef166
SF
878value from the end. In our example, C<$myarray[-1]> would have been
8795000, and C<$myarray[-2]> would have been 500.
880
881Hash subscripts are similar, only instead of square brackets curly brackets
8fdd8881 882are used. For example:
692ef166
SF
883
884 %scientists =
885 (
886 "Newton" => "Isaac",
887 "Einstein" => "Albert",
888 "Darwin" => "Charles",
889 "Feynman" => "Richard",
890 );
891
892 print "Darwin's First Name is ", $scientists{"Darwin"}, "\n";
893
aa80e1dc 894You can also subscript a list to get a single element from it:
d55a8828 895
aa80e1dc 896 $dir = (getpwnam("daemon"))[7];
d55a8828 897
9ed2a148
IG
898=head2 Multi-dimensional array emulation
899
900Multidimensional arrays may be emulated by subscripting a hash with a
8fdd8881 901list. The elements of the list are joined with the subscript separator
b8db74f2 902(see L<perlvar/$;>).
9ed2a148
IG
903
904 $foo{$a,$b,$c}
905
906is equivalent to
907
908 $foo{join($;, $a, $b, $c)}
909
910The default subscript separator is "\034", the same as SUBSEP in B<awk>.
911
aa80e1dc
FC
912=head2 Slices
913X<slice> X<array, slice> X<hash, slice>
d55a8828
TC
914
915A slice accesses several elements of a list, an array, or a hash
56d7751a
GS
916simultaneously using a list of subscripts. It's more convenient
917than writing out the individual elements as a list of separate
d55a8828
TC
918scalar values.
919
7e3b091d
DA
920 ($him, $her) = @folks[0,-1]; # array slice
921 @them = @folks[0 .. 3]; # array slice
922 ($who, $home) = @ENV{"USER", "HOME"}; # hash slice
923 ($uid, $dir) = (getpwnam("daemon"))[2,7]; # list slice
d55a8828
TC
924
925Since you can assign to a list of variables, you can also assign to
926an array or hash slice.
927
84f709e7 928 @days[3..5] = qw/Wed Thu Fri/;
d55a8828 929 @colors{'red','blue','green'}
7e3b091d 930 = (0xff0000, 0x0000ff, 0x00ff00);
d55a8828
TC
931 @folks[0, -1] = @folks[-1, 0];
932
933The previous assignments are exactly equivalent to
934
84f709e7
JH
935 ($days[3], $days[4], $days[5]) = qw/Wed Thu Fri/;
936 ($colors{'red'}, $colors{'blue'}, $colors{'green'})
7e3b091d 937 = (0xff0000, 0x0000ff, 0x00ff00);
88fd19e3 938 ($folks[0], $folks[-1]) = ($folks[-1], $folks[0]);
d55a8828
TC
939
940Since changing a slice changes the original array or hash that it's
56d7751a
GS
941slicing, a C<foreach> construct will alter some--or even all--of the
942values of the array or hash.
d55a8828
TC
943
944 foreach (@array[ 4 .. 10 ]) { s/peter/paul/ }
945
00cb5da1 946 foreach (@hash{qw[key1 key2]}) {
7e3b091d
DA
947 s/^\s+//; # trim leading whitespace
948 s/\s+$//; # trim trailing whitespace
949 s/(\w+)/\u\L$1/g; # "titlecase" words
d55a8828
TC
950 }
951
e2ec1b05
AP
952As a special exception, when you slice a list (but not an array or a hash),
953if the list evaluates to empty, then taking a slice of that empty list will
954always yield the empty list in turn. Thus:
08cd8952 955
e2ec1b05
AP
956 @a = ()[0,1]; # @a has no elements
957 @b = (@a)[0,1]; # @b has no elements
958 @c = (sub{}->())[0,1]; # @c has no elements
959 @d = ('a','b')[0,1]; # @d has two elements
960 @e = (@d)[0,1,8,9]; # @e has four elements
961 @f = (@d)[8,9]; # @f has two elements
f51152ef 962
19799a22
GS
963This makes it easy to write loops that terminate when a null list
964is returned:
d55a8828 965
e2ec1b05 966 while ( ($home, $user) = (getpwent)[7,0] ) {
7e3b091d 967 printf "%-8s %s\n", $user, $home;
d55a8828
TC
968 }
969
970As noted earlier in this document, the scalar sense of list assignment
971is the number of elements on the right-hand side of the assignment.
19799a22 972The null list contains no elements, so when the password file is
d55a8828
TC
973exhausted, the result is 0, not 2.
974
ad1de9c6
ML
975Slices in scalar context return the last item of the slice.
976
977 @a = qw/first second third/;
978 %h = (first => 'A', second => 'B');
979 $t = @a[0, 1]; # $t is now 'second'
0de10106 980 $u = @h{'first', 'second'}; # $u is now 'B'
ad1de9c6 981
d55a8828
TC
982If you're confused about why you use an '@' there on a hash slice
983instead of a '%', think of it like this. The type of bracket (square
984or curly) governs whether it's an array or a hash being looked at.
985On the other hand, the leading symbol ('$' or '@') on the array or
986hash indicates whether you are getting back a singular value (a
987scalar) or a plural one (a list).
988
8a7ab7dc 989=head3 Key/Value Hash Slices
23a22365 990
c44d7536
FC
991Starting in Perl 5.20, a hash slice operation
992with the % symbol is a variant of slice operation
190c3990 993returning a list of key/value pairs rather than just values:
23a22365 994
190c3990
FC
995 %h = (blonk => 2, foo => 3, squink => 5, bar => 8);
996 %subset = %h{'foo', 'bar'}; # key/value hash slice
997 # %subset is now (foo => 3, bar => 8)
23a22365 998
190c3990
FC
999However, the result of such a slice cannot be localized, deleted or used
1000in assignment. These are otherwise very much consistent with hash slices
1001using the @ symbol.
23a22365 1002
8a7ab7dc 1003=head3 Index/Value Array Slices
23a22365 1004
c44d7536
FC
1005Similar to key/value hash slices (and also introduced
1006in Perl 5.20), the % array slice syntax returns a list
190c3990 1007of index/value pairs:
23a22365 1008
190c3990
FC
1009 @a = "a".."z";
1010 @list = %a[3,4,6];
1011 # @list is now (3, "d", 4, "e", 6, "g")
23a22365 1012
5f05dabc 1013=head2 Typeglobs and Filehandles
d74e8afc 1014X<typeglob> X<filehandle> X<*>
cb1a09d0
AD
1015
1016Perl uses an internal type called a I<typeglob> to hold an entire
1017symbol table entry. The type prefix of a typeglob is a C<*>, because
54310121 1018it represents all types. This used to be the preferred way to
cb1a09d0 1019pass arrays and hashes by reference into a function, but now that
5a964f20
TC
1020we have real references, this is seldom needed.
1021
1022The main use of typeglobs in modern Perl is create symbol table aliases.
1023This assignment:
1024
1025 *this = *that;
1026
1027makes $this an alias for $that, @this an alias for @that, %this an alias
1028for %that, &this an alias for &that, etc. Much safer is to use a reference.
1029This:
5f05dabc 1030
5a964f20
TC
1031 local *Here::blue = \$There::green;
1032
1033temporarily makes $Here::blue an alias for $There::green, but doesn't
1034make @Here::blue an alias for @There::green, or %Here::blue an alias for
1035%There::green, etc. See L<perlmod/"Symbol Tables"> for more examples
1036of this. Strange though this may seem, this is the basis for the whole
84f709e7 1037module import/export system.
5a964f20 1038
d55a8828 1039Another use for typeglobs is to pass filehandles into a function or
5a964f20
TC
1040to create new filehandles. If you need to use a typeglob to save away
1041a filehandle, do it this way:
5f05dabc 1042
84f709e7 1043 $fh = *STDOUT;
5f05dabc
PP
1044
1045or perhaps as a real reference, like this:
1046
84f709e7 1047 $fh = \*STDOUT;
5f05dabc 1048
5a964f20
TC
1049See L<perlsub> for examples of using these as indirect filehandles
1050in functions.
1051
1052Typeglobs are also a way to create a local filehandle using the local()
1053operator. These last until their block is exited, but may be passed back.
1054For example:
5f05dabc
PP
1055
1056 sub newopen {
7e3b091d
DA
1057 my $path = shift;
1058 local *FH; # not my!
1059 open (FH, $path) or return undef;
1060 return *FH;
5f05dabc 1061 }
84f709e7 1062 $fh = newopen('/etc/passwd');
5f05dabc 1063
d55a8828 1064Now that we have the C<*foo{THING}> notation, typeglobs aren't used as much
5a964f20 1065for filehandle manipulations, although they're still needed to pass brand
8fdd8881 1066new file and directory handles into or out of functions. That's because
d55a8828
TC
1067C<*HANDLE{IO}> only works if HANDLE has already been used as a handle.
1068In other words, C<*FH> must be used to create new symbol table entries;
1069C<*foo{THING}> cannot. When in doubt, use C<*FH>.
1070
36392fcf
GS
1071All functions that are capable of creating filehandles (open(),
1072opendir(), pipe(), socketpair(), sysopen(), socket(), and accept())
1073automatically create an anonymous filehandle if the handle passed to
8fdd8881 1074them is an uninitialized scalar variable. This allows the constructs
36392fcf
GS
1075such as C<open(my $fh, ...)> and C<open(local $fh,...)> to be used to
1076create filehandles that will conveniently be closed automatically when
8fdd8881 1077the scope ends, provided there are no other references to them. This
36392fcf
GS
1078largely eliminates the need for typeglobs when opening filehandles
1079that must be passed around, as in the following example:
1080
1081 sub myopen {
84f709e7 1082 open my $fh, "@_"
7e3b091d
DA
1083 or die "Can't open '@_': $!";
1084 return $fh;
36392fcf
GS
1085 }
1086
1087 {
1088 my $f = myopen("</etc/motd");
7e3b091d
DA
1089 print <$f>;
1090 # $f implicitly closed here
36392fcf
GS
1091 }
1092
b92795fe
AMS
1093Note that if an initialized scalar variable is used instead the
1094result is different: C<my $fh='zzz'; open($fh, ...)> is equivalent
1095to C<open( *{'zzz'}, ...)>.
d83fe814
AT
1096C<use strict 'refs'> forbids such practice.
1097
d55a8828
TC
1098Another way to create anonymous filehandles is with the Symbol
1099module or with the IO::Handle module and its ilk. These modules
1100have the advantage of not hiding different types of the same name
66b6e4ad
KW
1101during the local(). See the bottom of L<perlfunc/open> for an
1102example.
d55a8828
TC
1103
1104=head1 SEE ALSO
1105
1106See L<perlvar> for a description of Perl's built-in variables and
1107a discussion of legal variable names. See L<perlref>, L<perlsub>,
1108and L<perlmod/"Symbol Tables"> for more discussion on typeglobs and
1109the C<*foo{THING}> syntax.