1 ;# $Id: Storable.pm,v 1.0.1.7 2001/01/03 09:39:02 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 1.0.1.7 2001/01/03 09:39:02 ram
10 ;# patch7: added CAN_FLOCK to determine whether we can flock() or not
12 ;# Revision 1.0.1.6 2000/11/05 17:20:25 ram
13 ;# patch6: increased version number
15 ;# Revision 1.0.1.5 2000/10/26 17:10:18 ram
16 ;# patch5: documented that store() and retrieve() can return undef
17 ;# patch5: added paragraph explaining the auto require for thaw hooks
19 ;# Revision 1.0.1.4 2000/10/23 18:02:57 ram
20 ;# patch4: protected calls to flock() for dos platform
21 ;# patch4: added logcarp emulation if they don't have Log::Agent
23 ;# $Log: Storable.pm,v $
24 ;# Revision 1.0 2000/09/01 19:40:41 ram
25 ;# Baseline for first official release.
30 package Storable; @ISA = qw(Exporter DynaLoader);
32 @EXPORT = qw(store retrieve);
34 nstore store_fd nstore_fd fd_retrieve
38 lock_store lock_nstore lock_retrieve
42 use vars qw($forgive_me $VERSION);
45 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
48 # Use of Log::Agent is optional
51 eval "use Log::Agent";
53 unless (defined @Log::Agent::EXPORT) {
67 # They might miss :flock in Fcntl
72 if (exists $Fcntl::EXPORT_TAGS{'flock'}) {
73 Fcntl->import(':flock');
85 sub retrieve_fd { &fd_retrieve } # Backward compatibility
88 # Determine whether locking is possible, but only when needed.
94 return $CAN_FLOCK if defined $CAN_FLOCK;
95 require Config; import Config;
98 $Config{'d_fcntl_can_lock'} ||
109 # Store target object hierarchy, identified by a reference to its root.
110 # The stored object tree may later be retrieved to memory via retrieve.
111 # Returns undef if an I/O error occurred, in which case the file is
115 return _store(\&pstore, @_, 0);
121 # Same as store, but in network order.
124 return _store(\&net_pstore, @_, 0);
130 # Same as store, but flock the file first (advisory locking).
133 return _store(\&pstore, @_, 1);
139 # Same as nstore, but flock the file first (advisory locking).
142 return _store(\&net_pstore, @_, 1);
145 # Internal store to file routine
149 my ($file, $use_locking) = @_;
150 logcroak "not a reference" unless ref($self);
151 logcroak "too many arguments" unless @_ == 2; # No @foo in arglist
153 open(FILE, ">$file") || logcroak "can't create $file: $!";
154 binmode FILE; # Archaic systems...
156 unless (&CAN_FLOCK) {
157 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
160 flock(FILE, LOCK_EX) ||
161 logcroak "can't get exclusive lock on $file: $!";
163 # Unlocking will happen when FILE is closed
165 my $da = $@; # Don't mess if called from exception handler
167 # Call C routine nstore or pstore, depending on network order
168 eval { $ret = &$xsptr(*FILE, $self) };
169 close(FILE) or $ret = undef;
170 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
171 logcroak $@ if $@ =~ s/\.?\n$/,/;
173 return $ret ? $ret : undef;
179 # Same as store, but perform on an already opened file descriptor instead.
180 # Returns undef if an I/O error occurred.
183 return _store_fd(\&pstore, @_);
189 # Same as store_fd, but in network order.
192 my ($self, $file) = @_;
193 return _store_fd(\&net_pstore, @_);
196 # Internal store routine on opened file descriptor
201 logcroak "not a reference" unless ref($self);
202 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
203 my $fd = fileno($file);
204 logcroak "not a valid file descriptor" unless defined $fd;
205 my $da = $@; # Don't mess if called from exception handler
207 # Call C routine nstore or pstore, depending on network order
208 eval { $ret = &$xsptr($file, $self) };
209 logcroak $@ if $@ =~ s/\.?\n$/,/;
211 return $ret ? $ret : undef;
217 # Store oject and its hierarchy in memory and return a scalar
218 # containing the result.
221 _freeze(\&mstore, @_);
227 # Same as freeze but in network order.
230 _freeze(\&net_mstore, @_);
233 # Internal freeze routine
237 logcroak "not a reference" unless ref($self);
238 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
239 my $da = $@; # Don't mess if called from exception handler
241 # Call C routine mstore or net_mstore, depending on network order
242 eval { $ret = &$xsptr($self) };
243 logcroak $@ if $@ =~ s/\.?\n$/,/;
245 return $ret ? $ret : undef;
251 # Retrieve object hierarchy from disk, returning a reference to the root
252 # object of that tree.
261 # Same as retrieve, but with advisory locking.
267 # Internal retrieve routine
269 my ($file, $use_locking) = @_;
271 open(FILE, $file) || logcroak "can't open $file: $!";
272 binmode FILE; # Archaic systems...
274 my $da = $@; # Could be from exception handler
276 unless (&CAN_FLOCK) {
277 logcarp "Storable::lock_retrieve: fcntl/flock emulation broken on $^O";
280 flock(FILE, LOCK_SH) ||
281 logcroak "can't get shared lock on $file: $!";
282 # Unlocking will happen when FILE is closed
284 eval { $self = pretrieve(*FILE) }; # Call C routine
286 logcroak $@ if $@ =~ s/\.?\n$/,/;
294 # Same as retrieve, but perform from an already opened file descriptor instead.
298 my $fd = fileno($file);
299 logcroak "not a valid file descriptor" unless defined $fd;
301 my $da = $@; # Could be from exception handler
302 eval { $self = pretrieve($file) }; # Call C routine
303 logcroak $@ if $@ =~ s/\.?\n$/,/;
311 # Recreate objects in memory from an existing frozen image created
312 # by freeze. If the frozen image passed is undef, return undef.
316 return undef unless defined $frozen;
318 my $da = $@; # Could be from exception handler
319 eval { $self = mretrieve($frozen) }; # Call C routine
320 logcroak $@ if $@ =~ s/\.?\n$/,/;
327 Storable - persistency for perl data structures
332 store \%table, 'file';
333 $hashref = retrieve('file');
335 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
338 nstore \%table, 'file';
339 $hashref = retrieve('file'); # There is NO nretrieve()
341 # Storing to and retrieving from an already opened file
342 store_fd \@array, \*STDOUT;
343 nstore_fd \%table, \*STDOUT;
344 $aryref = fd_retrieve(\*SOCKET);
345 $hashref = fd_retrieve(\*SOCKET);
347 # Serializing to memory
348 $serialized = freeze \%table;
349 %table_clone = %{ thaw($serialized) };
351 # Deep (recursive) cloning
352 $cloneref = dclone($ref);
355 use Storable qw(lock_store lock_nstore lock_retrieve)
356 lock_store \%table, 'file';
357 lock_nstore \%table, 'file';
358 $hashref = lock_retrieve('file');
362 The Storable package brings persistency to your perl data structures
363 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
364 convenientely stored to disk and retrieved at a later time.
366 It can be used in the regular procedural way by calling C<store> with
367 a reference to the object to be stored, along with the file name where
368 the image should be written.
369 The routine returns C<undef> for I/O problems or other internal error,
370 a true value otherwise. Serious errors are propagated as a C<die> exception.
372 To retrieve data stored to disk, use C<retrieve> with a file name,
373 and the objects stored into that file are recreated into memory for you,
374 a I<reference> to the root object being returned. In case an I/O error
375 occurs while reading, C<undef> is returned instead. Other serious
376 errors are propagated via C<die>.
378 Since storage is performed recursively, you might want to stuff references
379 to objects that share a lot of common data into a single array or hash
380 table, and then store that object. That way, when you retrieve back the
381 whole thing, the objects will continue to share what they originally shared.
383 At the cost of a slight header overhead, you may store to an already
384 opened file descriptor using the C<store_fd> routine, and retrieve
385 from a file via C<fd_retrieve>. Those names aren't imported by default,
386 so you will have to do that explicitely if you need those routines.
387 The file descriptor you supply must be already opened, for read
388 if you're going to retrieve and for write if you wish to store.
390 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
391 $hashref = fd_retrieve(*STDIN);
393 You can also store data in network order to allow easy sharing across
394 multiple platforms, or when storing on a socket known to be remotely
395 connected. The routines to call have an initial C<n> prefix for I<network>,
396 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
397 correctly restored so you don't have to know whether you're restoring
398 from native or network ordered data. Double values are stored stringified
399 to ensure portability as well, at the slight risk of loosing some precision
400 in the last decimals.
402 When using C<fd_retrieve>, objects are retrieved in sequence, one
403 object (i.e. one recursive tree) per associated C<store_fd>.
405 If you're more from the object-oriented camp, you can inherit from
406 Storable and directly store your objects by invoking C<store> as
407 a method. The fact that the root of the to-be-stored tree is a
408 blessed reference (i.e. an object) is special-cased so that the
409 retrieve does not provide a reference to that object but rather the
410 blessed object reference itself. (Otherwise, you'd get a reference
411 to that blessed object).
415 The Storable engine can also store data into a Perl scalar instead, to
416 later retrieve them. This is mainly used to freeze a complex structure in
417 some safe compact memory place (where it can possibly be sent to another
418 process via some IPC, since freezing the structure also serializes it in
419 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
420 out and recreate the original complex structure in memory.
422 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
423 If you wish to send out the frozen scalar to another machine, use
424 C<nfreeze> instead to get a portable image.
426 Note that freezing an object structure and immediately thawing it
427 actually achieves a deep cloning of that structure:
429 dclone(.) = thaw(freeze(.))
431 Storable provides you with a C<dclone> interface which does not create
432 that intermediary scalar but instead freezes the structure in some
433 internal memory space and then immediatly thaws it out.
435 =head1 ADVISORY LOCKING
437 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
438 and C<nstore>, only they get an exclusive lock on the file before
439 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
440 gets a shared lock on the file before reading.
442 Like with any advisory locking scheme, the protection only works if
443 you systematically use C<lock_store> and C<lock_retrieve>. If one
444 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
445 you will get no protection at all.
447 The internal advisory locking is implemented using Perl's flock() routine.
448 If your system does not support any form of flock(), or if you share
449 your files across NFS, you might wish to use other forms of locking by
450 using modules like LockFile::Simple which lock a file using a filesystem
451 entry, instead of locking the file descriptor.
455 The heart of Storable is written in C for decent speed. Extra low-level
456 optimization have been made when manipulating perl internals, to
457 sacrifice encapsulation for the benefit of a greater speed.
459 =head1 CANONICAL REPRESENTATION
461 Normally Storable stores elements of hashes in the order they are
462 stored internally by Perl, i.e. pseudo-randomly. If you set
463 C<$Storable::canonical> to some C<TRUE> value, Storable will store
464 hashes with the elements sorted by their key. This allows you to
465 compare data structures by comparing their frozen representations (or
466 even the compressed frozen representations), which can be useful for
467 creating lookup tables for complicated queries.
469 Canonical order does not imply network order, those are two orthogonal
472 =head1 ERROR REPORTING
474 Storable uses the "exception" paradigm, in that it does not try to workaround
475 failures: if something bad happens, an exception is generated from the
476 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
479 When Storable croaks, it tries to report the error via the C<logcroak()>
480 routine from the C<Log::Agent> package, if it is available.
482 Normal errors are reported by having store() or retrieve() return C<undef>.
483 Such errors are usually I/O errors (or truncated stream errors at retrieval).
489 Any class may define hooks that will be called during the serialization
490 and deserialization process on objects that are instances of that class.
491 Those hooks can redefine the way serialization is performed (and therefore,
492 how the symetrical deserialization should be conducted).
494 Since we said earlier:
496 dclone(.) = thaw(freeze(.))
498 everything we say about hooks should also hold for deep cloning. However,
499 hooks get to know whether the operation is a mere serialization, or a cloning.
501 Therefore, when serializing hooks are involved,
503 dclone(.) <> thaw(freeze(.))
505 Well, you could keep them in sync, but there's no guarantee it will always
506 hold on classes somebody else wrote. Besides, there is little to gain in
507 doing so: a serializing hook could only keep one attribute of an object,
508 which is probably not what should happen during a deep cloning of that
511 Here is the hooking interface:
515 =item C<STORABLE_freeze> I<obj>, I<cloning>
517 The serializing hook, called on the object during serialization. It can be
518 inherited, or defined in the class itself, like any other method.
520 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
521 whether we're in a dclone() or a regular serialization via store() or freeze().
523 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
524 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
525 extra references that you wish to let the Storable engine serialize.
527 At deserialization time, you will be given back the same LIST, but all the
528 extra references will be pointing into the deserialized structure.
530 The B<first time> the hook is hit in a serialization flow, you may have it
531 return an empty list. That will signal the Storable engine to further
532 discard that hook for this class and to therefore revert to the default
533 serialization of the underlying Perl data. The hook will again be normally
534 processed in the next serialization.
536 Unless you know better, serializing hook should always say:
538 sub STORABLE_freeze {
539 my ($self, $cloning) = @_;
540 return if $cloning; # Regular default serialization
544 in order to keep reasonable dclone() semantics.
546 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
548 The deserializing hook called on the object during deserialization.
549 But wait. If we're deserializing, there's no object yet... right?
551 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
552 you can view C<STORABLE_thaw> as an alternate creation routine.
554 This means the hook can be inherited like any other method, and that
555 I<obj> is your blessed reference for this particular instance.
557 The other arguments should look familiar if you know C<STORABLE_freeze>:
558 I<cloning> is true when we're part of a deep clone operation, I<serialized>
559 is the serialized string you returned to the engine in C<STORABLE_freeze>,
560 and there may be an optional list of references, in the same order you gave
561 them at serialization time, pointing to the deserialized objects (which
562 have been processed courtesy of the Storable engine).
564 When the Storable engine does not find any C<STORABLE_thaw> hook routine,
565 it tries to load the class by requiring the package dynamically (using
566 the blessed package name), and then re-attempts the lookup. If at that
567 time the hook cannot be located, the engine croaks. Note that this mechanism
568 will fail if you define several classes in the same file, but perlmod(1)
571 It is up to you to use these information to populate I<obj> the way you want.
573 Returned value: none.
579 Predicates are not exportable. They must be called by explicitely prefixing
580 them with the Storable package name.
584 =item C<Storable::last_op_in_netorder>
586 The C<Storable::last_op_in_netorder()> predicate will tell you whether
587 network order was used in the last store or retrieve operation. If you
588 don't know how to use this, just forget about it.
590 =item C<Storable::is_storing>
592 Returns true if within a store operation (via STORABLE_freeze hook).
594 =item C<Storable::is_retrieving>
596 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
602 With hooks comes the ability to recurse back to the Storable engine. Indeed,
603 hooks are regular Perl code, and Storable is convenient when it comes to
604 serialize and deserialize things, so why not use it to handle the
605 serialization string?
607 There are a few things you need to know however:
613 You can create endless loops if the things you serialize via freeze()
614 (for instance) point back to the object we're trying to serialize in the hook.
618 Shared references among objects will not stay shared: if we're serializing
619 the list of object [A, C] where both object A and C refer to the SAME object
620 B, and if there is a serializing hook in A that says freeze(B), then when
621 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
622 a deep clone of B'. The topology was not preserved.
626 That's why C<STORABLE_freeze> lets you provide a list of references
627 to serialize. The engine guarantees that those will be serialized in the
628 same context as the other objects, and therefore that shared objects will
631 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
633 ("something", $self->{B})
635 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
636 would get back the reference to the B' object, deserialized for you.
638 Therefore, recursion should normally be avoided, but is nonetheless supported.
642 There is a new Clone module available on CPAN which implements deep cloning
643 natively, i.e. without freezing to memory and thawing the result. It is
644 aimed to replace Storable's dclone() some day. However, it does not currently
645 support Storable hooks to redefine the way deep cloning is performed.
649 Here are some code samples showing a possible usage of Storable:
651 use Storable qw(store retrieve freeze thaw dclone);
653 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
655 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
657 $colref = retrieve('/tmp/colors');
658 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
659 printf "Blue is still %lf\n", $colref->{'Blue'};
661 $colref2 = dclone(\%color);
663 $str = freeze(\%color);
664 printf "Serialization of %%color is %d bytes long.\n", length($str);
665 $colref3 = thaw($str);
667 which prints (on my machine):
669 Blue is still 0.100000
670 Serialization of %color is 102 bytes long.
674 If you're using references as keys within your hash tables, you're bound
675 to disapointment when retrieving your data. Indeed, Perl stringifies
676 references used as hash table keys. If you later wish to access the
677 items via another reference stringification (i.e. using the same
678 reference that was used for the key originally to record the value into
679 the hash table), it will work because both references stringify to the
682 It won't work across a C<store> and C<retrieve> operations however, because
683 the addresses in the retrieved objects, which are part of the stringified
684 references, will probably differ from the original addresses. The
685 topology of your structure is preserved, but not hidden semantics
688 On platforms where it matters, be sure to call C<binmode()> on the
689 descriptors that you pass to Storable functions.
691 Storing data canonically that contains large hashes can be
692 significantly slower than storing the same data normally, as
693 temprorary arrays to hold the keys for each hash have to be allocated,
694 populated, sorted and freed. Some tests have shown a halving of the
695 speed of storing -- the exact penalty will depend on the complexity of
696 your data. There is no slowdown on retrieval.
700 You can't store GLOB, CODE, FORMLINE, etc... If you can define
701 semantics for those operations, feel free to enhance Storable so that
702 it can deal with them.
704 The store functions will C<croak> if they run into such references
705 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
706 case, the fatal message is turned in a warning and some
707 meaningless string is stored instead.
709 Setting C<$Storable::canonical> may not yield frozen strings that
710 compare equal due to possible stringification of numbers. When the
711 string version of a scalar exists, it is the form stored, therefore
712 if you happen to use your numbers as strings between two freezing
713 operations on the same data structures, you will get different
716 When storing doubles in network order, their value is stored as text.
717 However, you should also not expect non-numeric floating-point values
718 such as infinity and "not a number" to pass successfully through a
719 nstore()/retrieve() pair.
721 As Storable neither knows nor cares about character sets (although it
722 does know that characters may be more than eight bits wide), any difference
723 in the interpretation of character codes between a host and a target
724 system is your problem. In particular, if host and target use different
725 code points to represent the characters used in the text representation
726 of floating-point numbers, you will not be able be able to exchange
727 floating-point data, even with nstore().
731 Thank you to (in chronological order):
733 Jarkko Hietaniemi <jhi@iki.fi>
734 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
735 Benjamin A. Holzman <bah@ecnvantage.com>
736 Andrew Ford <A.Ford@ford-mason.co.uk>
737 Gisle Aas <gisle@aas.no>
738 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
739 Murray Nesbitt <murray@activestate.com>
740 Marc Lehmann <pcg@opengroup.org>
741 Justin Banks <justinb@wamnet.com>
742 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
743 Salvador Ortiz Garcia <sog@msg.com.mx>
744 Dominic Dunlop <domo@computer.org>
745 Erik Haugan <erik@solbors.no>
747 for their bug reports, suggestions and contributions.
749 Benjamin Holzman contributed the tied variable support, Andrew Ford
750 contributed the canonical order for hashes, and Gisle Aas fixed
751 a few misunderstandings of mine regarding the Perl internals,
752 and optimized the emission of "tags" in the output streams by
753 simply counting the objects instead of tagging them (leading to
754 a binary incompatibility for the Storable image starting at version
755 0.6--older images are of course still properly understood).
756 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
757 and reference to tied items support.
761 There is a Japanese translation of this man page available at
762 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
763 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
767 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>