This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
cmp() compares chars, not bytes.
[perl5.git] / ext / Storable / Storable.pm
CommitLineData
596596d5 1;# $Id: Storable.pm,v 1.0.1.13 2001/12/01 13:34:49 ram Exp $
7a6a85bf
RG
2;#
3;# Copyright (c) 1995-2000, Raphael Manfredi
4;#
9e21b3d0
JH
5;# You may redistribute only under the same terms as Perl 5, as specified
6;# in the README file that comes with the distribution.
7a6a85bf
RG
7;#
8;# $Log: Storable.pm,v $
596596d5
JH
9;# Revision 1.0.1.13 2001/12/01 13:34:49 ram
10;# patch14: avoid requiring Fcntl upfront, useful to embedded runtimes
11;# patch14: store_fd() will now correctly autoflush file if needed
12;#
6e0ac6f5
JH
13;# Revision 1.0.1.12 2001/08/28 21:51:51 ram
14;# patch13: fixed truncation race with lock_retrieve() in lock_store()
15;#
e993d95c
JH
16;# Revision 1.0.1.11 2001/07/01 11:22:14 ram
17;# patch12: systematically use "=over 4" for POD linters
18;# patch12: updated version number
19;#
8be2b38b
JH
20;# Revision 1.0.1.10 2001/03/15 00:20:25 ram
21;# patch11: updated version number
22;#
23;# Revision 1.0.1.9 2001/02/17 12:37:32 ram
24;# patch10: forgot to increase version number at previous patch
25;#
b12202d0
JH
26;# Revision 1.0.1.8 2001/02/17 12:24:37 ram
27;# patch8: fixed incorrect error message
28;#
862382c7
JH
29;# Revision 1.0.1.7 2001/01/03 09:39:02 ram
30;# patch7: added CAN_FLOCK to determine whether we can flock() or not
31;#
90826881
JH
32;# Revision 1.0.1.6 2000/11/05 17:20:25 ram
33;# patch6: increased version number
34;#
212e9bde
JH
35;# Revision 1.0.1.5 2000/10/26 17:10:18 ram
36;# patch5: documented that store() and retrieve() can return undef
37;# patch5: added paragraph explaining the auto require for thaw hooks
38;#
39;# Revision 1.0.1.4 2000/10/23 18:02:57 ram
40;# patch4: protected calls to flock() for dos platform
41;# patch4: added logcarp emulation if they don't have Log::Agent
42;#
8be2b38b
JH
43;# Revision 1.0.1.3 2000/09/29 19:49:01 ram
44;# patch3: updated version number
45;#
46;# Revision 1.0.1.2 2000/09/28 21:42:51 ram
47;# patch2: added lock_store lock_nstore lock_retrieve
48;#
49;# Revision 1.0.1.1 2000/09/17 16:46:21 ram
50;# patch1: documented that doubles are stringified by nstore()
51;# patch1: added Salvador Ortiz Garcia in CREDITS section
52;#
9e21b3d0
JH
53;# Revision 1.0 2000/09/01 19:40:41 ram
54;# Baseline for first official release.
7a6a85bf
RG
55;#
56
57require DynaLoader;
58require Exporter;
59package Storable; @ISA = qw(Exporter DynaLoader);
60
61@EXPORT = qw(store retrieve);
62@EXPORT_OK = qw(
9e21b3d0 63 nstore store_fd nstore_fd fd_retrieve
7a6a85bf
RG
64 freeze nfreeze thaw
65 dclone
9e21b3d0 66 retrieve_fd
dd19458b 67 lock_store lock_nstore lock_retrieve
7a6a85bf
RG
68);
69
70use AutoLoader;
71use vars qw($forgive_me $VERSION);
72
caa9d880 73$VERSION = '2.0';
7a6a85bf
RG
74*AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
75
76#
77# Use of Log::Agent is optional
78#
79
80eval "use Log::Agent";
81
530b72ba 82require Carp;
7a6a85bf 83
dd19458b
JH
84#
85# They might miss :flock in Fcntl
86#
87
88BEGIN {
596596d5 89 if (eval { require Fcntl; 1 } && exists $Fcntl::EXPORT_TAGS{'flock'}) {
dd19458b
JH
90 Fcntl->import(':flock');
91 } else {
92 eval q{
93 sub LOCK_SH () {1}
94 sub LOCK_EX () {2}
95 };
96 }
97}
98
b8778c7c 99# Can't Autoload cleanly as this clashes 8.3 with &retrieve
9e21b3d0 100sub retrieve_fd { &fd_retrieve } # Backward compatibility
cb3d9de5 101
530b72ba
NC
102# By default restricted hashes are downgraded on earlier perls.
103
104$Storable::downgrade_restricted = 1;
e8189732 105$Storable::accept_future_minor = 1;
b8778c7c
NC
106bootstrap Storable;
1071;
108__END__
530b72ba
NC
109#
110# Use of Log::Agent is optional. If it hasn't imported these subs then
111# Autoloader will kindly supply our fallback implementation.
112#
113
114sub logcroak {
115 Carp::croak(@_);
116}
117
118sub logcarp {
119 Carp::carp(@_);
120}
b8778c7c 121
862382c7
JH
122#
123# Determine whether locking is possible, but only when needed.
124#
125
530b72ba 126sub CAN_FLOCK; my $CAN_FLOCK; sub CAN_FLOCK {
862382c7
JH
127 return $CAN_FLOCK if defined $CAN_FLOCK;
128 require Config; import Config;
129 return $CAN_FLOCK =
130 $Config{'d_flock'} ||
131 $Config{'d_fcntl_can_lock'} ||
132 $Config{'d_lockf'};
133}
134
0a0da639
JH
135sub show_file_magic {
136 print <<EOM;
137#
138# To recognize the data files of the Perl module Storable,
139# the following lines need to be added to the local magic(5) file,
140# usually either /usr/share/misc/magic or /etc/magic.
0a0da639
JH
141#
1420 string perl-store perl Storable(v0.6) data
8b793558
JH
143>4 byte >0 (net-order %d)
144>>4 byte &01 (network-ordered)
145>>4 byte =3 (major 1)
146>>4 byte =2 (major 1)
147
0a0da639 1480 string pst0 perl Storable(v0.7) data
8b793558
JH
149>4 byte >0
150>>4 byte &01 (network-ordered)
151>>4 byte =5 (major 2)
152>>4 byte =4 (major 2)
153>>5 byte >0 (minor %d)
0a0da639
JH
154EOM
155}
156
b8778c7c
NC
157sub read_magic {
158 my $header = shift;
159 return unless defined $header and length $header > 11;
160 my $result;
161 if ($header =~ s/^perl-store//) {
162 die "Can't deal with version 0 headers";
163 } elsif ($header =~ s/^pst0//) {
164 $result->{file} = 1;
165 }
166 # Assume it's a string.
167 my ($major, $minor, $bytelen) = unpack "C3", $header;
168
169 my $net_order = $major & 1;
170 $major >>= 1;
171 @$result{qw(major minor netorder)} = ($major, $minor, $net_order);
172
173 return $result if $net_order;
174
175 # I assume that it is rare to find v1 files, so this is an intentionally
176 # inefficient way of doing it, to make the rest of the code constant.
177 if ($major < 2) {
178 delete $result->{minor};
179 $header = '.' . $header;
180 $bytelen = $minor;
181 }
182
183 @$result{qw(byteorder intsize longsize ptrsize)} =
184 unpack "x3 A$bytelen C3", $header;
185
186 if ($major >= 2 and $minor >= 2) {
187 $result->{nvsize} = unpack "x6 x$bytelen C", $header;
188 }
189 $result;
190}
7a6a85bf
RG
191
192#
193# store
194#
195# Store target object hierarchy, identified by a reference to its root.
196# The stored object tree may later be retrieved to memory via retrieve.
197# Returns undef if an I/O error occurred, in which case the file is
198# removed.
199#
200sub store {
dd19458b 201 return _store(\&pstore, @_, 0);
7a6a85bf
RG
202}
203
204#
205# nstore
206#
207# Same as store, but in network order.
208#
209sub nstore {
dd19458b
JH
210 return _store(\&net_pstore, @_, 0);
211}
212
213#
214# lock_store
215#
216# Same as store, but flock the file first (advisory locking).
217#
218sub lock_store {
219 return _store(\&pstore, @_, 1);
220}
221
222#
223# lock_nstore
224#
225# Same as nstore, but flock the file first (advisory locking).
226#
227sub lock_nstore {
228 return _store(\&net_pstore, @_, 1);
7a6a85bf
RG
229}
230
231# Internal store to file routine
232sub _store {
233 my $xsptr = shift;
234 my $self = shift;
dd19458b 235 my ($file, $use_locking) = @_;
7a6a85bf 236 logcroak "not a reference" unless ref($self);
b12202d0 237 logcroak "wrong argument number" unless @_ == 2; # No @foo in arglist
7a6a85bf 238 local *FILE;
dd19458b 239 if ($use_locking) {
6e0ac6f5 240 open(FILE, ">>$file") || logcroak "can't write into $file: $!";
862382c7 241 unless (&CAN_FLOCK) {
b29b780f
RM
242 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
243 return undef;
f567092b 244 }
dd19458b
JH
245 flock(FILE, LOCK_EX) ||
246 logcroak "can't get exclusive lock on $file: $!";
247 truncate FILE, 0;
248 # Unlocking will happen when FILE is closed
6e0ac6f5
JH
249 } else {
250 open(FILE, ">$file") || logcroak "can't create $file: $!";
dd19458b 251 }
6e0ac6f5 252 binmode FILE; # Archaic systems...
7a6a85bf
RG
253 my $da = $@; # Don't mess if called from exception handler
254 my $ret;
255 # Call C routine nstore or pstore, depending on network order
256 eval { $ret = &$xsptr(*FILE, $self) };
257 close(FILE) or $ret = undef;
258 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
259 logcroak $@ if $@ =~ s/\.?\n$/,/;
260 $@ = $da;
261 return $ret ? $ret : undef;
262}
263
264#
265# store_fd
266#
267# Same as store, but perform on an already opened file descriptor instead.
268# Returns undef if an I/O error occurred.
269#
270sub store_fd {
271 return _store_fd(\&pstore, @_);
272}
273
274#
275# nstore_fd
276#
277# Same as store_fd, but in network order.
278#
279sub nstore_fd {
280 my ($self, $file) = @_;
281 return _store_fd(\&net_pstore, @_);
282}
283
284# Internal store routine on opened file descriptor
285sub _store_fd {
286 my $xsptr = shift;
287 my $self = shift;
288 my ($file) = @_;
289 logcroak "not a reference" unless ref($self);
290 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
291 my $fd = fileno($file);
292 logcroak "not a valid file descriptor" unless defined $fd;
293 my $da = $@; # Don't mess if called from exception handler
294 my $ret;
295 # Call C routine nstore or pstore, depending on network order
296 eval { $ret = &$xsptr($file, $self) };
297 logcroak $@ if $@ =~ s/\.?\n$/,/;
596596d5 298 local $\; print $file ''; # Autoflush the file if wanted
7a6a85bf
RG
299 $@ = $da;
300 return $ret ? $ret : undef;
301}
302
303#
304# freeze
305#
306# Store oject and its hierarchy in memory and return a scalar
307# containing the result.
308#
309sub freeze {
310 _freeze(\&mstore, @_);
311}
312
313#
314# nfreeze
315#
316# Same as freeze but in network order.
317#
318sub nfreeze {
319 _freeze(\&net_mstore, @_);
320}
321
322# Internal freeze routine
323sub _freeze {
324 my $xsptr = shift;
325 my $self = shift;
326 logcroak "not a reference" unless ref($self);
327 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
328 my $da = $@; # Don't mess if called from exception handler
329 my $ret;
330 # Call C routine mstore or net_mstore, depending on network order
331 eval { $ret = &$xsptr($self) };
332 logcroak $@ if $@ =~ s/\.?\n$/,/;
333 $@ = $da;
334 return $ret ? $ret : undef;
335}
336
337#
338# retrieve
339#
340# Retrieve object hierarchy from disk, returning a reference to the root
341# object of that tree.
342#
343sub retrieve {
dd19458b
JH
344 _retrieve($_[0], 0);
345}
346
347#
348# lock_retrieve
349#
350# Same as retrieve, but with advisory locking.
351#
352sub lock_retrieve {
353 _retrieve($_[0], 1);
354}
355
356# Internal retrieve routine
357sub _retrieve {
358 my ($file, $use_locking) = @_;
7a6a85bf 359 local *FILE;
dd19458b 360 open(FILE, $file) || logcroak "can't open $file: $!";
7a6a85bf
RG
361 binmode FILE; # Archaic systems...
362 my $self;
363 my $da = $@; # Could be from exception handler
dd19458b 364 if ($use_locking) {
862382c7 365 unless (&CAN_FLOCK) {
8be2b38b 366 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
b29b780f
RM
367 return undef;
368 }
8be2b38b 369 flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!";
dd19458b
JH
370 # Unlocking will happen when FILE is closed
371 }
7a6a85bf
RG
372 eval { $self = pretrieve(*FILE) }; # Call C routine
373 close(FILE);
374 logcroak $@ if $@ =~ s/\.?\n$/,/;
375 $@ = $da;
376 return $self;
377}
378
379#
9e21b3d0 380# fd_retrieve
7a6a85bf
RG
381#
382# Same as retrieve, but perform from an already opened file descriptor instead.
383#
9e21b3d0 384sub fd_retrieve {
7a6a85bf
RG
385 my ($file) = @_;
386 my $fd = fileno($file);
387 logcroak "not a valid file descriptor" unless defined $fd;
388 my $self;
389 my $da = $@; # Could be from exception handler
390 eval { $self = pretrieve($file) }; # Call C routine
391 logcroak $@ if $@ =~ s/\.?\n$/,/;
392 $@ = $da;
393 return $self;
394}
395
396#
397# thaw
398#
399# Recreate objects in memory from an existing frozen image created
400# by freeze. If the frozen image passed is undef, return undef.
401#
402sub thaw {
403 my ($frozen) = @_;
404 return undef unless defined $frozen;
405 my $self;
406 my $da = $@; # Could be from exception handler
407 eval { $self = mretrieve($frozen) }; # Call C routine
408 logcroak $@ if $@ =~ s/\.?\n$/,/;
409 $@ = $da;
410 return $self;
411}
412
413=head1 NAME
414
f062ea6c 415Storable - persistence for Perl data structures
7a6a85bf
RG
416
417=head1 SYNOPSIS
418
419 use Storable;
420 store \%table, 'file';
421 $hashref = retrieve('file');
422
423 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
424
425 # Network order
426 nstore \%table, 'file';
427 $hashref = retrieve('file'); # There is NO nretrieve()
428
429 # Storing to and retrieving from an already opened file
430 store_fd \@array, \*STDOUT;
431 nstore_fd \%table, \*STDOUT;
9e21b3d0
JH
432 $aryref = fd_retrieve(\*SOCKET);
433 $hashref = fd_retrieve(\*SOCKET);
7a6a85bf
RG
434
435 # Serializing to memory
436 $serialized = freeze \%table;
437 %table_clone = %{ thaw($serialized) };
438
439 # Deep (recursive) cloning
440 $cloneref = dclone($ref);
441
dd19458b
JH
442 # Advisory locking
443 use Storable qw(lock_store lock_nstore lock_retrieve)
444 lock_store \%table, 'file';
445 lock_nstore \%table, 'file';
446 $hashref = lock_retrieve('file');
447
7a6a85bf
RG
448=head1 DESCRIPTION
449
f062ea6c 450The Storable package brings persistence to your Perl data structures
7a6a85bf 451containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
c261f00e 452conveniently stored to disk and retrieved at a later time.
7a6a85bf
RG
453
454It can be used in the regular procedural way by calling C<store> with
455a reference to the object to be stored, along with the file name where
456the image should be written.
775ecd75 457
7a6a85bf
RG
458The routine returns C<undef> for I/O problems or other internal error,
459a true value otherwise. Serious errors are propagated as a C<die> exception.
460
f062ea6c
PN
461To retrieve data stored to disk, use C<retrieve> with a file name.
462The objects stored into that file are recreated into memory for you,
463and a I<reference> to the root object is returned. In case an I/O error
7a6a85bf
RG
464occurs while reading, C<undef> is returned instead. Other serious
465errors are propagated via C<die>.
466
467Since storage is performed recursively, you might want to stuff references
468to objects that share a lot of common data into a single array or hash
469table, and then store that object. That way, when you retrieve back the
470whole thing, the objects will continue to share what they originally shared.
471
472At the cost of a slight header overhead, you may store to an already
473opened file descriptor using the C<store_fd> routine, and retrieve
9e21b3d0 474from a file via C<fd_retrieve>. Those names aren't imported by default,
c261f00e 475so you will have to do that explicitly if you need those routines.
7a6a85bf
RG
476The file descriptor you supply must be already opened, for read
477if you're going to retrieve and for write if you wish to store.
478
479 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
9e21b3d0 480 $hashref = fd_retrieve(*STDIN);
7a6a85bf
RG
481
482You can also store data in network order to allow easy sharing across
483multiple platforms, or when storing on a socket known to be remotely
484connected. The routines to call have an initial C<n> prefix for I<network>,
485as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
486correctly restored so you don't have to know whether you're restoring
dd19458b
JH
487from native or network ordered data. Double values are stored stringified
488to ensure portability as well, at the slight risk of loosing some precision
489in the last decimals.
7a6a85bf 490
9e21b3d0 491When using C<fd_retrieve>, objects are retrieved in sequence, one
7a6a85bf
RG
492object (i.e. one recursive tree) per associated C<store_fd>.
493
494If you're more from the object-oriented camp, you can inherit from
495Storable and directly store your objects by invoking C<store> as
496a method. The fact that the root of the to-be-stored tree is a
497blessed reference (i.e. an object) is special-cased so that the
498retrieve does not provide a reference to that object but rather the
499blessed object reference itself. (Otherwise, you'd get a reference
500to that blessed object).
501
502=head1 MEMORY STORE
503
504The Storable engine can also store data into a Perl scalar instead, to
505later retrieve them. This is mainly used to freeze a complex structure in
506some safe compact memory place (where it can possibly be sent to another
507process via some IPC, since freezing the structure also serializes it in
508effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
509out and recreate the original complex structure in memory.
510
511Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
512If you wish to send out the frozen scalar to another machine, use
513C<nfreeze> instead to get a portable image.
514
515Note that freezing an object structure and immediately thawing it
516actually achieves a deep cloning of that structure:
517
518 dclone(.) = thaw(freeze(.))
519
520Storable provides you with a C<dclone> interface which does not create
521that intermediary scalar but instead freezes the structure in some
c261f00e 522internal memory space and then immediately thaws it out.
7a6a85bf 523
dd19458b
JH
524=head1 ADVISORY LOCKING
525
f062ea6c
PN
526The C<lock_store> and C<lock_nstore> routine are equivalent to
527C<store> and C<nstore>, except that they get an exclusive lock on
528the file before writing. Likewise, C<lock_retrieve> does the same
529as C<retrieve>, but also gets a shared lock on the file before reading.
dd19458b 530
f062ea6c
PN
531As with any advisory locking scheme, the protection only works if you
532systematically use C<lock_store> and C<lock_retrieve>. If one side of
533your application uses C<store> whilst the other uses C<lock_retrieve>,
dd19458b
JH
534you will get no protection at all.
535
f062ea6c
PN
536The internal advisory locking is implemented using Perl's flock()
537routine. If your system does not support any form of flock(), or if
538you share your files across NFS, you might wish to use other forms
539of locking by using modules such as LockFile::Simple which lock a
540file using a filesystem entry, instead of locking the file descriptor.
dd19458b 541
7a6a85bf
RG
542=head1 SPEED
543
544The heart of Storable is written in C for decent speed. Extra low-level
4d3295e3
PN
545optimizations have been made when manipulating perl internals, to
546sacrifice encapsulation for the benefit of greater speed.
7a6a85bf
RG
547
548=head1 CANONICAL REPRESENTATION
549
f062ea6c 550Normally, Storable stores elements of hashes in the order they are
7a6a85bf
RG
551stored internally by Perl, i.e. pseudo-randomly. If you set
552C<$Storable::canonical> to some C<TRUE> value, Storable will store
553hashes with the elements sorted by their key. This allows you to
554compare data structures by comparing their frozen representations (or
555even the compressed frozen representations), which can be useful for
556creating lookup tables for complicated queries.
557
f062ea6c 558Canonical order does not imply network order; those are two orthogonal
7a6a85bf
RG
559settings.
560
c261f00e
NC
561=head1 FORWARD COMPATIBILITY
562
563This release of Storable can be used on a newer version of Perl to
f062ea6c 564serialize data which is not supported by earlier Perls. By default,
c261f00e 565Storable will attempt to do the right thing, by C<croak()>ing if it
775ecd75 566encounters data that it cannot deserialize. However, the defaults
f062ea6c 567can be changed as follows:
c261f00e
NC
568
569=over 4
570
571=item utf8 data
572
573Perl 5.6 added support for Unicode characters with code points > 255,
574and Perl 5.8 has full support for Unicode characters in hash keys.
575Perl internally encodes strings with these characters using utf8, and
576Storable serializes them as utf8. By default, if an older version of
577Perl encounters a utf8 value it cannot represent, it will C<croak()>.
578To change this behaviour so that Storable deserializes utf8 encoded
579values as the string of bytes (effectively dropping the I<is_utf8> flag)
580set C<$Storable::drop_utf8> to some C<TRUE> value. This is a form of
581data loss, because with C<$drop_utf8> true, it becomes impossible to tell
582whether the original data was the Unicode string, or a series of bytes
583that happen to be valid utf8.
584
585=item restricted hashes
586
f062ea6c
PN
587Perl 5.8 adds support for restricted hashes, which have keys
588restricted to a given set, and can have values locked to be read only.
589By default, when Storable encounters a restricted hash on a perl
590that doesn't support them, it will deserialize it as a normal hash,
591silently discarding any placeholder keys and leaving the keys and
592all values unlocked. To make Storable C<croak()> instead, set
593C<$Storable::downgrade_restricted> to a C<FALSE> value. To restore
594the default set it back to some C<TRUE> value.
c261f00e 595
e8189732
NC
596=item files from future versions of Storable
597
598Earlier versions of Storable would immediately croak if they encountered
599a file with a higher internal version number than the reading Storable
600knew about. Internal version numbers are increased each time new data
601types (such as restricted hashes) are added to the vocabulary of the file
602format. This meant that a newer Storable module had no way of writing a
f062ea6c 603file readable by an older Storable, even if the writer didn't store newer
e8189732
NC
604data types.
605
606This version of Storable will defer croaking until it encounters a data
607type in the file that it does not recognize. This means that it will
608continue to read files generated by newer Storable modules which are careful
609in what they write out, making it easier to upgrade Storable modules in a
610mixed environment.
611
612The old behaviour of immediate croaking can be re-instated by setting
f062ea6c 613C<$Storable::accept_future_minor> to some C<FALSE> value.
e8189732 614
c261f00e
NC
615=back
616
f062ea6c 617All these variables have no effect on a newer Perl which supports the
c261f00e
NC
618relevant feature.
619
7a6a85bf
RG
620=head1 ERROR REPORTING
621
622Storable uses the "exception" paradigm, in that it does not try to workaround
623failures: if something bad happens, an exception is generated from the
624caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
625those exceptions.
626
627When Storable croaks, it tries to report the error via the C<logcroak()>
628routine from the C<Log::Agent> package, if it is available.
629
212e9bde
JH
630Normal errors are reported by having store() or retrieve() return C<undef>.
631Such errors are usually I/O errors (or truncated stream errors at retrieval).
632
7a6a85bf
RG
633=head1 WIZARDS ONLY
634
635=head2 Hooks
636
637Any class may define hooks that will be called during the serialization
638and deserialization process on objects that are instances of that class.
639Those hooks can redefine the way serialization is performed (and therefore,
c261f00e 640how the symmetrical deserialization should be conducted).
7a6a85bf
RG
641
642Since we said earlier:
643
644 dclone(.) = thaw(freeze(.))
645
646everything we say about hooks should also hold for deep cloning. However,
647hooks get to know whether the operation is a mere serialization, or a cloning.
648
649Therefore, when serializing hooks are involved,
650
651 dclone(.) <> thaw(freeze(.))
652
653Well, you could keep them in sync, but there's no guarantee it will always
654hold on classes somebody else wrote. Besides, there is little to gain in
f062ea6c 655doing so: a serializing hook could keep only one attribute of an object,
7a6a85bf
RG
656which is probably not what should happen during a deep cloning of that
657same object.
658
659Here is the hooking interface:
660
bbc7dcd2 661=over 4
7a6a85bf
RG
662
663=item C<STORABLE_freeze> I<obj>, I<cloning>
664
665The serializing hook, called on the object during serialization. It can be
666inherited, or defined in the class itself, like any other method.
667
668Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
669whether we're in a dclone() or a regular serialization via store() or freeze().
670
671Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
672is the serialized form to be used, and the optional $ref1, $ref2, etc... are
673extra references that you wish to let the Storable engine serialize.
674
675At deserialization time, you will be given back the same LIST, but all the
676extra references will be pointing into the deserialized structure.
677
678The B<first time> the hook is hit in a serialization flow, you may have it
679return an empty list. That will signal the Storable engine to further
680discard that hook for this class and to therefore revert to the default
681serialization of the underlying Perl data. The hook will again be normally
682processed in the next serialization.
683
684Unless you know better, serializing hook should always say:
685
686 sub STORABLE_freeze {
687 my ($self, $cloning) = @_;
688 return if $cloning; # Regular default serialization
689 ....
690 }
691
692in order to keep reasonable dclone() semantics.
693
694=item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
695
696The deserializing hook called on the object during deserialization.
f062ea6c 697But wait: if we're deserializing, there's no object yet... right?
7a6a85bf
RG
698
699Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
700you can view C<STORABLE_thaw> as an alternate creation routine.
701
702This means the hook can be inherited like any other method, and that
703I<obj> is your blessed reference for this particular instance.
704
705The other arguments should look familiar if you know C<STORABLE_freeze>:
706I<cloning> is true when we're part of a deep clone operation, I<serialized>
707is the serialized string you returned to the engine in C<STORABLE_freeze>,
708and there may be an optional list of references, in the same order you gave
709them at serialization time, pointing to the deserialized objects (which
710have been processed courtesy of the Storable engine).
711
212e9bde
JH
712When the Storable engine does not find any C<STORABLE_thaw> hook routine,
713it tries to load the class by requiring the package dynamically (using
714the blessed package name), and then re-attempts the lookup. If at that
715time the hook cannot be located, the engine croaks. Note that this mechanism
c261f00e 716will fail if you define several classes in the same file, but L<perlmod>
212e9bde
JH
717warned you.
718
f062ea6c 719It is up to you to use this information to populate I<obj> the way you want.
7a6a85bf
RG
720
721Returned value: none.
722
723=back
724
725=head2 Predicates
726
c261f00e 727Predicates are not exportable. They must be called by explicitly prefixing
7a6a85bf
RG
728them with the Storable package name.
729
bbc7dcd2 730=over 4
7a6a85bf
RG
731
732=item C<Storable::last_op_in_netorder>
733
734The C<Storable::last_op_in_netorder()> predicate will tell you whether
735network order was used in the last store or retrieve operation. If you
736don't know how to use this, just forget about it.
737
738=item C<Storable::is_storing>
739
740Returns true if within a store operation (via STORABLE_freeze hook).
741
742=item C<Storable::is_retrieving>
743
f062ea6c 744Returns true if within a retrieve operation (via STORABLE_thaw hook).
7a6a85bf
RG
745
746=back
747
748=head2 Recursion
749
f062ea6c
PN
750With hooks comes the ability to recurse back to the Storable engine.
751Indeed, hooks are regular Perl code, and Storable is convenient when
752it comes to serializing and deserializing things, so why not use it
753to handle the serialization string?
7a6a85bf 754
f062ea6c 755There are a few things you need to know, however:
7a6a85bf 756
bbc7dcd2 757=over 4
7a6a85bf
RG
758
759=item *
760
761You can create endless loops if the things you serialize via freeze()
f062ea6c
PN
762(for instance) point back to the object we're trying to serialize in
763the hook.
7a6a85bf
RG
764
765=item *
766
767Shared references among objects will not stay shared: if we're serializing
768the list of object [A, C] where both object A and C refer to the SAME object
769B, and if there is a serializing hook in A that says freeze(B), then when
770deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
771a deep clone of B'. The topology was not preserved.
772
773=back
774
775That's why C<STORABLE_freeze> lets you provide a list of references
776to serialize. The engine guarantees that those will be serialized in the
777same context as the other objects, and therefore that shared objects will
778stay shared.
779
780In the above [A, C] example, the C<STORABLE_freeze> hook could return:
781
782 ("something", $self->{B})
783
784and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
785would get back the reference to the B' object, deserialized for you.
786
787Therefore, recursion should normally be avoided, but is nonetheless supported.
788
789=head2 Deep Cloning
790
f062ea6c 791There is a Clone module available on CPAN which implements deep cloning
7a6a85bf
RG
792natively, i.e. without freezing to memory and thawing the result. It is
793aimed to replace Storable's dclone() some day. However, it does not currently
794support Storable hooks to redefine the way deep cloning is performed.
795
0a0da639
JH
796=head1 Storable magic
797
798Yes, there's a lot of that :-) But more precisely, in UNIX systems
799there's a utility called C<file>, which recognizes data files based on
800their contents (usually their first few bytes). For this to work,
8b793558 801a certain file called F<magic> needs to taught about the I<signature>
0a0da639 802of the data. Where that configuration file lives depends on the UNIX
f062ea6c 803flavour; often it's something like F</usr/share/misc/magic> or
8b793558
JH
804F</etc/magic>. Your system administrator needs to do the updating of
805the F<magic> file. The necessary signature information is output to
f062ea6c
PN
806STDOUT by invoking Storable::show_file_magic(). Note that the GNU
807implementation of the C<file> utility, version 3.38 or later,
808is expected to contain support for recognising Storable files
809out-of-the-box, in addition to other kinds of Perl files.
0a0da639 810
7a6a85bf
RG
811=head1 EXAMPLES
812
813Here are some code samples showing a possible usage of Storable:
814
815 use Storable qw(store retrieve freeze thaw dclone);
816
817 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
818
819 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
820
821 $colref = retrieve('/tmp/colors');
822 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
823 printf "Blue is still %lf\n", $colref->{'Blue'};
824
825 $colref2 = dclone(\%color);
826
827 $str = freeze(\%color);
828 printf "Serialization of %%color is %d bytes long.\n", length($str);
829 $colref3 = thaw($str);
830
831which prints (on my machine):
832
833 Blue is still 0.100000
834 Serialization of %color is 102 bytes long.
835
836=head1 WARNING
837
838If you're using references as keys within your hash tables, you're bound
f062ea6c 839to be disappointed when retrieving your data. Indeed, Perl stringifies
7a6a85bf
RG
840references used as hash table keys. If you later wish to access the
841items via another reference stringification (i.e. using the same
842reference that was used for the key originally to record the value into
843the hash table), it will work because both references stringify to the
844same string.
845
f062ea6c
PN
846It won't work across a C<store> and C<retrieve> operations, however,
847because the addresses in the retrieved objects, which are part of
848the stringified references, will probably differ from the original
849addresses. The topology of your structure is preserved, but not hidden
850semantics like those.
7a6a85bf
RG
851
852On platforms where it matters, be sure to call C<binmode()> on the
853descriptors that you pass to Storable functions.
854
855Storing data canonically that contains large hashes can be
856significantly slower than storing the same data normally, as
c261f00e 857temporary arrays to hold the keys for each hash have to be allocated,
7a6a85bf
RG
858populated, sorted and freed. Some tests have shown a halving of the
859speed of storing -- the exact penalty will depend on the complexity of
860your data. There is no slowdown on retrieval.
861
862=head1 BUGS
863
f062ea6c 864You can't store GLOB, CODE, FORMLINE, etc.... If you can define
7a6a85bf
RG
865semantics for those operations, feel free to enhance Storable so that
866it can deal with them.
867
868The store functions will C<croak> if they run into such references
869unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
870case, the fatal message is turned in a warning and some
871meaningless string is stored instead.
872
873Setting C<$Storable::canonical> may not yield frozen strings that
874compare equal due to possible stringification of numbers. When the
f062ea6c 875string version of a scalar exists, it is the form stored; therefore,
7a6a85bf
RG
876if you happen to use your numbers as strings between two freezing
877operations on the same data structures, you will get different
878results.
879
dd19458b
JH
880When storing doubles in network order, their value is stored as text.
881However, you should also not expect non-numeric floating-point values
882such as infinity and "not a number" to pass successfully through a
883nstore()/retrieve() pair.
884
885As Storable neither knows nor cares about character sets (although it
886does know that characters may be more than eight bits wide), any difference
887in the interpretation of character codes between a host and a target
888system is your problem. In particular, if host and target use different
889code points to represent the characters used in the text representation
890of floating-point numbers, you will not be able be able to exchange
891floating-point data, even with nstore().
892
c261f00e
NC
893C<Storable::drop_utf8> is a blunt tool. There is no facility either to
894return B<all> strings as utf8 sequences, or to attempt to convert utf8
895data back to 8 bit and C<croak()> if the conversion fails.
896
7a6a85bf
RG
897=head1 CREDITS
898
899Thank you to (in chronological order):
900
901 Jarkko Hietaniemi <jhi@iki.fi>
902 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
903 Benjamin A. Holzman <bah@ecnvantage.com>
904 Andrew Ford <A.Ford@ford-mason.co.uk>
905 Gisle Aas <gisle@aas.no>
906 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
907 Murray Nesbitt <murray@activestate.com>
908 Marc Lehmann <pcg@opengroup.org>
9e21b3d0
JH
909 Justin Banks <justinb@wamnet.com>
910 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
dd19458b
JH
911 Salvador Ortiz Garcia <sog@msg.com.mx>
912 Dominic Dunlop <domo@computer.org>
913 Erik Haugan <erik@solbors.no>
7a6a85bf
RG
914
915for their bug reports, suggestions and contributions.
916
917Benjamin Holzman contributed the tied variable support, Andrew Ford
918contributed the canonical order for hashes, and Gisle Aas fixed
f062ea6c 919a few misunderstandings of mine regarding the perl internals,
7a6a85bf
RG
920and optimized the emission of "tags" in the output streams by
921simply counting the objects instead of tagging them (leading to
922a binary incompatibility for the Storable image starting at version
f062ea6c 9230.6--older images are, of course, still properly understood).
7a6a85bf 924Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
f062ea6c 925and references to tied items support.
7a6a85bf 926
7a6a85bf
RG
927=head1 AUTHOR
928
0ba8809e 929Storable was written by Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>
775ecd75 930Maintenance is now done by the perl5-porters F<E<lt>perl5-porters@perl.orgE<gt>>
0ba8809e
NC
931
932Please e-mail us with problems, bug fixes, comments and complaints,
933although if you have complements you should send them to Raphael.
934Please don't e-mail Raphael with problems, as he no longer works on
935Storable, and your message will be delayed while he forwards it to us.
7a6a85bf
RG
936
937=head1 SEE ALSO
938
c261f00e 939L<Clone>.
7a6a85bf
RG
940
941=cut
942