This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add a line about 2.05.
[perl5.git] / ext / Storable / Storable.pm
CommitLineData
25f64a11
AMS
1#
2# Copyright (c) 1995-2000, Raphael Manfredi
3#
4# You may redistribute only under the same terms as Perl 5, as specified
5# in the README file that comes with the distribution.
6#
7a6a85bf
RG
7
8require DynaLoader;
9require Exporter;
10package Storable; @ISA = qw(Exporter DynaLoader);
11
12@EXPORT = qw(store retrieve);
13@EXPORT_OK = qw(
9e21b3d0 14 nstore store_fd nstore_fd fd_retrieve
7a6a85bf
RG
15 freeze nfreeze thaw
16 dclone
9e21b3d0 17 retrieve_fd
dd19458b 18 lock_store lock_nstore lock_retrieve
7a6a85bf
RG
19);
20
21use AutoLoader;
01d7b99e 22use vars qw($canonical $forgive_me $VERSION);
7a6a85bf 23
d2b96869 24$VERSION = '2.05';
7a6a85bf
RG
25*AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
26
27#
28# Use of Log::Agent is optional
29#
30
31eval "use Log::Agent";
32
530b72ba 33require Carp;
7a6a85bf 34
dd19458b
JH
35#
36# They might miss :flock in Fcntl
37#
38
39BEGIN {
596596d5 40 if (eval { require Fcntl; 1 } && exists $Fcntl::EXPORT_TAGS{'flock'}) {
dd19458b
JH
41 Fcntl->import(':flock');
42 } else {
43 eval q{
44 sub LOCK_SH () {1}
45 sub LOCK_EX () {2}
46 };
47 }
48}
49
b8778c7c 50# Can't Autoload cleanly as this clashes 8.3 with &retrieve
9e21b3d0 51sub retrieve_fd { &fd_retrieve } # Backward compatibility
cb3d9de5 52
530b72ba
NC
53# By default restricted hashes are downgraded on earlier perls.
54
55$Storable::downgrade_restricted = 1;
e8189732 56$Storable::accept_future_minor = 1;
b8778c7c
NC
57bootstrap Storable;
581;
59__END__
530b72ba
NC
60#
61# Use of Log::Agent is optional. If it hasn't imported these subs then
62# Autoloader will kindly supply our fallback implementation.
63#
64
65sub logcroak {
66 Carp::croak(@_);
67}
68
69sub logcarp {
70 Carp::carp(@_);
71}
b8778c7c 72
862382c7
JH
73#
74# Determine whether locking is possible, but only when needed.
75#
76
530b72ba 77sub CAN_FLOCK; my $CAN_FLOCK; sub CAN_FLOCK {
862382c7
JH
78 return $CAN_FLOCK if defined $CAN_FLOCK;
79 require Config; import Config;
80 return $CAN_FLOCK =
81 $Config{'d_flock'} ||
82 $Config{'d_fcntl_can_lock'} ||
83 $Config{'d_lockf'};
84}
85
0a0da639
JH
86sub show_file_magic {
87 print <<EOM;
88#
89# To recognize the data files of the Perl module Storable,
90# the following lines need to be added to the local magic(5) file,
91# usually either /usr/share/misc/magic or /etc/magic.
0a0da639
JH
92#
930 string perl-store perl Storable(v0.6) data
8b793558
JH
94>4 byte >0 (net-order %d)
95>>4 byte &01 (network-ordered)
96>>4 byte =3 (major 1)
97>>4 byte =2 (major 1)
98
0a0da639 990 string pst0 perl Storable(v0.7) data
8b793558
JH
100>4 byte >0
101>>4 byte &01 (network-ordered)
102>>4 byte =5 (major 2)
103>>4 byte =4 (major 2)
104>>5 byte >0 (minor %d)
0a0da639
JH
105EOM
106}
107
b8778c7c
NC
108sub read_magic {
109 my $header = shift;
110 return unless defined $header and length $header > 11;
111 my $result;
112 if ($header =~ s/^perl-store//) {
113 die "Can't deal with version 0 headers";
114 } elsif ($header =~ s/^pst0//) {
115 $result->{file} = 1;
116 }
117 # Assume it's a string.
118 my ($major, $minor, $bytelen) = unpack "C3", $header;
119
120 my $net_order = $major & 1;
121 $major >>= 1;
122 @$result{qw(major minor netorder)} = ($major, $minor, $net_order);
123
124 return $result if $net_order;
125
126 # I assume that it is rare to find v1 files, so this is an intentionally
127 # inefficient way of doing it, to make the rest of the code constant.
128 if ($major < 2) {
129 delete $result->{minor};
130 $header = '.' . $header;
131 $bytelen = $minor;
132 }
133
134 @$result{qw(byteorder intsize longsize ptrsize)} =
135 unpack "x3 A$bytelen C3", $header;
136
137 if ($major >= 2 and $minor >= 2) {
138 $result->{nvsize} = unpack "x6 x$bytelen C", $header;
139 }
140 $result;
141}
7a6a85bf
RG
142
143#
144# store
145#
146# Store target object hierarchy, identified by a reference to its root.
147# The stored object tree may later be retrieved to memory via retrieve.
148# Returns undef if an I/O error occurred, in which case the file is
149# removed.
150#
151sub store {
dd19458b 152 return _store(\&pstore, @_, 0);
7a6a85bf
RG
153}
154
155#
156# nstore
157#
158# Same as store, but in network order.
159#
160sub nstore {
dd19458b
JH
161 return _store(\&net_pstore, @_, 0);
162}
163
164#
165# lock_store
166#
167# Same as store, but flock the file first (advisory locking).
168#
169sub lock_store {
170 return _store(\&pstore, @_, 1);
171}
172
173#
174# lock_nstore
175#
176# Same as nstore, but flock the file first (advisory locking).
177#
178sub lock_nstore {
179 return _store(\&net_pstore, @_, 1);
7a6a85bf
RG
180}
181
182# Internal store to file routine
183sub _store {
184 my $xsptr = shift;
185 my $self = shift;
dd19458b 186 my ($file, $use_locking) = @_;
7a6a85bf 187 logcroak "not a reference" unless ref($self);
b12202d0 188 logcroak "wrong argument number" unless @_ == 2; # No @foo in arglist
7a6a85bf 189 local *FILE;
dd19458b 190 if ($use_locking) {
6e0ac6f5 191 open(FILE, ">>$file") || logcroak "can't write into $file: $!";
862382c7 192 unless (&CAN_FLOCK) {
b29b780f
RM
193 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
194 return undef;
f567092b 195 }
dd19458b
JH
196 flock(FILE, LOCK_EX) ||
197 logcroak "can't get exclusive lock on $file: $!";
198 truncate FILE, 0;
199 # Unlocking will happen when FILE is closed
6e0ac6f5
JH
200 } else {
201 open(FILE, ">$file") || logcroak "can't create $file: $!";
dd19458b 202 }
6e0ac6f5 203 binmode FILE; # Archaic systems...
7a6a85bf
RG
204 my $da = $@; # Don't mess if called from exception handler
205 my $ret;
206 # Call C routine nstore or pstore, depending on network order
207 eval { $ret = &$xsptr(*FILE, $self) };
208 close(FILE) or $ret = undef;
209 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
210 logcroak $@ if $@ =~ s/\.?\n$/,/;
211 $@ = $da;
212 return $ret ? $ret : undef;
213}
214
215#
216# store_fd
217#
218# Same as store, but perform on an already opened file descriptor instead.
219# Returns undef if an I/O error occurred.
220#
221sub store_fd {
222 return _store_fd(\&pstore, @_);
223}
224
225#
226# nstore_fd
227#
228# Same as store_fd, but in network order.
229#
230sub nstore_fd {
231 my ($self, $file) = @_;
232 return _store_fd(\&net_pstore, @_);
233}
234
235# Internal store routine on opened file descriptor
236sub _store_fd {
237 my $xsptr = shift;
238 my $self = shift;
239 my ($file) = @_;
240 logcroak "not a reference" unless ref($self);
241 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
242 my $fd = fileno($file);
243 logcroak "not a valid file descriptor" unless defined $fd;
244 my $da = $@; # Don't mess if called from exception handler
245 my $ret;
246 # Call C routine nstore or pstore, depending on network order
247 eval { $ret = &$xsptr($file, $self) };
248 logcroak $@ if $@ =~ s/\.?\n$/,/;
596596d5 249 local $\; print $file ''; # Autoflush the file if wanted
7a6a85bf
RG
250 $@ = $da;
251 return $ret ? $ret : undef;
252}
253
254#
255# freeze
256#
257# Store oject and its hierarchy in memory and return a scalar
258# containing the result.
259#
260sub freeze {
261 _freeze(\&mstore, @_);
262}
263
264#
265# nfreeze
266#
267# Same as freeze but in network order.
268#
269sub nfreeze {
270 _freeze(\&net_mstore, @_);
271}
272
273# Internal freeze routine
274sub _freeze {
275 my $xsptr = shift;
276 my $self = shift;
277 logcroak "not a reference" unless ref($self);
278 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
279 my $da = $@; # Don't mess if called from exception handler
280 my $ret;
281 # Call C routine mstore or net_mstore, depending on network order
282 eval { $ret = &$xsptr($self) };
283 logcroak $@ if $@ =~ s/\.?\n$/,/;
284 $@ = $da;
285 return $ret ? $ret : undef;
286}
287
288#
289# retrieve
290#
291# Retrieve object hierarchy from disk, returning a reference to the root
292# object of that tree.
293#
294sub retrieve {
dd19458b
JH
295 _retrieve($_[0], 0);
296}
297
298#
299# lock_retrieve
300#
301# Same as retrieve, but with advisory locking.
302#
303sub lock_retrieve {
304 _retrieve($_[0], 1);
305}
306
307# Internal retrieve routine
308sub _retrieve {
309 my ($file, $use_locking) = @_;
7a6a85bf 310 local *FILE;
dd19458b 311 open(FILE, $file) || logcroak "can't open $file: $!";
7a6a85bf
RG
312 binmode FILE; # Archaic systems...
313 my $self;
314 my $da = $@; # Could be from exception handler
dd19458b 315 if ($use_locking) {
862382c7 316 unless (&CAN_FLOCK) {
8be2b38b 317 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
b29b780f
RM
318 return undef;
319 }
8be2b38b 320 flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!";
dd19458b
JH
321 # Unlocking will happen when FILE is closed
322 }
7a6a85bf
RG
323 eval { $self = pretrieve(*FILE) }; # Call C routine
324 close(FILE);
325 logcroak $@ if $@ =~ s/\.?\n$/,/;
326 $@ = $da;
327 return $self;
328}
329
330#
9e21b3d0 331# fd_retrieve
7a6a85bf
RG
332#
333# Same as retrieve, but perform from an already opened file descriptor instead.
334#
9e21b3d0 335sub fd_retrieve {
7a6a85bf
RG
336 my ($file) = @_;
337 my $fd = fileno($file);
338 logcroak "not a valid file descriptor" unless defined $fd;
339 my $self;
340 my $da = $@; # Could be from exception handler
341 eval { $self = pretrieve($file) }; # Call C routine
342 logcroak $@ if $@ =~ s/\.?\n$/,/;
343 $@ = $da;
344 return $self;
345}
346
347#
348# thaw
349#
350# Recreate objects in memory from an existing frozen image created
351# by freeze. If the frozen image passed is undef, return undef.
352#
353sub thaw {
354 my ($frozen) = @_;
355 return undef unless defined $frozen;
356 my $self;
357 my $da = $@; # Could be from exception handler
358 eval { $self = mretrieve($frozen) }; # Call C routine
359 logcroak $@ if $@ =~ s/\.?\n$/,/;
360 $@ = $da;
361 return $self;
362}
363
364=head1 NAME
365
f062ea6c 366Storable - persistence for Perl data structures
7a6a85bf
RG
367
368=head1 SYNOPSIS
369
370 use Storable;
371 store \%table, 'file';
372 $hashref = retrieve('file');
373
374 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
375
376 # Network order
377 nstore \%table, 'file';
378 $hashref = retrieve('file'); # There is NO nretrieve()
379
380 # Storing to and retrieving from an already opened file
381 store_fd \@array, \*STDOUT;
382 nstore_fd \%table, \*STDOUT;
9e21b3d0
JH
383 $aryref = fd_retrieve(\*SOCKET);
384 $hashref = fd_retrieve(\*SOCKET);
7a6a85bf
RG
385
386 # Serializing to memory
387 $serialized = freeze \%table;
388 %table_clone = %{ thaw($serialized) };
389
390 # Deep (recursive) cloning
391 $cloneref = dclone($ref);
392
dd19458b
JH
393 # Advisory locking
394 use Storable qw(lock_store lock_nstore lock_retrieve)
395 lock_store \%table, 'file';
396 lock_nstore \%table, 'file';
397 $hashref = lock_retrieve('file');
398
7a6a85bf
RG
399=head1 DESCRIPTION
400
f062ea6c 401The Storable package brings persistence to your Perl data structures
7a6a85bf 402containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
c261f00e 403conveniently stored to disk and retrieved at a later time.
7a6a85bf
RG
404
405It can be used in the regular procedural way by calling C<store> with
406a reference to the object to be stored, along with the file name where
407the image should be written.
775ecd75 408
7a6a85bf
RG
409The routine returns C<undef> for I/O problems or other internal error,
410a true value otherwise. Serious errors are propagated as a C<die> exception.
411
f062ea6c
PN
412To retrieve data stored to disk, use C<retrieve> with a file name.
413The objects stored into that file are recreated into memory for you,
414and a I<reference> to the root object is returned. In case an I/O error
7a6a85bf
RG
415occurs while reading, C<undef> is returned instead. Other serious
416errors are propagated via C<die>.
417
418Since storage is performed recursively, you might want to stuff references
419to objects that share a lot of common data into a single array or hash
420table, and then store that object. That way, when you retrieve back the
421whole thing, the objects will continue to share what they originally shared.
422
423At the cost of a slight header overhead, you may store to an already
424opened file descriptor using the C<store_fd> routine, and retrieve
9e21b3d0 425from a file via C<fd_retrieve>. Those names aren't imported by default,
c261f00e 426so you will have to do that explicitly if you need those routines.
7a6a85bf
RG
427The file descriptor you supply must be already opened, for read
428if you're going to retrieve and for write if you wish to store.
429
430 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
9e21b3d0 431 $hashref = fd_retrieve(*STDIN);
7a6a85bf
RG
432
433You can also store data in network order to allow easy sharing across
434multiple platforms, or when storing on a socket known to be remotely
435connected. The routines to call have an initial C<n> prefix for I<network>,
436as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
437correctly restored so you don't have to know whether you're restoring
dd19458b
JH
438from native or network ordered data. Double values are stored stringified
439to ensure portability as well, at the slight risk of loosing some precision
440in the last decimals.
7a6a85bf 441
9e21b3d0 442When using C<fd_retrieve>, objects are retrieved in sequence, one
7a6a85bf
RG
443object (i.e. one recursive tree) per associated C<store_fd>.
444
445If you're more from the object-oriented camp, you can inherit from
446Storable and directly store your objects by invoking C<store> as
447a method. The fact that the root of the to-be-stored tree is a
448blessed reference (i.e. an object) is special-cased so that the
449retrieve does not provide a reference to that object but rather the
450blessed object reference itself. (Otherwise, you'd get a reference
451to that blessed object).
452
453=head1 MEMORY STORE
454
455The Storable engine can also store data into a Perl scalar instead, to
456later retrieve them. This is mainly used to freeze a complex structure in
457some safe compact memory place (where it can possibly be sent to another
458process via some IPC, since freezing the structure also serializes it in
459effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
460out and recreate the original complex structure in memory.
461
462Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
463If you wish to send out the frozen scalar to another machine, use
464C<nfreeze> instead to get a portable image.
465
466Note that freezing an object structure and immediately thawing it
467actually achieves a deep cloning of that structure:
468
469 dclone(.) = thaw(freeze(.))
470
471Storable provides you with a C<dclone> interface which does not create
472that intermediary scalar but instead freezes the structure in some
c261f00e 473internal memory space and then immediately thaws it out.
7a6a85bf 474
dd19458b
JH
475=head1 ADVISORY LOCKING
476
f062ea6c
PN
477The C<lock_store> and C<lock_nstore> routine are equivalent to
478C<store> and C<nstore>, except that they get an exclusive lock on
479the file before writing. Likewise, C<lock_retrieve> does the same
480as C<retrieve>, but also gets a shared lock on the file before reading.
dd19458b 481
f062ea6c
PN
482As with any advisory locking scheme, the protection only works if you
483systematically use C<lock_store> and C<lock_retrieve>. If one side of
484your application uses C<store> whilst the other uses C<lock_retrieve>,
dd19458b
JH
485you will get no protection at all.
486
f062ea6c
PN
487The internal advisory locking is implemented using Perl's flock()
488routine. If your system does not support any form of flock(), or if
489you share your files across NFS, you might wish to use other forms
490of locking by using modules such as LockFile::Simple which lock a
491file using a filesystem entry, instead of locking the file descriptor.
dd19458b 492
7a6a85bf
RG
493=head1 SPEED
494
495The heart of Storable is written in C for decent speed. Extra low-level
4d3295e3
PN
496optimizations have been made when manipulating perl internals, to
497sacrifice encapsulation for the benefit of greater speed.
7a6a85bf
RG
498
499=head1 CANONICAL REPRESENTATION
500
f062ea6c 501Normally, Storable stores elements of hashes in the order they are
7a6a85bf
RG
502stored internally by Perl, i.e. pseudo-randomly. If you set
503C<$Storable::canonical> to some C<TRUE> value, Storable will store
504hashes with the elements sorted by their key. This allows you to
505compare data structures by comparing their frozen representations (or
506even the compressed frozen representations), which can be useful for
507creating lookup tables for complicated queries.
508
f062ea6c 509Canonical order does not imply network order; those are two orthogonal
7a6a85bf
RG
510settings.
511
d2b96869
SR
512=head1 CODE REFERENCES
513
514Since Storable version 2.05, CODE references may be serialized with
515the help of L<B::Deparse>. To enable this feature, set
516C<$Storable::Deparse> to a true value. To enable deserializazion,
517C<$Storable::Eval> should be set to a true value. Be aware that
518deserialization is done through C<eval>, which is dangerous if the
519Storable file contains malicious data. You can set C<$Storable::Eval>
520to a subroutine reference which would be used instead of C<eval>. See
521below for an example using a L<Safe> compartment for deserialization
522of CODE references.
523
c261f00e
NC
524=head1 FORWARD COMPATIBILITY
525
526This release of Storable can be used on a newer version of Perl to
f062ea6c 527serialize data which is not supported by earlier Perls. By default,
c261f00e 528Storable will attempt to do the right thing, by C<croak()>ing if it
775ecd75 529encounters data that it cannot deserialize. However, the defaults
f062ea6c 530can be changed as follows:
c261f00e
NC
531
532=over 4
533
534=item utf8 data
535
536Perl 5.6 added support for Unicode characters with code points > 255,
537and Perl 5.8 has full support for Unicode characters in hash keys.
538Perl internally encodes strings with these characters using utf8, and
539Storable serializes them as utf8. By default, if an older version of
540Perl encounters a utf8 value it cannot represent, it will C<croak()>.
541To change this behaviour so that Storable deserializes utf8 encoded
542values as the string of bytes (effectively dropping the I<is_utf8> flag)
543set C<$Storable::drop_utf8> to some C<TRUE> value. This is a form of
544data loss, because with C<$drop_utf8> true, it becomes impossible to tell
545whether the original data was the Unicode string, or a series of bytes
546that happen to be valid utf8.
547
548=item restricted hashes
549
f062ea6c
PN
550Perl 5.8 adds support for restricted hashes, which have keys
551restricted to a given set, and can have values locked to be read only.
552By default, when Storable encounters a restricted hash on a perl
553that doesn't support them, it will deserialize it as a normal hash,
554silently discarding any placeholder keys and leaving the keys and
555all values unlocked. To make Storable C<croak()> instead, set
556C<$Storable::downgrade_restricted> to a C<FALSE> value. To restore
557the default set it back to some C<TRUE> value.
c261f00e 558
e8189732
NC
559=item files from future versions of Storable
560
561Earlier versions of Storable would immediately croak if they encountered
562a file with a higher internal version number than the reading Storable
563knew about. Internal version numbers are increased each time new data
564types (such as restricted hashes) are added to the vocabulary of the file
565format. This meant that a newer Storable module had no way of writing a
f062ea6c 566file readable by an older Storable, even if the writer didn't store newer
e8189732
NC
567data types.
568
569This version of Storable will defer croaking until it encounters a data
570type in the file that it does not recognize. This means that it will
571continue to read files generated by newer Storable modules which are careful
572in what they write out, making it easier to upgrade Storable modules in a
573mixed environment.
574
575The old behaviour of immediate croaking can be re-instated by setting
f062ea6c 576C<$Storable::accept_future_minor> to some C<FALSE> value.
e8189732 577
c261f00e
NC
578=back
579
f062ea6c 580All these variables have no effect on a newer Perl which supports the
c261f00e
NC
581relevant feature.
582
7a6a85bf
RG
583=head1 ERROR REPORTING
584
585Storable uses the "exception" paradigm, in that it does not try to workaround
586failures: if something bad happens, an exception is generated from the
587caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
588those exceptions.
589
590When Storable croaks, it tries to report the error via the C<logcroak()>
591routine from the C<Log::Agent> package, if it is available.
592
212e9bde
JH
593Normal errors are reported by having store() or retrieve() return C<undef>.
594Such errors are usually I/O errors (or truncated stream errors at retrieval).
595
7a6a85bf
RG
596=head1 WIZARDS ONLY
597
598=head2 Hooks
599
600Any class may define hooks that will be called during the serialization
601and deserialization process on objects that are instances of that class.
602Those hooks can redefine the way serialization is performed (and therefore,
c261f00e 603how the symmetrical deserialization should be conducted).
7a6a85bf
RG
604
605Since we said earlier:
606
607 dclone(.) = thaw(freeze(.))
608
609everything we say about hooks should also hold for deep cloning. However,
610hooks get to know whether the operation is a mere serialization, or a cloning.
611
612Therefore, when serializing hooks are involved,
613
614 dclone(.) <> thaw(freeze(.))
615
616Well, you could keep them in sync, but there's no guarantee it will always
617hold on classes somebody else wrote. Besides, there is little to gain in
f062ea6c 618doing so: a serializing hook could keep only one attribute of an object,
7a6a85bf
RG
619which is probably not what should happen during a deep cloning of that
620same object.
621
622Here is the hooking interface:
623
bbc7dcd2 624=over 4
7a6a85bf
RG
625
626=item C<STORABLE_freeze> I<obj>, I<cloning>
627
628The serializing hook, called on the object during serialization. It can be
629inherited, or defined in the class itself, like any other method.
630
631Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
632whether we're in a dclone() or a regular serialization via store() or freeze().
633
634Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
635is the serialized form to be used, and the optional $ref1, $ref2, etc... are
636extra references that you wish to let the Storable engine serialize.
637
638At deserialization time, you will be given back the same LIST, but all the
639extra references will be pointing into the deserialized structure.
640
641The B<first time> the hook is hit in a serialization flow, you may have it
642return an empty list. That will signal the Storable engine to further
643discard that hook for this class and to therefore revert to the default
644serialization of the underlying Perl data. The hook will again be normally
645processed in the next serialization.
646
647Unless you know better, serializing hook should always say:
648
649 sub STORABLE_freeze {
650 my ($self, $cloning) = @_;
651 return if $cloning; # Regular default serialization
652 ....
653 }
654
655in order to keep reasonable dclone() semantics.
656
657=item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
658
659The deserializing hook called on the object during deserialization.
f062ea6c 660But wait: if we're deserializing, there's no object yet... right?
7a6a85bf
RG
661
662Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
663you can view C<STORABLE_thaw> as an alternate creation routine.
664
665This means the hook can be inherited like any other method, and that
666I<obj> is your blessed reference for this particular instance.
667
668The other arguments should look familiar if you know C<STORABLE_freeze>:
669I<cloning> is true when we're part of a deep clone operation, I<serialized>
670is the serialized string you returned to the engine in C<STORABLE_freeze>,
671and there may be an optional list of references, in the same order you gave
672them at serialization time, pointing to the deserialized objects (which
673have been processed courtesy of the Storable engine).
674
212e9bde
JH
675When the Storable engine does not find any C<STORABLE_thaw> hook routine,
676it tries to load the class by requiring the package dynamically (using
677the blessed package name), and then re-attempts the lookup. If at that
678time the hook cannot be located, the engine croaks. Note that this mechanism
c261f00e 679will fail if you define several classes in the same file, but L<perlmod>
212e9bde
JH
680warned you.
681
f062ea6c 682It is up to you to use this information to populate I<obj> the way you want.
7a6a85bf
RG
683
684Returned value: none.
685
686=back
687
688=head2 Predicates
689
c261f00e 690Predicates are not exportable. They must be called by explicitly prefixing
7a6a85bf
RG
691them with the Storable package name.
692
bbc7dcd2 693=over 4
7a6a85bf
RG
694
695=item C<Storable::last_op_in_netorder>
696
697The C<Storable::last_op_in_netorder()> predicate will tell you whether
698network order was used in the last store or retrieve operation. If you
699don't know how to use this, just forget about it.
700
701=item C<Storable::is_storing>
702
703Returns true if within a store operation (via STORABLE_freeze hook).
704
705=item C<Storable::is_retrieving>
706
f062ea6c 707Returns true if within a retrieve operation (via STORABLE_thaw hook).
7a6a85bf
RG
708
709=back
710
711=head2 Recursion
712
f062ea6c
PN
713With hooks comes the ability to recurse back to the Storable engine.
714Indeed, hooks are regular Perl code, and Storable is convenient when
715it comes to serializing and deserializing things, so why not use it
716to handle the serialization string?
7a6a85bf 717
f062ea6c 718There are a few things you need to know, however:
7a6a85bf 719
bbc7dcd2 720=over 4
7a6a85bf
RG
721
722=item *
723
724You can create endless loops if the things you serialize via freeze()
f062ea6c
PN
725(for instance) point back to the object we're trying to serialize in
726the hook.
7a6a85bf
RG
727
728=item *
729
730Shared references among objects will not stay shared: if we're serializing
731the list of object [A, C] where both object A and C refer to the SAME object
732B, and if there is a serializing hook in A that says freeze(B), then when
733deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
734a deep clone of B'. The topology was not preserved.
735
736=back
737
738That's why C<STORABLE_freeze> lets you provide a list of references
739to serialize. The engine guarantees that those will be serialized in the
740same context as the other objects, and therefore that shared objects will
741stay shared.
742
743In the above [A, C] example, the C<STORABLE_freeze> hook could return:
744
745 ("something", $self->{B})
746
747and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
748would get back the reference to the B' object, deserialized for you.
749
750Therefore, recursion should normally be avoided, but is nonetheless supported.
751
752=head2 Deep Cloning
753
f062ea6c 754There is a Clone module available on CPAN which implements deep cloning
7a6a85bf
RG
755natively, i.e. without freezing to memory and thawing the result. It is
756aimed to replace Storable's dclone() some day. However, it does not currently
757support Storable hooks to redefine the way deep cloning is performed.
758
0a0da639
JH
759=head1 Storable magic
760
761Yes, there's a lot of that :-) But more precisely, in UNIX systems
762there's a utility called C<file>, which recognizes data files based on
763their contents (usually their first few bytes). For this to work,
8b793558 764a certain file called F<magic> needs to taught about the I<signature>
0a0da639 765of the data. Where that configuration file lives depends on the UNIX
f062ea6c 766flavour; often it's something like F</usr/share/misc/magic> or
8b793558
JH
767F</etc/magic>. Your system administrator needs to do the updating of
768the F<magic> file. The necessary signature information is output to
f062ea6c
PN
769STDOUT by invoking Storable::show_file_magic(). Note that the GNU
770implementation of the C<file> utility, version 3.38 or later,
771is expected to contain support for recognising Storable files
772out-of-the-box, in addition to other kinds of Perl files.
0a0da639 773
7a6a85bf
RG
774=head1 EXAMPLES
775
776Here are some code samples showing a possible usage of Storable:
777
778 use Storable qw(store retrieve freeze thaw dclone);
779
780 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
781
782 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
783
784 $colref = retrieve('/tmp/colors');
785 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
786 printf "Blue is still %lf\n", $colref->{'Blue'};
787
788 $colref2 = dclone(\%color);
789
790 $str = freeze(\%color);
791 printf "Serialization of %%color is %d bytes long.\n", length($str);
792 $colref3 = thaw($str);
793
794which prints (on my machine):
795
796 Blue is still 0.100000
797 Serialization of %color is 102 bytes long.
798
d2b96869
SR
799Serialization of CODE references and deserialization in a safe
800compartment:
801
802 use Storable qw(freeze thaw);
803 use Safe;
804 use strict;
805 my $safe = new Safe;
806 # permitting the "require" opcode is necessary when using "use strict"
807 $safe->permit(qw(:default require));
808 local $Storable::Deparse = 1;
809 local $Storable::Eval = sub { $safe->reval($_[0]) };
810 my $serialized = freeze(sub { print "42\n" });
811 my $code = thaw($serialized);
812 $code->(); # prints 42
813
7a6a85bf
RG
814=head1 WARNING
815
816If you're using references as keys within your hash tables, you're bound
f062ea6c 817to be disappointed when retrieving your data. Indeed, Perl stringifies
7a6a85bf
RG
818references used as hash table keys. If you later wish to access the
819items via another reference stringification (i.e. using the same
820reference that was used for the key originally to record the value into
821the hash table), it will work because both references stringify to the
822same string.
823
6fe6778b
PN
824It won't work across a sequence of C<store> and C<retrieve> operations,
825however, because the addresses in the retrieved objects, which are
826part of the stringified references, will probably differ from the
827original addresses. The topology of your structure is preserved,
828but not hidden semantics like those.
7a6a85bf
RG
829
830On platforms where it matters, be sure to call C<binmode()> on the
831descriptors that you pass to Storable functions.
832
833Storing data canonically that contains large hashes can be
834significantly slower than storing the same data normally, as
c261f00e 835temporary arrays to hold the keys for each hash have to be allocated,
7a6a85bf
RG
836populated, sorted and freed. Some tests have shown a halving of the
837speed of storing -- the exact penalty will depend on the complexity of
838your data. There is no slowdown on retrieval.
839
840=head1 BUGS
841
f062ea6c 842You can't store GLOB, CODE, FORMLINE, etc.... If you can define
7a6a85bf
RG
843semantics for those operations, feel free to enhance Storable so that
844it can deal with them.
845
846The store functions will C<croak> if they run into such references
847unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
848case, the fatal message is turned in a warning and some
849meaningless string is stored instead.
850
851Setting C<$Storable::canonical> may not yield frozen strings that
852compare equal due to possible stringification of numbers. When the
f062ea6c 853string version of a scalar exists, it is the form stored; therefore,
7a6a85bf
RG
854if you happen to use your numbers as strings between two freezing
855operations on the same data structures, you will get different
856results.
857
dd19458b
JH
858When storing doubles in network order, their value is stored as text.
859However, you should also not expect non-numeric floating-point values
860such as infinity and "not a number" to pass successfully through a
861nstore()/retrieve() pair.
862
863As Storable neither knows nor cares about character sets (although it
864does know that characters may be more than eight bits wide), any difference
865in the interpretation of character codes between a host and a target
866system is your problem. In particular, if host and target use different
867code points to represent the characters used in the text representation
868of floating-point numbers, you will not be able be able to exchange
869floating-point data, even with nstore().
870
c261f00e
NC
871C<Storable::drop_utf8> is a blunt tool. There is no facility either to
872return B<all> strings as utf8 sequences, or to attempt to convert utf8
873data back to 8 bit and C<croak()> if the conversion fails.
874
ee0f7aac
NC
875Prior to Storable 2.01, no distinction was made between signed and
876unsigned integers on storing. By default Storable prefers to store a
877scalars string representation (if it has one) so this would only cause
878problems when storing large unsigned integers that had never been coverted
879to string or floating point. In other words values that had been generated
880by integer operations such as logic ops and then not used in any string or
881arithmetic context before storing.
882
883=head2 64 bit data in perl 5.6.0 and 5.6.1
884
885This section only applies to you if you have existing data written out
886by Storable 2.02 or earlier on perl 5.6.0 or 5.6.1 on Unix or Linux which
887has been configured with 64 bit integer support (not the default)
888If you got a precompiled perl, rather than running Configure to build
889your own perl from source, then it almost certainly does not affect you,
890and you can stop reading now (unless you're curious). If you're using perl
891on Windows it does not affect you.
892
893Storable writes a file header which contains the sizes of various C
894language types for the C compiler that built Storable (when not writing in
895network order), and will refuse to load files written by a Storable not
896on the same (or compatible) architecture. This check and a check on
897machine byteorder is needed because the size of various fields in the file
898are given by the sizes of the C language types, and so files written on
899different architectures are incompatible. This is done for increased speed.
900(When writing in network order, all fields are written out as standard
901lengths, which allows full interworking, but takes longer to read and write)
902
903Perl 5.6.x introduced the ability to optional configure the perl interpreter
904to use C's C<long long> type to allow scalars to store 64 bit integers on 32
905bit systems. However, due to the way the Perl configuration system
906generated the C configuration files on non-Windows platforms, and the way
907Storable generates its header, nothing in the Storable file header reflected
908whether the perl writing was using 32 or 64 bit integers, despite the fact
909that Storable was storing some data differently in the file. Hence Storable
910running on perl with 64 bit integers will read the header from a file
911written by a 32 bit perl, not realise that the data is actually in a subtly
912incompatible format, and then go horribly wrong (possibly crashing) if it
913encountered a stored integer. This is a design failure.
914
915Storable has now been changed to write out and read in a file header with
916information about the size of integers. It's impossible to detect whether
917an old file being read in was written with 32 or 64 bit integers (they have
918the same header) so it's impossible to automatically switch to a correct
919backwards compatibility mode. Hence this Storable defaults to the new,
920correct behaviour.
921
922What this means is that if you have data written by Storable 1.x running
923on perl 5.6.0 or 5.6.1 configured with 64 bit integers on Unix or Linux
924then by default this Storable will refuse to read it, giving the error
925I<Byte order is not compatible>. If you have such data then you you
926should set C<$Storable::interwork_56_64bit> to a true value to make this
927Storable read and write files with the old header. You should also
928migrate your data, or any older perl you are communicating with, to this
929current version of Storable.
930
931If you don't have data written with specific configuration of perl described
932above, then you do not and should not do anything. Don't set the flag -
933not only will Storable on an identically configured perl refuse to load them,
934but Storable a differently configured perl will load them believing them
935to be correct for it, and then may well fail or crash part way through
936reading them.
937
7a6a85bf
RG
938=head1 CREDITS
939
940Thank you to (in chronological order):
941
942 Jarkko Hietaniemi <jhi@iki.fi>
943 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
944 Benjamin A. Holzman <bah@ecnvantage.com>
945 Andrew Ford <A.Ford@ford-mason.co.uk>
946 Gisle Aas <gisle@aas.no>
947 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
948 Murray Nesbitt <murray@activestate.com>
949 Marc Lehmann <pcg@opengroup.org>
9e21b3d0
JH
950 Justin Banks <justinb@wamnet.com>
951 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
dd19458b
JH
952 Salvador Ortiz Garcia <sog@msg.com.mx>
953 Dominic Dunlop <domo@computer.org>
954 Erik Haugan <erik@solbors.no>
7a6a85bf
RG
955
956for their bug reports, suggestions and contributions.
957
958Benjamin Holzman contributed the tied variable support, Andrew Ford
959contributed the canonical order for hashes, and Gisle Aas fixed
f062ea6c 960a few misunderstandings of mine regarding the perl internals,
7a6a85bf
RG
961and optimized the emission of "tags" in the output streams by
962simply counting the objects instead of tagging them (leading to
963a binary incompatibility for the Storable image starting at version
f062ea6c 9640.6--older images are, of course, still properly understood).
7a6a85bf 965Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
f062ea6c 966and references to tied items support.
7a6a85bf 967
7a6a85bf
RG
968=head1 AUTHOR
969
0ba8809e 970Storable was written by Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>
775ecd75 971Maintenance is now done by the perl5-porters F<E<lt>perl5-porters@perl.orgE<gt>>
0ba8809e
NC
972
973Please e-mail us with problems, bug fixes, comments and complaints,
974although if you have complements you should send them to Raphael.
975Please don't e-mail Raphael with problems, as he no longer works on
976Storable, and your message will be delayed while he forwards it to us.
7a6a85bf
RG
977
978=head1 SEE ALSO
979
c261f00e 980L<Clone>.
7a6a85bf
RG
981
982=cut