1 ;# $Id: Storable.pm,v 1.0.1.5 2000/10/26 17:10:18 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 1.0.1.6 2000/11/05 17:20:25 ram
10 ;# patch6: increased version number
12 ;# Revision 1.0.1.5 2000/10/26 17:10:18 ram
13 ;# patch5: documented that store() and retrieve() can return undef
14 ;# patch5: added paragraph explaining the auto require for thaw hooks
16 ;# Revision 1.0.1.4 2000/10/23 18:02:57 ram
17 ;# patch4: protected calls to flock() for dos platform
18 ;# patch4: added logcarp emulation if they don't have Log::Agent
20 ;# $Log: Storable.pm,v $
21 ;# Revision 1.0 2000/09/01 19:40:41 ram
22 ;# Baseline for first official release.
27 package Storable; @ISA = qw(Exporter DynaLoader);
29 @EXPORT = qw(store retrieve);
31 nstore store_fd nstore_fd fd_retrieve
35 lock_store lock_nstore lock_retrieve
39 use vars qw($forgive_me $VERSION);
42 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
45 # Use of Log::Agent is optional
48 eval "use Log::Agent";
50 unless (defined @Log::Agent::EXPORT) {
64 # They might miss :flock in Fcntl
69 if (exists $Fcntl::EXPORT_TAGS{'flock'}) {
70 Fcntl->import(':flock');
82 sub retrieve_fd { &fd_retrieve } # Backward compatibility
91 # Store target object hierarchy, identified by a reference to its root.
92 # The stored object tree may later be retrieved to memory via retrieve.
93 # Returns undef if an I/O error occurred, in which case the file is
97 return _store(\&pstore, @_, 0);
103 # Same as store, but in network order.
106 return _store(\&net_pstore, @_, 0);
112 # Same as store, but flock the file first (advisory locking).
115 return _store(\&pstore, @_, 1);
121 # Same as nstore, but flock the file first (advisory locking).
124 return _store(\&net_pstore, @_, 1);
127 # Internal store to file routine
131 my ($file, $use_locking) = @_;
132 logcroak "not a reference" unless ref($self);
133 logcroak "too many arguments" unless @_ == 2; # No @foo in arglist
135 open(FILE, ">$file") || logcroak "can't create $file: $!";
136 binmode FILE; # Archaic systems...
138 require Config; import Config;
139 if (!$Config{'d_flock'} &&
140 !$Config{'d_fcntl_can_lock'} &&
141 !$Config{'d_lockf'}) {
142 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
145 flock(FILE, LOCK_EX) ||
146 logcroak "can't get exclusive lock on $file: $!";
148 # Unlocking will happen when FILE is closed
150 my $da = $@; # Don't mess if called from exception handler
152 # Call C routine nstore or pstore, depending on network order
153 eval { $ret = &$xsptr(*FILE, $self) };
154 close(FILE) or $ret = undef;
155 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
156 logcroak $@ if $@ =~ s/\.?\n$/,/;
158 return $ret ? $ret : undef;
164 # Same as store, but perform on an already opened file descriptor instead.
165 # Returns undef if an I/O error occurred.
168 return _store_fd(\&pstore, @_);
174 # Same as store_fd, but in network order.
177 my ($self, $file) = @_;
178 return _store_fd(\&net_pstore, @_);
181 # Internal store routine on opened file descriptor
186 logcroak "not a reference" unless ref($self);
187 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
188 my $fd = fileno($file);
189 logcroak "not a valid file descriptor" unless defined $fd;
190 my $da = $@; # Don't mess if called from exception handler
192 # Call C routine nstore or pstore, depending on network order
193 eval { $ret = &$xsptr($file, $self) };
194 logcroak $@ if $@ =~ s/\.?\n$/,/;
196 return $ret ? $ret : undef;
202 # Store oject and its hierarchy in memory and return a scalar
203 # containing the result.
206 _freeze(\&mstore, @_);
212 # Same as freeze but in network order.
215 _freeze(\&net_mstore, @_);
218 # Internal freeze routine
222 logcroak "not a reference" unless ref($self);
223 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
224 my $da = $@; # Don't mess if called from exception handler
226 # Call C routine mstore or net_mstore, depending on network order
227 eval { $ret = &$xsptr($self) };
228 logcroak $@ if $@ =~ s/\.?\n$/,/;
230 return $ret ? $ret : undef;
236 # Retrieve object hierarchy from disk, returning a reference to the root
237 # object of that tree.
246 # Same as retrieve, but with advisory locking.
252 # Internal retrieve routine
254 my ($file, $use_locking) = @_;
256 open(FILE, $file) || logcroak "can't open $file: $!";
257 binmode FILE; # Archaic systems...
259 my $da = $@; # Could be from exception handler
261 require Config; import Config;
262 if (!$Config{'d_flock'} &&
263 !$Config{'d_fcntl_can_lock'} &&
264 !$Config{'d_lockf'}) {
265 logcarp "Storable::lock_retrieve: fcntl/flock emulation broken on $^O";
268 flock(FILE, LOCK_SH) ||
269 logcroak "can't get shared lock on $file: $!";
270 # Unlocking will happen when FILE is closed
272 eval { $self = pretrieve(*FILE) }; # Call C routine
274 logcroak $@ if $@ =~ s/\.?\n$/,/;
282 # Same as retrieve, but perform from an already opened file descriptor instead.
286 my $fd = fileno($file);
287 logcroak "not a valid file descriptor" unless defined $fd;
289 my $da = $@; # Could be from exception handler
290 eval { $self = pretrieve($file) }; # Call C routine
291 logcroak $@ if $@ =~ s/\.?\n$/,/;
299 # Recreate objects in memory from an existing frozen image created
300 # by freeze. If the frozen image passed is undef, return undef.
304 return undef unless defined $frozen;
306 my $da = $@; # Could be from exception handler
307 eval { $self = mretrieve($frozen) }; # Call C routine
308 logcroak $@ if $@ =~ s/\.?\n$/,/;
315 Storable - persistency for perl data structures
320 store \%table, 'file';
321 $hashref = retrieve('file');
323 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
326 nstore \%table, 'file';
327 $hashref = retrieve('file'); # There is NO nretrieve()
329 # Storing to and retrieving from an already opened file
330 store_fd \@array, \*STDOUT;
331 nstore_fd \%table, \*STDOUT;
332 $aryref = fd_retrieve(\*SOCKET);
333 $hashref = fd_retrieve(\*SOCKET);
335 # Serializing to memory
336 $serialized = freeze \%table;
337 %table_clone = %{ thaw($serialized) };
339 # Deep (recursive) cloning
340 $cloneref = dclone($ref);
343 use Storable qw(lock_store lock_nstore lock_retrieve)
344 lock_store \%table, 'file';
345 lock_nstore \%table, 'file';
346 $hashref = lock_retrieve('file');
350 The Storable package brings persistency to your perl data structures
351 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
352 convenientely stored to disk and retrieved at a later time.
354 It can be used in the regular procedural way by calling C<store> with
355 a reference to the object to be stored, along with the file name where
356 the image should be written.
357 The routine returns C<undef> for I/O problems or other internal error,
358 a true value otherwise. Serious errors are propagated as a C<die> exception.
360 To retrieve data stored to disk, use C<retrieve> with a file name,
361 and the objects stored into that file are recreated into memory for you,
362 a I<reference> to the root object being returned. In case an I/O error
363 occurs while reading, C<undef> is returned instead. Other serious
364 errors are propagated via C<die>.
366 Since storage is performed recursively, you might want to stuff references
367 to objects that share a lot of common data into a single array or hash
368 table, and then store that object. That way, when you retrieve back the
369 whole thing, the objects will continue to share what they originally shared.
371 At the cost of a slight header overhead, you may store to an already
372 opened file descriptor using the C<store_fd> routine, and retrieve
373 from a file via C<fd_retrieve>. Those names aren't imported by default,
374 so you will have to do that explicitely if you need those routines.
375 The file descriptor you supply must be already opened, for read
376 if you're going to retrieve and for write if you wish to store.
378 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
379 $hashref = fd_retrieve(*STDIN);
381 You can also store data in network order to allow easy sharing across
382 multiple platforms, or when storing on a socket known to be remotely
383 connected. The routines to call have an initial C<n> prefix for I<network>,
384 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
385 correctly restored so you don't have to know whether you're restoring
386 from native or network ordered data. Double values are stored stringified
387 to ensure portability as well, at the slight risk of loosing some precision
388 in the last decimals.
390 When using C<fd_retrieve>, objects are retrieved in sequence, one
391 object (i.e. one recursive tree) per associated C<store_fd>.
393 If you're more from the object-oriented camp, you can inherit from
394 Storable and directly store your objects by invoking C<store> as
395 a method. The fact that the root of the to-be-stored tree is a
396 blessed reference (i.e. an object) is special-cased so that the
397 retrieve does not provide a reference to that object but rather the
398 blessed object reference itself. (Otherwise, you'd get a reference
399 to that blessed object).
403 The Storable engine can also store data into a Perl scalar instead, to
404 later retrieve them. This is mainly used to freeze a complex structure in
405 some safe compact memory place (where it can possibly be sent to another
406 process via some IPC, since freezing the structure also serializes it in
407 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
408 out and recreate the original complex structure in memory.
410 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
411 If you wish to send out the frozen scalar to another machine, use
412 C<nfreeze> instead to get a portable image.
414 Note that freezing an object structure and immediately thawing it
415 actually achieves a deep cloning of that structure:
417 dclone(.) = thaw(freeze(.))
419 Storable provides you with a C<dclone> interface which does not create
420 that intermediary scalar but instead freezes the structure in some
421 internal memory space and then immediatly thaws it out.
423 =head1 ADVISORY LOCKING
425 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
426 and C<nstore>, only they get an exclusive lock on the file before
427 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
428 gets a shared lock on the file before reading.
430 Like with any advisory locking scheme, the protection only works if
431 you systematically use C<lock_store> and C<lock_retrieve>. If one
432 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
433 you will get no protection at all.
435 The internal advisory locking is implemented using Perl's flock() routine.
436 If your system does not support any form of flock(), or if you share
437 your files across NFS, you might wish to use other forms of locking by
438 using modules like LockFile::Simple which lock a file using a filesystem
439 entry, instead of locking the file descriptor.
443 The heart of Storable is written in C for decent speed. Extra low-level
444 optimization have been made when manipulating perl internals, to
445 sacrifice encapsulation for the benefit of a greater speed.
447 =head1 CANONICAL REPRESENTATION
449 Normally Storable stores elements of hashes in the order they are
450 stored internally by Perl, i.e. pseudo-randomly. If you set
451 C<$Storable::canonical> to some C<TRUE> value, Storable will store
452 hashes with the elements sorted by their key. This allows you to
453 compare data structures by comparing their frozen representations (or
454 even the compressed frozen representations), which can be useful for
455 creating lookup tables for complicated queries.
457 Canonical order does not imply network order, those are two orthogonal
460 =head1 ERROR REPORTING
462 Storable uses the "exception" paradigm, in that it does not try to workaround
463 failures: if something bad happens, an exception is generated from the
464 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
467 When Storable croaks, it tries to report the error via the C<logcroak()>
468 routine from the C<Log::Agent> package, if it is available.
470 Normal errors are reported by having store() or retrieve() return C<undef>.
471 Such errors are usually I/O errors (or truncated stream errors at retrieval).
477 Any class may define hooks that will be called during the serialization
478 and deserialization process on objects that are instances of that class.
479 Those hooks can redefine the way serialization is performed (and therefore,
480 how the symetrical deserialization should be conducted).
482 Since we said earlier:
484 dclone(.) = thaw(freeze(.))
486 everything we say about hooks should also hold for deep cloning. However,
487 hooks get to know whether the operation is a mere serialization, or a cloning.
489 Therefore, when serializing hooks are involved,
491 dclone(.) <> thaw(freeze(.))
493 Well, you could keep them in sync, but there's no guarantee it will always
494 hold on classes somebody else wrote. Besides, there is little to gain in
495 doing so: a serializing hook could only keep one attribute of an object,
496 which is probably not what should happen during a deep cloning of that
499 Here is the hooking interface:
503 =item C<STORABLE_freeze> I<obj>, I<cloning>
505 The serializing hook, called on the object during serialization. It can be
506 inherited, or defined in the class itself, like any other method.
508 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
509 whether we're in a dclone() or a regular serialization via store() or freeze().
511 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
512 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
513 extra references that you wish to let the Storable engine serialize.
515 At deserialization time, you will be given back the same LIST, but all the
516 extra references will be pointing into the deserialized structure.
518 The B<first time> the hook is hit in a serialization flow, you may have it
519 return an empty list. That will signal the Storable engine to further
520 discard that hook for this class and to therefore revert to the default
521 serialization of the underlying Perl data. The hook will again be normally
522 processed in the next serialization.
524 Unless you know better, serializing hook should always say:
526 sub STORABLE_freeze {
527 my ($self, $cloning) = @_;
528 return if $cloning; # Regular default serialization
532 in order to keep reasonable dclone() semantics.
534 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
536 The deserializing hook called on the object during deserialization.
537 But wait. If we're deserializing, there's no object yet... right?
539 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
540 you can view C<STORABLE_thaw> as an alternate creation routine.
542 This means the hook can be inherited like any other method, and that
543 I<obj> is your blessed reference for this particular instance.
545 The other arguments should look familiar if you know C<STORABLE_freeze>:
546 I<cloning> is true when we're part of a deep clone operation, I<serialized>
547 is the serialized string you returned to the engine in C<STORABLE_freeze>,
548 and there may be an optional list of references, in the same order you gave
549 them at serialization time, pointing to the deserialized objects (which
550 have been processed courtesy of the Storable engine).
552 When the Storable engine does not find any C<STORABLE_thaw> hook routine,
553 it tries to load the class by requiring the package dynamically (using
554 the blessed package name), and then re-attempts the lookup. If at that
555 time the hook cannot be located, the engine croaks. Note that this mechanism
556 will fail if you define several classes in the same file, but perlmod(1)
559 It is up to you to use these information to populate I<obj> the way you want.
561 Returned value: none.
567 Predicates are not exportable. They must be called by explicitely prefixing
568 them with the Storable package name.
572 =item C<Storable::last_op_in_netorder>
574 The C<Storable::last_op_in_netorder()> predicate will tell you whether
575 network order was used in the last store or retrieve operation. If you
576 don't know how to use this, just forget about it.
578 =item C<Storable::is_storing>
580 Returns true if within a store operation (via STORABLE_freeze hook).
582 =item C<Storable::is_retrieving>
584 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
590 With hooks comes the ability to recurse back to the Storable engine. Indeed,
591 hooks are regular Perl code, and Storable is convenient when it comes to
592 serialize and deserialize things, so why not use it to handle the
593 serialization string?
595 There are a few things you need to know however:
601 You can create endless loops if the things you serialize via freeze()
602 (for instance) point back to the object we're trying to serialize in the hook.
606 Shared references among objects will not stay shared: if we're serializing
607 the list of object [A, C] where both object A and C refer to the SAME object
608 B, and if there is a serializing hook in A that says freeze(B), then when
609 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
610 a deep clone of B'. The topology was not preserved.
614 That's why C<STORABLE_freeze> lets you provide a list of references
615 to serialize. The engine guarantees that those will be serialized in the
616 same context as the other objects, and therefore that shared objects will
619 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
621 ("something", $self->{B})
623 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
624 would get back the reference to the B' object, deserialized for you.
626 Therefore, recursion should normally be avoided, but is nonetheless supported.
630 There is a new Clone module available on CPAN which implements deep cloning
631 natively, i.e. without freezing to memory and thawing the result. It is
632 aimed to replace Storable's dclone() some day. However, it does not currently
633 support Storable hooks to redefine the way deep cloning is performed.
637 Here are some code samples showing a possible usage of Storable:
639 use Storable qw(store retrieve freeze thaw dclone);
641 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
643 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
645 $colref = retrieve('/tmp/colors');
646 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
647 printf "Blue is still %lf\n", $colref->{'Blue'};
649 $colref2 = dclone(\%color);
651 $str = freeze(\%color);
652 printf "Serialization of %%color is %d bytes long.\n", length($str);
653 $colref3 = thaw($str);
655 which prints (on my machine):
657 Blue is still 0.100000
658 Serialization of %color is 102 bytes long.
662 If you're using references as keys within your hash tables, you're bound
663 to disapointment when retrieving your data. Indeed, Perl stringifies
664 references used as hash table keys. If you later wish to access the
665 items via another reference stringification (i.e. using the same
666 reference that was used for the key originally to record the value into
667 the hash table), it will work because both references stringify to the
670 It won't work across a C<store> and C<retrieve> operations however, because
671 the addresses in the retrieved objects, which are part of the stringified
672 references, will probably differ from the original addresses. The
673 topology of your structure is preserved, but not hidden semantics
676 On platforms where it matters, be sure to call C<binmode()> on the
677 descriptors that you pass to Storable functions.
679 Storing data canonically that contains large hashes can be
680 significantly slower than storing the same data normally, as
681 temprorary arrays to hold the keys for each hash have to be allocated,
682 populated, sorted and freed. Some tests have shown a halving of the
683 speed of storing -- the exact penalty will depend on the complexity of
684 your data. There is no slowdown on retrieval.
688 You can't store GLOB, CODE, FORMLINE, etc... If you can define
689 semantics for those operations, feel free to enhance Storable so that
690 it can deal with them.
692 The store functions will C<croak> if they run into such references
693 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
694 case, the fatal message is turned in a warning and some
695 meaningless string is stored instead.
697 Setting C<$Storable::canonical> may not yield frozen strings that
698 compare equal due to possible stringification of numbers. When the
699 string version of a scalar exists, it is the form stored, therefore
700 if you happen to use your numbers as strings between two freezing
701 operations on the same data structures, you will get different
704 When storing doubles in network order, their value is stored as text.
705 However, you should also not expect non-numeric floating-point values
706 such as infinity and "not a number" to pass successfully through a
707 nstore()/retrieve() pair.
709 As Storable neither knows nor cares about character sets (although it
710 does know that characters may be more than eight bits wide), any difference
711 in the interpretation of character codes between a host and a target
712 system is your problem. In particular, if host and target use different
713 code points to represent the characters used in the text representation
714 of floating-point numbers, you will not be able be able to exchange
715 floating-point data, even with nstore().
719 Thank you to (in chronological order):
721 Jarkko Hietaniemi <jhi@iki.fi>
722 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
723 Benjamin A. Holzman <bah@ecnvantage.com>
724 Andrew Ford <A.Ford@ford-mason.co.uk>
725 Gisle Aas <gisle@aas.no>
726 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
727 Murray Nesbitt <murray@activestate.com>
728 Marc Lehmann <pcg@opengroup.org>
729 Justin Banks <justinb@wamnet.com>
730 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
731 Salvador Ortiz Garcia <sog@msg.com.mx>
732 Dominic Dunlop <domo@computer.org>
733 Erik Haugan <erik@solbors.no>
735 for their bug reports, suggestions and contributions.
737 Benjamin Holzman contributed the tied variable support, Andrew Ford
738 contributed the canonical order for hashes, and Gisle Aas fixed
739 a few misunderstandings of mine regarding the Perl internals,
740 and optimized the emission of "tags" in the output streams by
741 simply counting the objects instead of tagging them (leading to
742 a binary incompatibility for the Storable image starting at version
743 0.6--older images are of course still properly understood).
744 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
745 and reference to tied items support.
749 There is a Japanese translation of this man page available at
750 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
751 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
755 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>