This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Clarify the description of SCALAR in perltie
[perl5.git] / pod / perltie.pod
CommitLineData
cb1a09d0
AD
1=head1 NAME
2
3perltie - how to hide an object class in a simple variable
4
5=head1 SYNOPSIS
6
7 tie VARIABLE, CLASSNAME, LIST
8
6fdf61fb
PP
9 $object = tied VARIABLE
10
cb1a09d0
AD
11 untie VARIABLE
12
13=head1 DESCRIPTION
14
15Prior to release 5.0 of Perl, a programmer could use dbmopen()
5f05dabc
PP
16to connect an on-disk database in the standard Unix dbm(3x)
17format magically to a %HASH in their program. However, their Perl was either
cb1a09d0
AD
18built with one particular dbm library or another, but not both, and
19you couldn't extend this mechanism to other packages or types of variables.
20
21Now you can.
22
23The tie() function binds a variable to a class (package) that will provide
24the implementation for access methods for that variable. Once this magic
25has been performed, accessing a tied variable automatically triggers
5a964f20 26method calls in the proper class. The complexity of the class is
cb1a09d0
AD
27hidden behind magic methods calls. The method names are in ALL CAPS,
28which is a convention that Perl uses to indicate that they're called
29implicitly rather than explicitly--just like the BEGIN() and END()
30functions.
31
32In the tie() call, C<VARIABLE> is the name of the variable to be
33enchanted. C<CLASSNAME> is the name of a class implementing objects of
34the correct type. Any additional arguments in the C<LIST> are passed to
35the appropriate constructor method for that class--meaning TIESCALAR(),
5f05dabc 36TIEARRAY(), TIEHASH(), or TIEHANDLE(). (Typically these are arguments
a7adf1f0
PP
37such as might be passed to the dbminit() function of C.) The object
38returned by the "new" method is also returned by the tie() function,
39which would be useful if you wanted to access other methods in
40C<CLASSNAME>. (You don't actually have to return a reference to a right
5f05dabc 41"type" (e.g., HASH or C<CLASSNAME>) so long as it's a properly blessed
a7adf1f0
PP
42object.) You can also retrieve a reference to the underlying object
43using the tied() function.
cb1a09d0
AD
44
45Unlike dbmopen(), the tie() function will not C<use> or C<require> a module
46for you--you need to do that explicitly yourself.
47
48=head2 Tying Scalars
49
50A class implementing a tied scalar should define the following methods:
301e8125 51TIESCALAR, FETCH, STORE, and possibly UNTIE and/or DESTROY.
cb1a09d0
AD
52
53Let's look at each in turn, using as an example a tie class for
54scalars that allows the user to do something like:
55
56 tie $his_speed, 'Nice', getppid();
57 tie $my_speed, 'Nice', $$;
58
59And now whenever either of those variables is accessed, its current
60system priority is retrieved and returned. If those variables are set,
61then the process's priority is changed!
62
5aabfad6
PP
63We'll use Jarkko Hietaniemi <F<jhi@iki.fi>>'s BSD::Resource class (not
64included) to access the PRIO_PROCESS, PRIO_MIN, and PRIO_MAX constants
65from your system, as well as the getpriority() and setpriority() system
66calls. Here's the preamble of the class.
cb1a09d0
AD
67
68 package Nice;
69 use Carp;
70 use BSD::Resource;
71 use strict;
72 $Nice::DEBUG = 0 unless defined $Nice::DEBUG;
73
13a2d996 74=over 4
cb1a09d0
AD
75
76=item TIESCALAR classname, LIST
77
78This is the constructor for the class. That means it is
79expected to return a blessed reference to a new scalar
80(probably anonymous) that it's creating. For example:
81
82 sub TIESCALAR {
83 my $class = shift;
84 my $pid = shift || $$; # 0 means me
85
86 if ($pid !~ /^\d+$/) {
6fdf61fb 87 carp "Nice::Tie::Scalar got non-numeric pid $pid" if $^W;
cb1a09d0
AD
88 return undef;
89 }
90
91 unless (kill 0, $pid) { # EPERM or ERSCH, no doubt
6fdf61fb 92 carp "Nice::Tie::Scalar got bad pid $pid: $!" if $^W;
cb1a09d0
AD
93 return undef;
94 }
95
96 return bless \$pid, $class;
97 }
98
99This tie class has chosen to return an error rather than raising an
100exception if its constructor should fail. While this is how dbmopen() works,
101other classes may well not wish to be so forgiving. It checks the global
102variable C<$^W> to see whether to emit a bit of noise anyway.
103
104=item FETCH this
105
106This method will be triggered every time the tied variable is accessed
107(read). It takes no arguments beyond its self reference, which is the
5f05dabc
PP
108object representing the scalar we're dealing with. Because in this case
109we're using just a SCALAR ref for the tied scalar object, a simple $$self
cb1a09d0
AD
110allows the method to get at the real value stored there. In our example
111below, that real value is the process ID to which we've tied our variable.
112
113 sub FETCH {
114 my $self = shift;
115 confess "wrong type" unless ref $self;
116 croak "usage error" if @_;
117 my $nicety;
118 local($!) = 0;
119 $nicety = getpriority(PRIO_PROCESS, $$self);
120 if ($!) { croak "getpriority failed: $!" }
121 return $nicety;
122 }
123
124This time we've decided to blow up (raise an exception) if the renice
125fails--there's no place for us to return an error otherwise, and it's
126probably the right thing to do.
127
128=item STORE this, value
129
130This method will be triggered every time the tied variable is set
131(assigned). Beyond its self reference, it also expects one (and only one)
a177e38d
DN
132argument--the new value the user is trying to assign. Don't worry about
133returning a value from STORE -- the semantic of assignment returning the
134assigned value is implemented with FETCH.
cb1a09d0
AD
135
136 sub STORE {
137 my $self = shift;
138 confess "wrong type" unless ref $self;
139 my $new_nicety = shift;
140 croak "usage error" if @_;
141
142 if ($new_nicety < PRIO_MIN) {
143 carp sprintf
144 "WARNING: priority %d less than minimum system priority %d",
145 $new_nicety, PRIO_MIN if $^W;
146 $new_nicety = PRIO_MIN;
147 }
148
149 if ($new_nicety > PRIO_MAX) {
150 carp sprintf
151 "WARNING: priority %d greater than maximum system priority %d",
152 $new_nicety, PRIO_MAX if $^W;
153 $new_nicety = PRIO_MAX;
154 }
155
156 unless (defined setpriority(PRIO_PROCESS, $$self, $new_nicety)) {
157 confess "setpriority failed: $!";
158 }
cb1a09d0
AD
159 }
160
301e8125
NIS
161=item UNTIE this
162
163This method will be triggered when the C<untie> occurs. This can be useful
164if the class needs to know when no further calls will be made. (Except DESTROY
d5582e24 165of course.) See L<The C<untie> Gotcha> below for more details.
301e8125 166
cb1a09d0
AD
167=item DESTROY this
168
169This method will be triggered when the tied variable needs to be destructed.
5f05dabc 170As with other object classes, such a method is seldom necessary, because Perl
cb1a09d0
AD
171deallocates its moribund object's memory for you automatically--this isn't
172C++, you know. We'll use a DESTROY method here for debugging purposes only.
173
174 sub DESTROY {
175 my $self = shift;
176 confess "wrong type" unless ref $self;
177 carp "[ Nice::DESTROY pid $$self ]" if $Nice::DEBUG;
178 }
179
180=back
181
182That's about all there is to it. Actually, it's more than all there
5f05dabc 183is to it, because we've done a few nice things here for the sake
cb1a09d0
AD
184of completeness, robustness, and general aesthetics. Simpler
185TIESCALAR classes are certainly possible.
186
187=head2 Tying Arrays
188
189A class implementing a tied ordinary array should define the following
301e8125 190methods: TIEARRAY, FETCH, STORE, FETCHSIZE, STORESIZE and perhaps UNTIE and/or DESTROY.
cb1a09d0 191
a60c0954
NIS
192FETCHSIZE and STORESIZE are used to provide C<$#array> and
193equivalent C<scalar(@array)> access.
c47ff5f1 194
01020589
GS
195The methods POP, PUSH, SHIFT, UNSHIFT, SPLICE, DELETE, and EXISTS are
196required if the perl operator with the corresponding (but lowercase) name
197is to operate on the tied array. The B<Tie::Array> class can be used as a
198base class to implement the first five of these in terms of the basic
199methods above. The default implementations of DELETE and EXISTS in
200B<Tie::Array> simply C<croak>.
a60c0954 201
301e8125 202In addition EXTEND will be called when perl would have pre-extended
a60c0954
NIS
203allocation in a real array.
204
4ae85618
CT
205For this discussion, we'll implement an array whose elements are a fixed
206size at creation. If you try to create an element larger than the fixed
207size, you'll take an exception. For example:
cb1a09d0 208
4ae85618
CT
209 use FixedElem_Array;
210 tie @array, 'FixedElem_Array', 3;
211 $array[0] = 'cat'; # ok.
212 $array[1] = 'dogs'; # exception, length('dogs') > 3.
cb1a09d0
AD
213
214The preamble code for the class is as follows:
215
4ae85618 216 package FixedElem_Array;
cb1a09d0
AD
217 use Carp;
218 use strict;
219
13a2d996 220=over 4
cb1a09d0
AD
221
222=item TIEARRAY classname, LIST
223
224This is the constructor for the class. That means it is expected to
225return a blessed reference through which the new array (probably an
226anonymous ARRAY ref) will be accessed.
227
228In our example, just to show you that you don't I<really> have to return an
229ARRAY reference, we'll choose a HASH reference to represent our object.
4ae85618
CT
230A HASH works out well as a generic record type: the C<{ELEMSIZE}> field will
231store the maximum element size allowed, and the C<{ARRAY}> field will hold the
cb1a09d0
AD
232true ARRAY ref. If someone outside the class tries to dereference the
233object returned (doubtless thinking it an ARRAY ref), they'll blow up.
234This just goes to show you that you should respect an object's privacy.
235
236 sub TIEARRAY {
4ae85618
CT
237 my $class = shift;
238 my $elemsize = shift;
239 if ( @_ || $elemsize =~ /\D/ ) {
240 croak "usage: tie ARRAY, '" . __PACKAGE__ . "', elem_size";
241 }
242 return bless {
243 ELEMSIZE => $elemsize,
244 ARRAY => [],
245 }, $class;
cb1a09d0
AD
246 }
247
248=item FETCH this, index
249
250This method will be triggered every time an individual element the tied array
251is accessed (read). It takes one argument beyond its self reference: the
252index whose value we're trying to fetch.
253
254 sub FETCH {
4ae85618
CT
255 my $self = shift;
256 my $index = shift;
257 return $self->{ARRAY}->[$index];
cb1a09d0
AD
258 }
259
301e8125 260If a negative array index is used to read from an array, the index
0b931be4 261will be translated to a positive one internally by calling FETCHSIZE
6f12eb6d
MJD
262before being passed to FETCH. You may disable this feature by
263assigning a true value to the variable C<$NEGATIVE_INDICES> in the
264tied array class.
301e8125 265
cb1a09d0
AD
266As you may have noticed, the name of the FETCH method (et al.) is the same
267for all accesses, even though the constructors differ in names (TIESCALAR
268vs TIEARRAY). While in theory you could have the same class servicing
269several tied types, in practice this becomes cumbersome, and it's easiest
5f05dabc 270to keep them at simply one tie type per class.
cb1a09d0
AD
271
272=item STORE this, index, value
273
274This method will be triggered every time an element in the tied array is set
275(written). It takes two arguments beyond its self reference: the index at
276which we're trying to store something and the value we're trying to put
4ae85618
CT
277there.
278
279In our example, C<undef> is really C<$self-E<gt>{ELEMSIZE}> number of
280spaces so we have a little more work to do here:
cb1a09d0
AD
281
282 sub STORE {
4ae85618
CT
283 my $self = shift;
284 my( $index, $value ) = @_;
285 if ( length $value > $self->{ELEMSIZE} ) {
286 croak "length of $value is greater than $self->{ELEMSIZE}";
cb1a09d0 287 }
4ae85618
CT
288 # fill in the blanks
289 $self->EXTEND( $index ) if $index > $self->FETCHSIZE();
290 # right justify to keep element size for smaller elements
291 $self->{ARRAY}->[$index] = sprintf "%$self->{ELEMSIZE}s", $value;
cb1a09d0 292 }
301e8125
NIS
293
294Negative indexes are treated the same as with FETCH.
295
4ae85618
CT
296=item FETCHSIZE this
297
298Returns the total number of items in the tied array associated with
299object I<this>. (Equivalent to C<scalar(@array)>). For example:
300
301 sub FETCHSIZE {
302 my $self = shift;
303 return scalar @{$self->{ARRAY}};
304 }
305
306=item STORESIZE this, count
307
308Sets the total number of items in the tied array associated with
309object I<this> to be I<count>. If this makes the array larger then
310class's mapping of C<undef> should be returned for new positions.
311If the array becomes smaller then entries beyond count should be
312deleted.
313
314In our example, 'undef' is really an element containing
315C<$self-E<gt>{ELEMSIZE}> number of spaces. Observe:
316
f9abed49
CT
317 sub STORESIZE {
318 my $self = shift;
319 my $count = shift;
320 if ( $count > $self->FETCHSIZE() ) {
321 foreach ( $count - $self->FETCHSIZE() .. $count ) {
322 $self->STORE( $_, '' );
323 }
324 } elsif ( $count < $self->FETCHSIZE() ) {
325 foreach ( 0 .. $self->FETCHSIZE() - $count - 2 ) {
326 $self->POP();
327 }
328 }
329 }
4ae85618
CT
330
331=item EXTEND this, count
332
333Informative call that array is likely to grow to have I<count> entries.
334Can be used to optimize allocation. This method need do nothing.
335
336In our example, we want to make sure there are no blank (C<undef>)
337entries, so C<EXTEND> will make use of C<STORESIZE> to fill elements
338as needed:
339
340 sub EXTEND {
341 my $self = shift;
342 my $count = shift;
343 $self->STORESIZE( $count );
344 }
345
346=item EXISTS this, key
347
348Verify that the element at index I<key> exists in the tied array I<this>.
349
350In our example, we will determine that if an element consists of
351C<$self-E<gt>{ELEMSIZE}> spaces only, it does not exist:
352
353 sub EXISTS {
354 my $self = shift;
355 my $index = shift;
f9abed49
CT
356 return 0 if ! defined $self->{ARRAY}->[$index] ||
357 $self->{ARRAY}->[$index] eq ' ' x $self->{ELEMSIZE};
358 return 1;
4ae85618
CT
359 }
360
361=item DELETE this, key
362
363Delete the element at index I<key> from the tied array I<this>.
364
ad0f383a 365In our example, a deleted item is C<$self-E<gt>{ELEMSIZE}> spaces:
4ae85618
CT
366
367 sub DELETE {
368 my $self = shift;
369 my $index = shift;
370 return $self->STORE( $index, '' );
371 }
372
373=item CLEAR this
374
375Clear (remove, delete, ...) all values from the tied array associated with
376object I<this>. For example:
377
378 sub CLEAR {
379 my $self = shift;
380 return $self->{ARRAY} = [];
381 }
382
383=item PUSH this, LIST
384
385Append elements of I<LIST> to the array. For example:
386
387 sub PUSH {
388 my $self = shift;
389 my @list = @_;
390 my $last = $self->FETCHSIZE();
391 $self->STORE( $last + $_, $list[$_] ) foreach 0 .. $#list;
392 return $self->FETCHSIZE();
393 }
394
395=item POP this
396
397Remove last element of the array and return it. For example:
398
399 sub POP {
400 my $self = shift;
401 return pop @{$self->{ARRAY}};
402 }
403
404=item SHIFT this
405
406Remove the first element of the array (shifting other elements down)
407and return it. For example:
408
409 sub SHIFT {
410 my $self = shift;
411 return shift @{$self->{ARRAY}};
412 }
413
414=item UNSHIFT this, LIST
415
416Insert LIST elements at the beginning of the array, moving existing elements
417up to make room. For example:
418
419 sub UNSHIFT {
420 my $self = shift;
421 my @list = @_;
422 my $size = scalar( @list );
423 # make room for our list
424 @{$self->{ARRAY}}[ $size .. $#{$self->{ARRAY}} + $size ]
425 = @{$self->{ARRAY}};
426 $self->STORE( $_, $list[$_] ) foreach 0 .. $#list;
427 }
428
429=item SPLICE this, offset, length, LIST
430
431Perform the equivalent of C<splice> on the array.
432
433I<offset> is optional and defaults to zero, negative values count back
434from the end of the array.
435
436I<length> is optional and defaults to rest of the array.
437
438I<LIST> may be empty.
439
440Returns a list of the original I<length> elements at I<offset>.
441
442In our example, we'll use a little shortcut if there is a I<LIST>:
443
444 sub SPLICE {
445 my $self = shift;
446 my $offset = shift || 0;
447 my $length = shift || $self->FETCHSIZE() - $offset;
448 my @list = ();
449 if ( @_ ) {
450 tie @list, __PACKAGE__, $self->{ELEMSIZE};
451 @list = @_;
452 }
453 return splice @{$self->{ARRAY}}, $offset, $length, @list;
454 }
455
301e8125
NIS
456=item UNTIE this
457
d5582e24 458Will be called when C<untie> happens. (See L<The C<untie> Gotcha> below.)
cb1a09d0
AD
459
460=item DESTROY this
461
462This method will be triggered when the tied variable needs to be destructed.
184e9718 463As with the scalar tie class, this is almost never needed in a
cb1a09d0
AD
464language that does its own garbage collection, so this time we'll
465just leave it out.
466
467=back
468
cb1a09d0
AD
469=head2 Tying Hashes
470
be3174d2
GS
471Hashes were the first Perl data type to be tied (see dbmopen()). A class
472implementing a tied hash should define the following methods: TIEHASH is
473the constructor. FETCH and STORE access the key and value pairs. EXISTS
474reports whether a key is present in the hash, and DELETE deletes one.
475CLEAR empties the hash by deleting all the key and value pairs. FIRSTKEY
476and NEXTKEY implement the keys() and each() functions to iterate over all
a3bcc51e
TP
477the keys. SCALAR is triggered when the tied hash is evaluated in scalar
478context. UNTIE is called when C<untie> happens, and DESTROY is called when
301e8125 479the tied variable is garbage collected.
aa689395
PP
480
481If this seems like a lot, then feel free to inherit from merely the
d5582e24 482standard Tie::StdHash module for most of your methods, redefining only the
aa689395 483interesting ones. See L<Tie::Hash> for details.
cb1a09d0
AD
484
485Remember that Perl distinguishes between a key not existing in the hash,
486and the key existing in the hash but having a corresponding value of
487C<undef>. The two possibilities can be tested with the C<exists()> and
488C<defined()> functions.
489
490Here's an example of a somewhat interesting tied hash class: it gives you
5f05dabc
PP
491a hash representing a particular user's dot files. You index into the hash
492with the name of the file (minus the dot) and you get back that dot file's
cb1a09d0
AD
493contents. For example:
494
495 use DotFiles;
1f57c600 496 tie %dot, 'DotFiles';
cb1a09d0
AD
497 if ( $dot{profile} =~ /MANPATH/ ||
498 $dot{login} =~ /MANPATH/ ||
499 $dot{cshrc} =~ /MANPATH/ )
500 {
5f05dabc 501 print "you seem to set your MANPATH\n";
cb1a09d0
AD
502 }
503
504Or here's another sample of using our tied class:
505
1f57c600 506 tie %him, 'DotFiles', 'daemon';
cb1a09d0
AD
507 foreach $f ( keys %him ) {
508 printf "daemon dot file %s is size %d\n",
509 $f, length $him{$f};
510 }
511
512In our tied hash DotFiles example, we use a regular
513hash for the object containing several important
514fields, of which only the C<{LIST}> field will be what the
515user thinks of as the real hash.
516
517=over 5
518
519=item USER
520
521whose dot files this object represents
522
523=item HOME
524
5f05dabc 525where those dot files live
cb1a09d0
AD
526
527=item CLOBBER
528
529whether we should try to change or remove those dot files
530
531=item LIST
532
5f05dabc 533the hash of dot file names and content mappings
cb1a09d0
AD
534
535=back
536
537Here's the start of F<Dotfiles.pm>:
538
539 package DotFiles;
540 use Carp;
541 sub whowasi { (caller(1))[3] . '()' }
542 my $DEBUG = 0;
543 sub debug { $DEBUG = @_ ? shift : 1 }
544
5f05dabc 545For our example, we want to be able to emit debugging info to help in tracing
cb1a09d0
AD
546during development. We keep also one convenience function around
547internally to help print out warnings; whowasi() returns the function name
548that calls it.
549
550Here are the methods for the DotFiles tied hash.
551
13a2d996 552=over 4
cb1a09d0
AD
553
554=item TIEHASH classname, LIST
555
556This is the constructor for the class. That means it is expected to
557return a blessed reference through which the new object (probably but not
558necessarily an anonymous hash) will be accessed.
559
560Here's the constructor:
561
562 sub TIEHASH {
563 my $self = shift;
564 my $user = shift || $>;
565 my $dotdir = shift || '';
566 croak "usage: @{[&whowasi]} [USER [DOTDIR]]" if @_;
567 $user = getpwuid($user) if $user =~ /^\d+$/;
568 my $dir = (getpwnam($user))[7]
569 || croak "@{[&whowasi]}: no user $user";
570 $dir .= "/$dotdir" if $dotdir;
571
572 my $node = {
573 USER => $user,
574 HOME => $dir,
575 LIST => {},
576 CLOBBER => 0,
577 };
578
579 opendir(DIR, $dir)
580 || croak "@{[&whowasi]}: can't opendir $dir: $!";
581 foreach $dot ( grep /^\./ && -f "$dir/$_", readdir(DIR)) {
582 $dot =~ s/^\.//;
583 $node->{LIST}{$dot} = undef;
584 }
585 closedir DIR;
586 return bless $node, $self;
587 }
588
589It's probably worth mentioning that if you're going to filetest the
590return values out of a readdir, you'd better prepend the directory
5f05dabc 591in question. Otherwise, because we didn't chdir() there, it would
2ae324a7 592have been testing the wrong file.
cb1a09d0
AD
593
594=item FETCH this, key
595
596This method will be triggered every time an element in the tied hash is
597accessed (read). It takes one argument beyond its self reference: the key
598whose value we're trying to fetch.
599
600Here's the fetch for our DotFiles example.
601
602 sub FETCH {
603 carp &whowasi if $DEBUG;
604 my $self = shift;
605 my $dot = shift;
606 my $dir = $self->{HOME};
607 my $file = "$dir/.$dot";
608
609 unless (exists $self->{LIST}->{$dot} || -f $file) {
610 carp "@{[&whowasi]}: no $dot file" if $DEBUG;
611 return undef;
612 }
613
614 if (defined $self->{LIST}->{$dot}) {
615 return $self->{LIST}->{$dot};
616 } else {
617 return $self->{LIST}->{$dot} = `cat $dir/.$dot`;
618 }
619 }
620
621It was easy to write by having it call the Unix cat(1) command, but it
622would probably be more portable to open the file manually (and somewhat
5f05dabc 623more efficient). Of course, because dot files are a Unixy concept, we're
cb1a09d0
AD
624not that concerned.
625
626=item STORE this, key, value
627
628This method will be triggered every time an element in the tied hash is set
629(written). It takes two arguments beyond its self reference: the index at
630which we're trying to store something, and the value we're trying to put
631there.
632
633Here in our DotFiles example, we'll be careful not to let
634them try to overwrite the file unless they've called the clobber()
635method on the original object reference returned by tie().
636
637 sub STORE {
638 carp &whowasi if $DEBUG;
639 my $self = shift;
640 my $dot = shift;
641 my $value = shift;
642 my $file = $self->{HOME} . "/.$dot";
643 my $user = $self->{USER};
644
645 croak "@{[&whowasi]}: $file not clobberable"
646 unless $self->{CLOBBER};
647
648 open(F, "> $file") || croak "can't open $file: $!";
649 print F $value;
650 close(F);
651 }
652
653If they wanted to clobber something, they might say:
654
655 $ob = tie %daemon_dots, 'daemon';
656 $ob->clobber(1);
657 $daemon_dots{signature} = "A true daemon\n";
658
6fdf61fb
PP
659Another way to lay hands on a reference to the underlying object is to
660use the tied() function, so they might alternately have set clobber
661using:
662
663 tie %daemon_dots, 'daemon';
664 tied(%daemon_dots)->clobber(1);
665
666The clobber method is simply:
cb1a09d0
AD
667
668 sub clobber {
669 my $self = shift;
670 $self->{CLOBBER} = @_ ? shift : 1;
671 }
672
673=item DELETE this, key
674
675This method is triggered when we remove an element from the hash,
676typically by using the delete() function. Again, we'll
677be careful to check whether they really want to clobber files.
678
679 sub DELETE {
680 carp &whowasi if $DEBUG;
681
682 my $self = shift;
683 my $dot = shift;
684 my $file = $self->{HOME} . "/.$dot";
685 croak "@{[&whowasi]}: won't remove file $file"
686 unless $self->{CLOBBER};
687 delete $self->{LIST}->{$dot};
1f57c600
PP
688 my $success = unlink($file);
689 carp "@{[&whowasi]}: can't unlink $file: $!" unless $success;
690 $success;
cb1a09d0
AD
691 }
692
1f57c600
PP
693The value returned by DELETE becomes the return value of the call
694to delete(). If you want to emulate the normal behavior of delete(),
695you should return whatever FETCH would have returned for this key.
696In this example, we have chosen instead to return a value which tells
697the caller whether the file was successfully deleted.
698
cb1a09d0
AD
699=item CLEAR this
700
701This method is triggered when the whole hash is to be cleared, usually by
702assigning the empty list to it.
703
5f05dabc 704In our example, that would remove all the user's dot files! It's such a
cb1a09d0
AD
705dangerous thing that they'll have to set CLOBBER to something higher than
7061 to make it happen.
707
708 sub CLEAR {
709 carp &whowasi if $DEBUG;
710 my $self = shift;
5f05dabc 711 croak "@{[&whowasi]}: won't remove all dot files for $self->{USER}"
cb1a09d0
AD
712 unless $self->{CLOBBER} > 1;
713 my $dot;
714 foreach $dot ( keys %{$self->{LIST}}) {
715 $self->DELETE($dot);
716 }
717 }
718
719=item EXISTS this, key
720
721This method is triggered when the user uses the exists() function
722on a particular hash. In our example, we'll look at the C<{LIST}>
723hash element for this:
724
725 sub EXISTS {
726 carp &whowasi if $DEBUG;
727 my $self = shift;
728 my $dot = shift;
729 return exists $self->{LIST}->{$dot};
730 }
731
732=item FIRSTKEY this
733
734This method will be triggered when the user is going
735to iterate through the hash, such as via a keys() or each()
736call.
737
738 sub FIRSTKEY {
739 carp &whowasi if $DEBUG;
740 my $self = shift;
6fdf61fb 741 my $a = keys %{$self->{LIST}}; # reset each() iterator
cb1a09d0
AD
742 each %{$self->{LIST}}
743 }
744
745=item NEXTKEY this, lastkey
746
747This method gets triggered during a keys() or each() iteration. It has a
748second argument which is the last key that had been accessed. This is
749useful if you're carrying about ordering or calling the iterator from more
750than one sequence, or not really storing things in a hash anywhere.
751
5f05dabc
PP
752For our example, we're using a real hash so we'll do just the simple
753thing, but we'll have to go through the LIST field indirectly.
cb1a09d0
AD
754
755 sub NEXTKEY {
756 carp &whowasi if $DEBUG;
757 my $self = shift;
758 return each %{ $self->{LIST} }
759 }
760
a3bcc51e
TP
761=item SCALAR this
762
763This is called when the hash is evaluated in scalar context. In order
764to mimic the behaviour of untied hashes, this method should return a
765false value when the tied hash is considered empty. If this method does
159b10bb
RGS
766not exist, perl will make some educated guesses and return true when
767the hash is inside an iteration. If this isn't the case, FIRSTKEY is
768called, and the result will be a false value if FIRSTKEY returns the empty
769list, true otherwise.
a3bcc51e
TP
770
771In our example we can just call C<scalar> on the underlying hash
772referenced by C<$self-E<gt>{LIST}>:
773
774 sub SCALAR {
775 carp &whowasi if $DEBUG;
776 my $self = shift;
777 return scalar %{ $self->{LIST} }
778 }
779
301e8125
NIS
780=item UNTIE this
781
d5582e24 782This is called when C<untie> occurs. See L<The C<untie> Gotcha> below.
301e8125 783
cb1a09d0
AD
784=item DESTROY this
785
786This method is triggered when a tied hash is about to go out of
787scope. You don't really need it unless you're trying to add debugging
788or have auxiliary state to clean up. Here's a very simple function:
789
790 sub DESTROY {
791 carp &whowasi if $DEBUG;
792 }
793
794=back
795
1d2dff63
GS
796Note that functions such as keys() and values() may return huge lists
797when used on large objects, like DBM files. You may prefer to use the
798each() function to iterate over such. Example:
cb1a09d0
AD
799
800 # print out history file offsets
801 use NDBM_File;
1f57c600 802 tie(%HIST, 'NDBM_File', '/usr/lib/news/history', 1, 0);
cb1a09d0
AD
803 while (($key,$val) = each %HIST) {
804 print $key, ' = ', unpack('L',$val), "\n";
805 }
806 untie(%HIST);
807
808=head2 Tying FileHandles
809
184e9718 810This is partially implemented now.
a7adf1f0 811
2ae324a7 812A class implementing a tied filehandle should define the following
1d603a67 813methods: TIEHANDLE, at least one of PRINT, PRINTF, WRITE, READLINE, GETC,
301e8125 814READ, and possibly CLOSE, UNTIE and DESTROY. The class can also provide: BINMODE,
4592e6ca
NIS
815OPEN, EOF, FILENO, SEEK, TELL - if the corresponding perl operators are
816used on the handle.
a7adf1f0 817
7ff03255
SG
818When STDERR is tied, its PRINT method will be called to issue warnings
819and error messages. This feature is temporarily disabled during the call,
820which means you can use C<warn()> inside PRINT without starting a recursive
821loop. And just like C<__WARN__> and C<__DIE__> handlers, STDERR's PRINT
822method may be called to report parser errors, so the caveats mentioned under
823L<perlvar/%SIG> apply.
824
825All of this is especially useful when perl is embedded in some other
826program, where output to STDOUT and STDERR may have to be redirected
827in some special way. See nvi and the Apache module for examples.
a7adf1f0
PP
828
829In our example we're going to create a shouting handle.
830
831 package Shout;
832
13a2d996 833=over 4
a7adf1f0
PP
834
835=item TIEHANDLE classname, LIST
836
837This is the constructor for the class. That means it is expected to
184e9718 838return a blessed reference of some sort. The reference can be used to
5f05dabc 839hold some internal information.
a7adf1f0 840
7e1af8bc 841 sub TIEHANDLE { print "<shout>\n"; my $i; bless \$i, shift }
a7adf1f0 842
1d603a67
GB
843=item WRITE this, LIST
844
845This method will be called when the handle is written to via the
846C<syswrite> function.
847
848 sub WRITE {
849 $r = shift;
850 my($buf,$len,$offset) = @_;
851 print "WRITE called, \$buf=$buf, \$len=$len, \$offset=$offset";
852 }
853
a7adf1f0
PP
854=item PRINT this, LIST
855
46fc3d4c
PP
856This method will be triggered every time the tied handle is printed to
857with the C<print()> function.
184e9718 858Beyond its self reference it also expects the list that was passed to
a7adf1f0
PP
859the print function.
860
58f51617
SV
861 sub PRINT { $r = shift; $$r++; print join($,,map(uc($_),@_)),$\ }
862
46fc3d4c
PP
863=item PRINTF this, LIST
864
865This method will be triggered every time the tied handle is printed to
866with the C<printf()> function.
867Beyond its self reference it also expects the format and list that was
868passed to the printf function.
869
870 sub PRINTF {
871 shift;
872 my $fmt = shift;
873 print sprintf($fmt, @_)."\n";
874 }
875
1d603a67 876=item READ this, LIST
2ae324a7
PP
877
878This method will be called when the handle is read from via the C<read>
879or C<sysread> functions.
880
881 sub READ {
889a76e8 882 my $self = shift;
69801a40 883 my $bufref = \$_[0];
889a76e8
GS
884 my(undef,$len,$offset) = @_;
885 print "READ called, \$buf=$bufref, \$len=$len, \$offset=$offset";
886 # add to $$bufref, set $len to number of characters read
887 $len;
2ae324a7
PP
888 }
889
58f51617
SV
890=item READLINE this
891
2ae324a7
PP
892This method will be called when the handle is read from via <HANDLE>.
893The method should return undef when there is no more data.
58f51617 894
889a76e8 895 sub READLINE { $r = shift; "READLINE called $$r times\n"; }
a7adf1f0 896
2ae324a7
PP
897=item GETC this
898
899This method will be called when the C<getc> function is called.
900
901 sub GETC { print "Don't GETC, Get Perl"; return "a"; }
902
1d603a67
GB
903=item CLOSE this
904
905This method will be called when the handle is closed via the C<close>
906function.
907
908 sub CLOSE { print "CLOSE called.\n" }
909
301e8125
NIS
910=item UNTIE this
911
912As with the other types of ties, this method will be called when C<untie> happens.
d5582e24
IZ
913It may be appropriate to "auto CLOSE" when this occurs. See
914L<The C<untie> Gotcha> below.
301e8125 915
a7adf1f0
PP
916=item DESTROY this
917
918As with the other types of ties, this method will be called when the
919tied handle is about to be destroyed. This is useful for debugging and
920possibly cleaning up.
921
922 sub DESTROY { print "</shout>\n" }
923
924=back
925
926Here's how to use our little example:
927
928 tie(*FOO,'Shout');
929 print FOO "hello\n";
930 $a = 4; $b = 6;
931 print FOO $a, " plus ", $b, " equals ", $a + $b, "\n";
58f51617 932 print <FOO>;
cb1a09d0 933
d7da42b7
JH
934=head2 UNTIE this
935
936You can define for all tie types an UNTIE method that will be called
d5582e24 937at untie(). See L<The C<untie> Gotcha> below.
d7da42b7 938
2752eb9f
PM
939=head2 The C<untie> Gotcha
940
941If you intend making use of the object returned from either tie() or
942tied(), and if the tie's target class defines a destructor, there is a
943subtle gotcha you I<must> guard against.
944
945As setup, consider this (admittedly rather contrived) example of a
946tie; all it does is use a file to keep a log of the values assigned to
947a scalar.
948
949 package Remember;
950
951 use strict;
9f1b1f2d 952 use warnings;
2752eb9f
PM
953 use IO::File;
954
955 sub TIESCALAR {
956 my $class = shift;
957 my $filename = shift;
958 my $handle = new IO::File "> $filename"
959 or die "Cannot open $filename: $!\n";
960
961 print $handle "The Start\n";
962 bless {FH => $handle, Value => 0}, $class;
963 }
964
965 sub FETCH {
966 my $self = shift;
967 return $self->{Value};
968 }
969
970 sub STORE {
971 my $self = shift;
972 my $value = shift;
973 my $handle = $self->{FH};
974 print $handle "$value\n";
975 $self->{Value} = $value;
976 }
977
978 sub DESTROY {
979 my $self = shift;
980 my $handle = $self->{FH};
981 print $handle "The End\n";
982 close $handle;
983 }
984
985 1;
986
987Here is an example that makes use of this tie:
988
989 use strict;
990 use Remember;
991
992 my $fred;
993 tie $fred, 'Remember', 'myfile.txt';
994 $fred = 1;
995 $fred = 4;
996 $fred = 5;
997 untie $fred;
998 system "cat myfile.txt";
999
1000This is the output when it is executed:
1001
1002 The Start
1003 1
1004 4
1005 5
1006 The End
1007
1008So far so good. Those of you who have been paying attention will have
1009spotted that the tied object hasn't been used so far. So lets add an
1010extra method to the Remember class to allow comments to be included in
1011the file -- say, something like this:
1012
1013 sub comment {
1014 my $self = shift;
1015 my $text = shift;
1016 my $handle = $self->{FH};
1017 print $handle $text, "\n";
1018 }
1019
1020And here is the previous example modified to use the C<comment> method
1021(which requires the tied object):
1022
1023 use strict;
1024 use Remember;
1025
1026 my ($fred, $x);
1027 $x = tie $fred, 'Remember', 'myfile.txt';
1028 $fred = 1;
1029 $fred = 4;
1030 comment $x "changing...";
1031 $fred = 5;
1032 untie $fred;
1033 system "cat myfile.txt";
1034
1035When this code is executed there is no output. Here's why:
1036
1037When a variable is tied, it is associated with the object which is the
1038return value of the TIESCALAR, TIEARRAY, or TIEHASH function. This
1039object normally has only one reference, namely, the implicit reference
1040from the tied variable. When untie() is called, that reference is
1041destroyed. Then, as in the first example above, the object's
1042destructor (DESTROY) is called, which is normal for objects that have
1043no more valid references; and thus the file is closed.
1044
1045In the second example, however, we have stored another reference to
19799a22 1046the tied object in $x. That means that when untie() gets called
2752eb9f
PM
1047there will still be a valid reference to the object in existence, so
1048the destructor is not called at that time, and thus the file is not
1049closed. The reason there is no output is because the file buffers
1050have not been flushed to disk.
1051
1052Now that you know what the problem is, what can you do to avoid it?
301e8125
NIS
1053Prior to the introduction of the optional UNTIE method the only way
1054was the good old C<-w> flag. Which will spot any instances where you call
2752eb9f 1055untie() and there are still valid references to the tied object. If
9f1b1f2d
GS
1056the second script above this near the top C<use warnings 'untie'>
1057or was run with the C<-w> flag, Perl prints this
2752eb9f
PM
1058warning message:
1059
1060 untie attempted while 1 inner references still exist
1061
1062To get the script to work properly and silence the warning make sure
1063there are no valid references to the tied object I<before> untie() is
1064called:
1065
1066 undef $x;
1067 untie $fred;
1068
301e8125
NIS
1069Now that UNTIE exists the class designer can decide which parts of the
1070class functionality are really associated with C<untie> and which with
1071the object being destroyed. What makes sense for a given class depends
1072on whether the inner references are being kept so that non-tie-related
1073methods can be called on the object. But in most cases it probably makes
1074sense to move the functionality that would have been in DESTROY to the UNTIE
1075method.
1076
1077If the UNTIE method exists then the warning above does not occur. Instead the
1078UNTIE method is passed the count of "extra" references and can issue its own
1079warning if appropriate. e.g. to replicate the no UNTIE case this method can
1080be used:
1081
1082 sub UNTIE
1083 {
1084 my ($obj,$count) = @_;
1085 carp "untie attempted while $count inner references still exist" if $count;
1086 }
1087
cb1a09d0
AD
1088=head1 SEE ALSO
1089
1090See L<DB_File> or L<Config> for some interesting tie() implementations.
3d0ae7ba
GS
1091A good starting point for many tie() implementations is with one of the
1092modules L<Tie::Scalar>, L<Tie::Array>, L<Tie::Hash>, or L<Tie::Handle>.
cb1a09d0
AD
1093
1094=head1 BUGS
1095
029149a3
JH
1096The bucket usage information provided by C<scalar(%hash)> is not
1097available. What this means is that using %tied_hash in boolean
1098context doesn't work right (currently this always tests false,
1099regardless of whether the hash is empty or hash elements).
1100
1101Localizing tied arrays or hashes does not work. After exiting the
1102scope the arrays or the hashes are not restored.
1103
e77edca3
JH
1104Counting the number of entries in a hash via C<scalar(keys(%hash))>
1105or C<scalar(values(%hash)>) is inefficient since it needs to iterate
1106through all the entries with FIRSTKEY/NEXTKEY.
1107
1108Tied hash/array slices cause multiple FETCH/STORE pairs, there are no
1109tie methods for slice operations.
1110
c07a80fd
PP
1111You cannot easily tie a multilevel data structure (such as a hash of
1112hashes) to a dbm file. The first problem is that all but GDBM and
1113Berkeley DB have size limitations, but beyond that, you also have problems
1114with how references are to be represented on disk. One experimental
5f05dabc 1115module that does attempt to address this need partially is the MLDBM
f102b883 1116module. Check your nearest CPAN site as described in L<perlmodlib> for
c07a80fd
PP
1117source code to MLDBM.
1118
e08f2115
GA
1119Tied filehandles are still incomplete. sysopen(), truncate(),
1120flock(), fcntl(), stat() and -X can't currently be trapped.
1121
cb1a09d0
AD
1122=head1 AUTHOR
1123
1124Tom Christiansen
a7adf1f0 1125
46fc3d4c 1126TIEHANDLE by Sven Verdoolaege <F<skimo@dns.ufsia.ac.be>> and Doug MacEachern <F<dougm@osf.org>>
301e8125
NIS
1127
1128UNTIE by Nick Ing-Simmons <F<nick@ing-simmons.net>>
1129
a3bcc51e
TP
1130SCALAR by Tassilo von Parseval <F<tassilo.von.parseval@rwth-aachen.de>>
1131
e1e60e72 1132Tying Arrays by Casey West <F<casey@geeknest.com>>