1 ;# $Id: Storable.pm,v 1.0.1.8 2001/02/17 12:24:37 ram Exp $
3 ;# Copyright (c) 1995-2000, Raphael Manfredi
5 ;# You may redistribute only under the same terms as Perl 5, as specified
6 ;# in the README file that comes with the distribution.
8 ;# $Log: Storable.pm,v $
9 ;# Revision 1.0.1.8 2001/02/17 12:24:37 ram
10 ;# patch8: fixed incorrect error message
12 ;# Revision 1.0.1.7 2001/01/03 09:39:02 ram
13 ;# patch7: added CAN_FLOCK to determine whether we can flock() or not
15 ;# Revision 1.0.1.6 2000/11/05 17:20:25 ram
16 ;# patch6: increased version number
18 ;# Revision 1.0.1.5 2000/10/26 17:10:18 ram
19 ;# patch5: documented that store() and retrieve() can return undef
20 ;# patch5: added paragraph explaining the auto require for thaw hooks
22 ;# Revision 1.0.1.4 2000/10/23 18:02:57 ram
23 ;# patch4: protected calls to flock() for dos platform
24 ;# patch4: added logcarp emulation if they don't have Log::Agent
26 ;# $Log: Storable.pm,v $
27 ;# Revision 1.0 2000/09/01 19:40:41 ram
28 ;# Baseline for first official release.
33 package Storable; @ISA = qw(Exporter DynaLoader);
35 @EXPORT = qw(store retrieve);
37 nstore store_fd nstore_fd fd_retrieve
41 lock_store lock_nstore lock_retrieve
45 use vars qw($forgive_me $VERSION);
48 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
51 # Use of Log::Agent is optional
54 eval "use Log::Agent";
56 unless (defined @Log::Agent::EXPORT) {
70 # They might miss :flock in Fcntl
75 if (exists $Fcntl::EXPORT_TAGS{'flock'}) {
76 Fcntl->import(':flock');
88 sub retrieve_fd { &fd_retrieve } # Backward compatibility
91 # Determine whether locking is possible, but only when needed.
97 return $CAN_FLOCK if defined $CAN_FLOCK;
98 require Config; import Config;
100 $Config{'d_flock'} ||
101 $Config{'d_fcntl_can_lock'} ||
112 # Store target object hierarchy, identified by a reference to its root.
113 # The stored object tree may later be retrieved to memory via retrieve.
114 # Returns undef if an I/O error occurred, in which case the file is
118 return _store(\&pstore, @_, 0);
124 # Same as store, but in network order.
127 return _store(\&net_pstore, @_, 0);
133 # Same as store, but flock the file first (advisory locking).
136 return _store(\&pstore, @_, 1);
142 # Same as nstore, but flock the file first (advisory locking).
145 return _store(\&net_pstore, @_, 1);
148 # Internal store to file routine
152 my ($file, $use_locking) = @_;
153 logcroak "not a reference" unless ref($self);
154 logcroak "wrong argument number" unless @_ == 2; # No @foo in arglist
156 open(FILE, ">$file") || logcroak "can't create $file: $!";
157 binmode FILE; # Archaic systems...
159 unless (&CAN_FLOCK) {
160 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
163 flock(FILE, LOCK_EX) ||
164 logcroak "can't get exclusive lock on $file: $!";
166 # Unlocking will happen when FILE is closed
168 my $da = $@; # Don't mess if called from exception handler
170 # Call C routine nstore or pstore, depending on network order
171 eval { $ret = &$xsptr(*FILE, $self) };
172 close(FILE) or $ret = undef;
173 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
174 logcroak $@ if $@ =~ s/\.?\n$/,/;
176 return $ret ? $ret : undef;
182 # Same as store, but perform on an already opened file descriptor instead.
183 # Returns undef if an I/O error occurred.
186 return _store_fd(\&pstore, @_);
192 # Same as store_fd, but in network order.
195 my ($self, $file) = @_;
196 return _store_fd(\&net_pstore, @_);
199 # Internal store routine on opened file descriptor
204 logcroak "not a reference" unless ref($self);
205 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
206 my $fd = fileno($file);
207 logcroak "not a valid file descriptor" unless defined $fd;
208 my $da = $@; # Don't mess if called from exception handler
210 # Call C routine nstore or pstore, depending on network order
211 eval { $ret = &$xsptr($file, $self) };
212 logcroak $@ if $@ =~ s/\.?\n$/,/;
214 return $ret ? $ret : undef;
220 # Store oject and its hierarchy in memory and return a scalar
221 # containing the result.
224 _freeze(\&mstore, @_);
230 # Same as freeze but in network order.
233 _freeze(\&net_mstore, @_);
236 # Internal freeze routine
240 logcroak "not a reference" unless ref($self);
241 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
242 my $da = $@; # Don't mess if called from exception handler
244 # Call C routine mstore or net_mstore, depending on network order
245 eval { $ret = &$xsptr($self) };
246 logcroak $@ if $@ =~ s/\.?\n$/,/;
248 return $ret ? $ret : undef;
254 # Retrieve object hierarchy from disk, returning a reference to the root
255 # object of that tree.
264 # Same as retrieve, but with advisory locking.
270 # Internal retrieve routine
272 my ($file, $use_locking) = @_;
274 open(FILE, $file) || logcroak "can't open $file: $!";
275 binmode FILE; # Archaic systems...
277 my $da = $@; # Could be from exception handler
279 unless (&CAN_FLOCK) {
280 logcarp "Storable::lock_retrieve: fcntl/flock emulation broken on $^O";
283 flock(FILE, LOCK_SH) ||
284 logcroak "can't get shared lock on $file: $!";
285 # Unlocking will happen when FILE is closed
287 eval { $self = pretrieve(*FILE) }; # Call C routine
289 logcroak $@ if $@ =~ s/\.?\n$/,/;
297 # Same as retrieve, but perform from an already opened file descriptor instead.
301 my $fd = fileno($file);
302 logcroak "not a valid file descriptor" unless defined $fd;
304 my $da = $@; # Could be from exception handler
305 eval { $self = pretrieve($file) }; # Call C routine
306 logcroak $@ if $@ =~ s/\.?\n$/,/;
314 # Recreate objects in memory from an existing frozen image created
315 # by freeze. If the frozen image passed is undef, return undef.
319 return undef unless defined $frozen;
321 my $da = $@; # Could be from exception handler
322 eval { $self = mretrieve($frozen) }; # Call C routine
323 logcroak $@ if $@ =~ s/\.?\n$/,/;
330 Storable - persistency for perl data structures
335 store \%table, 'file';
336 $hashref = retrieve('file');
338 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
341 nstore \%table, 'file';
342 $hashref = retrieve('file'); # There is NO nretrieve()
344 # Storing to and retrieving from an already opened file
345 store_fd \@array, \*STDOUT;
346 nstore_fd \%table, \*STDOUT;
347 $aryref = fd_retrieve(\*SOCKET);
348 $hashref = fd_retrieve(\*SOCKET);
350 # Serializing to memory
351 $serialized = freeze \%table;
352 %table_clone = %{ thaw($serialized) };
354 # Deep (recursive) cloning
355 $cloneref = dclone($ref);
358 use Storable qw(lock_store lock_nstore lock_retrieve)
359 lock_store \%table, 'file';
360 lock_nstore \%table, 'file';
361 $hashref = lock_retrieve('file');
365 The Storable package brings persistency to your perl data structures
366 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
367 convenientely stored to disk and retrieved at a later time.
369 It can be used in the regular procedural way by calling C<store> with
370 a reference to the object to be stored, along with the file name where
371 the image should be written.
372 The routine returns C<undef> for I/O problems or other internal error,
373 a true value otherwise. Serious errors are propagated as a C<die> exception.
375 To retrieve data stored to disk, use C<retrieve> with a file name,
376 and the objects stored into that file are recreated into memory for you,
377 a I<reference> to the root object being returned. In case an I/O error
378 occurs while reading, C<undef> is returned instead. Other serious
379 errors are propagated via C<die>.
381 Since storage is performed recursively, you might want to stuff references
382 to objects that share a lot of common data into a single array or hash
383 table, and then store that object. That way, when you retrieve back the
384 whole thing, the objects will continue to share what they originally shared.
386 At the cost of a slight header overhead, you may store to an already
387 opened file descriptor using the C<store_fd> routine, and retrieve
388 from a file via C<fd_retrieve>. Those names aren't imported by default,
389 so you will have to do that explicitely if you need those routines.
390 The file descriptor you supply must be already opened, for read
391 if you're going to retrieve and for write if you wish to store.
393 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
394 $hashref = fd_retrieve(*STDIN);
396 You can also store data in network order to allow easy sharing across
397 multiple platforms, or when storing on a socket known to be remotely
398 connected. The routines to call have an initial C<n> prefix for I<network>,
399 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
400 correctly restored so you don't have to know whether you're restoring
401 from native or network ordered data. Double values are stored stringified
402 to ensure portability as well, at the slight risk of loosing some precision
403 in the last decimals.
405 When using C<fd_retrieve>, objects are retrieved in sequence, one
406 object (i.e. one recursive tree) per associated C<store_fd>.
408 If you're more from the object-oriented camp, you can inherit from
409 Storable and directly store your objects by invoking C<store> as
410 a method. The fact that the root of the to-be-stored tree is a
411 blessed reference (i.e. an object) is special-cased so that the
412 retrieve does not provide a reference to that object but rather the
413 blessed object reference itself. (Otherwise, you'd get a reference
414 to that blessed object).
418 The Storable engine can also store data into a Perl scalar instead, to
419 later retrieve them. This is mainly used to freeze a complex structure in
420 some safe compact memory place (where it can possibly be sent to another
421 process via some IPC, since freezing the structure also serializes it in
422 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
423 out and recreate the original complex structure in memory.
425 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
426 If you wish to send out the frozen scalar to another machine, use
427 C<nfreeze> instead to get a portable image.
429 Note that freezing an object structure and immediately thawing it
430 actually achieves a deep cloning of that structure:
432 dclone(.) = thaw(freeze(.))
434 Storable provides you with a C<dclone> interface which does not create
435 that intermediary scalar but instead freezes the structure in some
436 internal memory space and then immediatly thaws it out.
438 =head1 ADVISORY LOCKING
440 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
441 and C<nstore>, only they get an exclusive lock on the file before
442 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
443 gets a shared lock on the file before reading.
445 Like with any advisory locking scheme, the protection only works if
446 you systematically use C<lock_store> and C<lock_retrieve>. If one
447 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
448 you will get no protection at all.
450 The internal advisory locking is implemented using Perl's flock() routine.
451 If your system does not support any form of flock(), or if you share
452 your files across NFS, you might wish to use other forms of locking by
453 using modules like LockFile::Simple which lock a file using a filesystem
454 entry, instead of locking the file descriptor.
458 The heart of Storable is written in C for decent speed. Extra low-level
459 optimization have been made when manipulating perl internals, to
460 sacrifice encapsulation for the benefit of a greater speed.
462 =head1 CANONICAL REPRESENTATION
464 Normally Storable stores elements of hashes in the order they are
465 stored internally by Perl, i.e. pseudo-randomly. If you set
466 C<$Storable::canonical> to some C<TRUE> value, Storable will store
467 hashes with the elements sorted by their key. This allows you to
468 compare data structures by comparing their frozen representations (or
469 even the compressed frozen representations), which can be useful for
470 creating lookup tables for complicated queries.
472 Canonical order does not imply network order, those are two orthogonal
475 =head1 ERROR REPORTING
477 Storable uses the "exception" paradigm, in that it does not try to workaround
478 failures: if something bad happens, an exception is generated from the
479 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
482 When Storable croaks, it tries to report the error via the C<logcroak()>
483 routine from the C<Log::Agent> package, if it is available.
485 Normal errors are reported by having store() or retrieve() return C<undef>.
486 Such errors are usually I/O errors (or truncated stream errors at retrieval).
492 Any class may define hooks that will be called during the serialization
493 and deserialization process on objects that are instances of that class.
494 Those hooks can redefine the way serialization is performed (and therefore,
495 how the symetrical deserialization should be conducted).
497 Since we said earlier:
499 dclone(.) = thaw(freeze(.))
501 everything we say about hooks should also hold for deep cloning. However,
502 hooks get to know whether the operation is a mere serialization, or a cloning.
504 Therefore, when serializing hooks are involved,
506 dclone(.) <> thaw(freeze(.))
508 Well, you could keep them in sync, but there's no guarantee it will always
509 hold on classes somebody else wrote. Besides, there is little to gain in
510 doing so: a serializing hook could only keep one attribute of an object,
511 which is probably not what should happen during a deep cloning of that
514 Here is the hooking interface:
518 =item C<STORABLE_freeze> I<obj>, I<cloning>
520 The serializing hook, called on the object during serialization. It can be
521 inherited, or defined in the class itself, like any other method.
523 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
524 whether we're in a dclone() or a regular serialization via store() or freeze().
526 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
527 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
528 extra references that you wish to let the Storable engine serialize.
530 At deserialization time, you will be given back the same LIST, but all the
531 extra references will be pointing into the deserialized structure.
533 The B<first time> the hook is hit in a serialization flow, you may have it
534 return an empty list. That will signal the Storable engine to further
535 discard that hook for this class and to therefore revert to the default
536 serialization of the underlying Perl data. The hook will again be normally
537 processed in the next serialization.
539 Unless you know better, serializing hook should always say:
541 sub STORABLE_freeze {
542 my ($self, $cloning) = @_;
543 return if $cloning; # Regular default serialization
547 in order to keep reasonable dclone() semantics.
549 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
551 The deserializing hook called on the object during deserialization.
552 But wait. If we're deserializing, there's no object yet... right?
554 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
555 you can view C<STORABLE_thaw> as an alternate creation routine.
557 This means the hook can be inherited like any other method, and that
558 I<obj> is your blessed reference for this particular instance.
560 The other arguments should look familiar if you know C<STORABLE_freeze>:
561 I<cloning> is true when we're part of a deep clone operation, I<serialized>
562 is the serialized string you returned to the engine in C<STORABLE_freeze>,
563 and there may be an optional list of references, in the same order you gave
564 them at serialization time, pointing to the deserialized objects (which
565 have been processed courtesy of the Storable engine).
567 When the Storable engine does not find any C<STORABLE_thaw> hook routine,
568 it tries to load the class by requiring the package dynamically (using
569 the blessed package name), and then re-attempts the lookup. If at that
570 time the hook cannot be located, the engine croaks. Note that this mechanism
571 will fail if you define several classes in the same file, but perlmod(1)
574 It is up to you to use these information to populate I<obj> the way you want.
576 Returned value: none.
582 Predicates are not exportable. They must be called by explicitely prefixing
583 them with the Storable package name.
587 =item C<Storable::last_op_in_netorder>
589 The C<Storable::last_op_in_netorder()> predicate will tell you whether
590 network order was used in the last store or retrieve operation. If you
591 don't know how to use this, just forget about it.
593 =item C<Storable::is_storing>
595 Returns true if within a store operation (via STORABLE_freeze hook).
597 =item C<Storable::is_retrieving>
599 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
605 With hooks comes the ability to recurse back to the Storable engine. Indeed,
606 hooks are regular Perl code, and Storable is convenient when it comes to
607 serialize and deserialize things, so why not use it to handle the
608 serialization string?
610 There are a few things you need to know however:
616 You can create endless loops if the things you serialize via freeze()
617 (for instance) point back to the object we're trying to serialize in the hook.
621 Shared references among objects will not stay shared: if we're serializing
622 the list of object [A, C] where both object A and C refer to the SAME object
623 B, and if there is a serializing hook in A that says freeze(B), then when
624 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
625 a deep clone of B'. The topology was not preserved.
629 That's why C<STORABLE_freeze> lets you provide a list of references
630 to serialize. The engine guarantees that those will be serialized in the
631 same context as the other objects, and therefore that shared objects will
634 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
636 ("something", $self->{B})
638 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
639 would get back the reference to the B' object, deserialized for you.
641 Therefore, recursion should normally be avoided, but is nonetheless supported.
645 There is a new Clone module available on CPAN which implements deep cloning
646 natively, i.e. without freezing to memory and thawing the result. It is
647 aimed to replace Storable's dclone() some day. However, it does not currently
648 support Storable hooks to redefine the way deep cloning is performed.
652 Here are some code samples showing a possible usage of Storable:
654 use Storable qw(store retrieve freeze thaw dclone);
656 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
658 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
660 $colref = retrieve('/tmp/colors');
661 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
662 printf "Blue is still %lf\n", $colref->{'Blue'};
664 $colref2 = dclone(\%color);
666 $str = freeze(\%color);
667 printf "Serialization of %%color is %d bytes long.\n", length($str);
668 $colref3 = thaw($str);
670 which prints (on my machine):
672 Blue is still 0.100000
673 Serialization of %color is 102 bytes long.
677 If you're using references as keys within your hash tables, you're bound
678 to disapointment when retrieving your data. Indeed, Perl stringifies
679 references used as hash table keys. If you later wish to access the
680 items via another reference stringification (i.e. using the same
681 reference that was used for the key originally to record the value into
682 the hash table), it will work because both references stringify to the
685 It won't work across a C<store> and C<retrieve> operations however, because
686 the addresses in the retrieved objects, which are part of the stringified
687 references, will probably differ from the original addresses. The
688 topology of your structure is preserved, but not hidden semantics
691 On platforms where it matters, be sure to call C<binmode()> on the
692 descriptors that you pass to Storable functions.
694 Storing data canonically that contains large hashes can be
695 significantly slower than storing the same data normally, as
696 temprorary arrays to hold the keys for each hash have to be allocated,
697 populated, sorted and freed. Some tests have shown a halving of the
698 speed of storing -- the exact penalty will depend on the complexity of
699 your data. There is no slowdown on retrieval.
703 You can't store GLOB, CODE, FORMLINE, etc... If you can define
704 semantics for those operations, feel free to enhance Storable so that
705 it can deal with them.
707 The store functions will C<croak> if they run into such references
708 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
709 case, the fatal message is turned in a warning and some
710 meaningless string is stored instead.
712 Setting C<$Storable::canonical> may not yield frozen strings that
713 compare equal due to possible stringification of numbers. When the
714 string version of a scalar exists, it is the form stored, therefore
715 if you happen to use your numbers as strings between two freezing
716 operations on the same data structures, you will get different
719 When storing doubles in network order, their value is stored as text.
720 However, you should also not expect non-numeric floating-point values
721 such as infinity and "not a number" to pass successfully through a
722 nstore()/retrieve() pair.
724 As Storable neither knows nor cares about character sets (although it
725 does know that characters may be more than eight bits wide), any difference
726 in the interpretation of character codes between a host and a target
727 system is your problem. In particular, if host and target use different
728 code points to represent the characters used in the text representation
729 of floating-point numbers, you will not be able be able to exchange
730 floating-point data, even with nstore().
734 Thank you to (in chronological order):
736 Jarkko Hietaniemi <jhi@iki.fi>
737 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
738 Benjamin A. Holzman <bah@ecnvantage.com>
739 Andrew Ford <A.Ford@ford-mason.co.uk>
740 Gisle Aas <gisle@aas.no>
741 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
742 Murray Nesbitt <murray@activestate.com>
743 Marc Lehmann <pcg@opengroup.org>
744 Justin Banks <justinb@wamnet.com>
745 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
746 Salvador Ortiz Garcia <sog@msg.com.mx>
747 Dominic Dunlop <domo@computer.org>
748 Erik Haugan <erik@solbors.no>
750 for their bug reports, suggestions and contributions.
752 Benjamin Holzman contributed the tied variable support, Andrew Ford
753 contributed the canonical order for hashes, and Gisle Aas fixed
754 a few misunderstandings of mine regarding the Perl internals,
755 and optimized the emission of "tags" in the output streams by
756 simply counting the objects instead of tagging them (leading to
757 a binary incompatibility for the Storable image starting at version
758 0.6--older images are of course still properly understood).
759 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
760 and reference to tied items support.
764 There is a Japanese translation of this man page available at
765 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
766 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
770 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>