1 ;# $Id: Storable.pm,v 1.0.1.5 2000/10/26 17:10:18 ram Exp ram $
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 1.0.1.5 2000/10/26 17:10:18 ram
10 ;# patch5: documented that store() and retrieve() can return undef
11 ;# patch5: added paragraph explaining the auto require for thaw hooks
13 ;# Revision 1.0.1.4 2000/10/23 18:02:57 ram
14 ;# patch4: protected calls to flock() for dos platform
15 ;# patch4: added logcarp emulation if they don't have Log::Agent
17 ;# $Log: Storable.pm,v $
18 ;# Revision 1.0 2000/09/01 19:40:41 ram
19 ;# Baseline for first official release.
24 package Storable; @ISA = qw(Exporter DynaLoader);
26 @EXPORT = qw(store retrieve);
28 nstore store_fd nstore_fd fd_retrieve
32 lock_store lock_nstore lock_retrieve
36 use vars qw($forgive_me $VERSION);
39 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
42 # Use of Log::Agent is optional
45 eval "use Log::Agent";
47 unless (defined @Log::Agent::EXPORT) {
61 # They might miss :flock in Fcntl
66 if (exists $Fcntl::EXPORT_TAGS{'flock'}) {
67 Fcntl->import(':flock');
79 sub retrieve_fd { &fd_retrieve } # Backward compatibility
88 # Store target object hierarchy, identified by a reference to its root.
89 # The stored object tree may later be retrieved to memory via retrieve.
90 # Returns undef if an I/O error occurred, in which case the file is
94 return _store(\&pstore, @_, 0);
100 # Same as store, but in network order.
103 return _store(\&net_pstore, @_, 0);
109 # Same as store, but flock the file first (advisory locking).
112 return _store(\&pstore, @_, 1);
118 # Same as nstore, but flock the file first (advisory locking).
121 return _store(\&net_pstore, @_, 1);
124 # Internal store to file routine
128 my ($file, $use_locking) = @_;
129 logcroak "not a reference" unless ref($self);
130 logcroak "too many arguments" unless @_ == 2; # No @foo in arglist
132 open(FILE, ">$file") || logcroak "can't create $file: $!";
133 binmode FILE; # Archaic systems...
136 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
139 flock(FILE, LOCK_EX) ||
140 logcroak "can't get exclusive lock on $file: $!";
142 # Unlocking will happen when FILE is closed
144 my $da = $@; # Don't mess if called from exception handler
146 # Call C routine nstore or pstore, depending on network order
147 eval { $ret = &$xsptr(*FILE, $self) };
148 close(FILE) or $ret = undef;
149 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
150 logcroak $@ if $@ =~ s/\.?\n$/,/;
152 return $ret ? $ret : undef;
158 # Same as store, but perform on an already opened file descriptor instead.
159 # Returns undef if an I/O error occurred.
162 return _store_fd(\&pstore, @_);
168 # Same as store_fd, but in network order.
171 my ($self, $file) = @_;
172 return _store_fd(\&net_pstore, @_);
175 # Internal store routine on opened file descriptor
180 logcroak "not a reference" unless ref($self);
181 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
182 my $fd = fileno($file);
183 logcroak "not a valid file descriptor" unless defined $fd;
184 my $da = $@; # Don't mess if called from exception handler
186 # Call C routine nstore or pstore, depending on network order
187 eval { $ret = &$xsptr($file, $self) };
188 logcroak $@ if $@ =~ s/\.?\n$/,/;
190 return $ret ? $ret : undef;
196 # Store oject and its hierarchy in memory and return a scalar
197 # containing the result.
200 _freeze(\&mstore, @_);
206 # Same as freeze but in network order.
209 _freeze(\&net_mstore, @_);
212 # Internal freeze routine
216 logcroak "not a reference" unless ref($self);
217 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
218 my $da = $@; # Don't mess if called from exception handler
220 # Call C routine mstore or net_mstore, depending on network order
221 eval { $ret = &$xsptr($self) };
222 logcroak $@ if $@ =~ s/\.?\n$/,/;
224 return $ret ? $ret : undef;
230 # Retrieve object hierarchy from disk, returning a reference to the root
231 # object of that tree.
240 # Same as retrieve, but with advisory locking.
246 # Internal retrieve routine
248 my ($file, $use_locking) = @_;
250 open(FILE, $file) || logcroak "can't open $file: $!";
251 binmode FILE; # Archaic systems...
253 my $da = $@; # Could be from exception handler
256 logcarp "Storable::lock_store: fcntl/flock emulation broken on $^O";
259 flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!";
260 # Unlocking will happen when FILE is closed
262 eval { $self = pretrieve(*FILE) }; # Call C routine
264 logcroak $@ if $@ =~ s/\.?\n$/,/;
272 # Same as retrieve, but perform from an already opened file descriptor instead.
276 my $fd = fileno($file);
277 logcroak "not a valid file descriptor" unless defined $fd;
279 my $da = $@; # Could be from exception handler
280 eval { $self = pretrieve($file) }; # Call C routine
281 logcroak $@ if $@ =~ s/\.?\n$/,/;
289 # Recreate objects in memory from an existing frozen image created
290 # by freeze. If the frozen image passed is undef, return undef.
294 return undef unless defined $frozen;
296 my $da = $@; # Could be from exception handler
297 eval { $self = mretrieve($frozen) }; # Call C routine
298 logcroak $@ if $@ =~ s/\.?\n$/,/;
305 Storable - persistency for perl data structures
310 store \%table, 'file';
311 $hashref = retrieve('file');
313 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
316 nstore \%table, 'file';
317 $hashref = retrieve('file'); # There is NO nretrieve()
319 # Storing to and retrieving from an already opened file
320 store_fd \@array, \*STDOUT;
321 nstore_fd \%table, \*STDOUT;
322 $aryref = fd_retrieve(\*SOCKET);
323 $hashref = fd_retrieve(\*SOCKET);
325 # Serializing to memory
326 $serialized = freeze \%table;
327 %table_clone = %{ thaw($serialized) };
329 # Deep (recursive) cloning
330 $cloneref = dclone($ref);
333 use Storable qw(lock_store lock_nstore lock_retrieve)
334 lock_store \%table, 'file';
335 lock_nstore \%table, 'file';
336 $hashref = lock_retrieve('file');
340 The Storable package brings persistency to your perl data structures
341 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
342 convenientely stored to disk and retrieved at a later time.
344 It can be used in the regular procedural way by calling C<store> with
345 a reference to the object to be stored, along with the file name where
346 the image should be written.
347 The routine returns C<undef> for I/O problems or other internal error,
348 a true value otherwise. Serious errors are propagated as a C<die> exception.
350 To retrieve data stored to disk, use C<retrieve> with a file name,
351 and the objects stored into that file are recreated into memory for you,
352 a I<reference> to the root object being returned. In case an I/O error
353 occurs while reading, C<undef> is returned instead. Other serious
354 errors are propagated via C<die>.
356 Since storage is performed recursively, you might want to stuff references
357 to objects that share a lot of common data into a single array or hash
358 table, and then store that object. That way, when you retrieve back the
359 whole thing, the objects will continue to share what they originally shared.
361 At the cost of a slight header overhead, you may store to an already
362 opened file descriptor using the C<store_fd> routine, and retrieve
363 from a file via C<fd_retrieve>. Those names aren't imported by default,
364 so you will have to do that explicitely if you need those routines.
365 The file descriptor you supply must be already opened, for read
366 if you're going to retrieve and for write if you wish to store.
368 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
369 $hashref = fd_retrieve(*STDIN);
371 You can also store data in network order to allow easy sharing across
372 multiple platforms, or when storing on a socket known to be remotely
373 connected. The routines to call have an initial C<n> prefix for I<network>,
374 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
375 correctly restored so you don't have to know whether you're restoring
376 from native or network ordered data. Double values are stored stringified
377 to ensure portability as well, at the slight risk of loosing some precision
378 in the last decimals.
380 When using C<fd_retrieve>, objects are retrieved in sequence, one
381 object (i.e. one recursive tree) per associated C<store_fd>.
383 If you're more from the object-oriented camp, you can inherit from
384 Storable and directly store your objects by invoking C<store> as
385 a method. The fact that the root of the to-be-stored tree is a
386 blessed reference (i.e. an object) is special-cased so that the
387 retrieve does not provide a reference to that object but rather the
388 blessed object reference itself. (Otherwise, you'd get a reference
389 to that blessed object).
393 The Storable engine can also store data into a Perl scalar instead, to
394 later retrieve them. This is mainly used to freeze a complex structure in
395 some safe compact memory place (where it can possibly be sent to another
396 process via some IPC, since freezing the structure also serializes it in
397 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
398 out and recreate the original complex structure in memory.
400 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
401 If you wish to send out the frozen scalar to another machine, use
402 C<nfreeze> instead to get a portable image.
404 Note that freezing an object structure and immediately thawing it
405 actually achieves a deep cloning of that structure:
407 dclone(.) = thaw(freeze(.))
409 Storable provides you with a C<dclone> interface which does not create
410 that intermediary scalar but instead freezes the structure in some
411 internal memory space and then immediatly thaws it out.
413 =head1 ADVISORY LOCKING
415 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
416 and C<nstore>, only they get an exclusive lock on the file before
417 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
418 gets a shared lock on the file before reading.
420 Like with any advisory locking scheme, the protection only works if
421 you systematically use C<lock_store> and C<lock_retrieve>. If one
422 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
423 you will get no protection at all.
425 The internal advisory locking is implemented using Perl's flock() routine.
426 If your system does not support any form of flock(), or if you share
427 your files across NFS, you might wish to use other forms of locking by
428 using modules like LockFile::Simple which lock a file using a filesystem
429 entry, instead of locking the file descriptor.
433 The heart of Storable is written in C for decent speed. Extra low-level
434 optimization have been made when manipulating perl internals, to
435 sacrifice encapsulation for the benefit of a greater speed.
437 =head1 CANONICAL REPRESENTATION
439 Normally Storable stores elements of hashes in the order they are
440 stored internally by Perl, i.e. pseudo-randomly. If you set
441 C<$Storable::canonical> to some C<TRUE> value, Storable will store
442 hashes with the elements sorted by their key. This allows you to
443 compare data structures by comparing their frozen representations (or
444 even the compressed frozen representations), which can be useful for
445 creating lookup tables for complicated queries.
447 Canonical order does not imply network order, those are two orthogonal
450 =head1 ERROR REPORTING
452 Storable uses the "exception" paradigm, in that it does not try to workaround
453 failures: if something bad happens, an exception is generated from the
454 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
457 When Storable croaks, it tries to report the error via the C<logcroak()>
458 routine from the C<Log::Agent> package, if it is available.
460 Normal errors are reported by having store() or retrieve() return C<undef>.
461 Such errors are usually I/O errors (or truncated stream errors at retrieval).
467 Any class may define hooks that will be called during the serialization
468 and deserialization process on objects that are instances of that class.
469 Those hooks can redefine the way serialization is performed (and therefore,
470 how the symetrical deserialization should be conducted).
472 Since we said earlier:
474 dclone(.) = thaw(freeze(.))
476 everything we say about hooks should also hold for deep cloning. However,
477 hooks get to know whether the operation is a mere serialization, or a cloning.
479 Therefore, when serializing hooks are involved,
481 dclone(.) <> thaw(freeze(.))
483 Well, you could keep them in sync, but there's no guarantee it will always
484 hold on classes somebody else wrote. Besides, there is little to gain in
485 doing so: a serializing hook could only keep one attribute of an object,
486 which is probably not what should happen during a deep cloning of that
489 Here is the hooking interface:
493 =item C<STORABLE_freeze> I<obj>, I<cloning>
495 The serializing hook, called on the object during serialization. It can be
496 inherited, or defined in the class itself, like any other method.
498 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
499 whether we're in a dclone() or a regular serialization via store() or freeze().
501 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
502 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
503 extra references that you wish to let the Storable engine serialize.
505 At deserialization time, you will be given back the same LIST, but all the
506 extra references will be pointing into the deserialized structure.
508 The B<first time> the hook is hit in a serialization flow, you may have it
509 return an empty list. That will signal the Storable engine to further
510 discard that hook for this class and to therefore revert to the default
511 serialization of the underlying Perl data. The hook will again be normally
512 processed in the next serialization.
514 Unless you know better, serializing hook should always say:
516 sub STORABLE_freeze {
517 my ($self, $cloning) = @_;
518 return if $cloning; # Regular default serialization
522 in order to keep reasonable dclone() semantics.
524 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
526 The deserializing hook called on the object during deserialization.
527 But wait. If we're deserializing, there's no object yet... right?
529 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
530 you can view C<STORABLE_thaw> as an alternate creation routine.
532 This means the hook can be inherited like any other method, and that
533 I<obj> is your blessed reference for this particular instance.
535 The other arguments should look familiar if you know C<STORABLE_freeze>:
536 I<cloning> is true when we're part of a deep clone operation, I<serialized>
537 is the serialized string you returned to the engine in C<STORABLE_freeze>,
538 and there may be an optional list of references, in the same order you gave
539 them at serialization time, pointing to the deserialized objects (which
540 have been processed courtesy of the Storable engine).
542 When the Storable engine does not find any C<STORABLE_thaw> hook routine,
543 it tries to load the class by requiring the package dynamically (using
544 the blessed package name), and then re-attempts the lookup. If at that
545 time the hook cannot be located, the engine croaks. Note that this mechanism
546 will fail if you define several classes in the same file, but perlmod(1)
549 It is up to you to use these information to populate I<obj> the way you want.
551 Returned value: none.
557 Predicates are not exportable. They must be called by explicitely prefixing
558 them with the Storable package name.
562 =item C<Storable::last_op_in_netorder>
564 The C<Storable::last_op_in_netorder()> predicate will tell you whether
565 network order was used in the last store or retrieve operation. If you
566 don't know how to use this, just forget about it.
568 =item C<Storable::is_storing>
570 Returns true if within a store operation (via STORABLE_freeze hook).
572 =item C<Storable::is_retrieving>
574 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
580 With hooks comes the ability to recurse back to the Storable engine. Indeed,
581 hooks are regular Perl code, and Storable is convenient when it comes to
582 serialize and deserialize things, so why not use it to handle the
583 serialization string?
585 There are a few things you need to know however:
591 You can create endless loops if the things you serialize via freeze()
592 (for instance) point back to the object we're trying to serialize in the hook.
596 Shared references among objects will not stay shared: if we're serializing
597 the list of object [A, C] where both object A and C refer to the SAME object
598 B, and if there is a serializing hook in A that says freeze(B), then when
599 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
600 a deep clone of B'. The topology was not preserved.
604 That's why C<STORABLE_freeze> lets you provide a list of references
605 to serialize. The engine guarantees that those will be serialized in the
606 same context as the other objects, and therefore that shared objects will
609 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
611 ("something", $self->{B})
613 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
614 would get back the reference to the B' object, deserialized for you.
616 Therefore, recursion should normally be avoided, but is nonetheless supported.
620 There is a new Clone module available on CPAN which implements deep cloning
621 natively, i.e. without freezing to memory and thawing the result. It is
622 aimed to replace Storable's dclone() some day. However, it does not currently
623 support Storable hooks to redefine the way deep cloning is performed.
627 Here are some code samples showing a possible usage of Storable:
629 use Storable qw(store retrieve freeze thaw dclone);
631 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
633 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
635 $colref = retrieve('/tmp/colors');
636 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
637 printf "Blue is still %lf\n", $colref->{'Blue'};
639 $colref2 = dclone(\%color);
641 $str = freeze(\%color);
642 printf "Serialization of %%color is %d bytes long.\n", length($str);
643 $colref3 = thaw($str);
645 which prints (on my machine):
647 Blue is still 0.100000
648 Serialization of %color is 102 bytes long.
652 If you're using references as keys within your hash tables, you're bound
653 to disapointment when retrieving your data. Indeed, Perl stringifies
654 references used as hash table keys. If you later wish to access the
655 items via another reference stringification (i.e. using the same
656 reference that was used for the key originally to record the value into
657 the hash table), it will work because both references stringify to the
660 It won't work across a C<store> and C<retrieve> operations however, because
661 the addresses in the retrieved objects, which are part of the stringified
662 references, will probably differ from the original addresses. The
663 topology of your structure is preserved, but not hidden semantics
666 On platforms where it matters, be sure to call C<binmode()> on the
667 descriptors that you pass to Storable functions.
669 Storing data canonically that contains large hashes can be
670 significantly slower than storing the same data normally, as
671 temprorary arrays to hold the keys for each hash have to be allocated,
672 populated, sorted and freed. Some tests have shown a halving of the
673 speed of storing -- the exact penalty will depend on the complexity of
674 your data. There is no slowdown on retrieval.
678 You can't store GLOB, CODE, FORMLINE, etc... If you can define
679 semantics for those operations, feel free to enhance Storable so that
680 it can deal with them.
682 The store functions will C<croak> if they run into such references
683 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
684 case, the fatal message is turned in a warning and some
685 meaningless string is stored instead.
687 Setting C<$Storable::canonical> may not yield frozen strings that
688 compare equal due to possible stringification of numbers. When the
689 string version of a scalar exists, it is the form stored, therefore
690 if you happen to use your numbers as strings between two freezing
691 operations on the same data structures, you will get different
694 When storing doubles in network order, their value is stored as text.
695 However, you should also not expect non-numeric floating-point values
696 such as infinity and "not a number" to pass successfully through a
697 nstore()/retrieve() pair.
699 As Storable neither knows nor cares about character sets (although it
700 does know that characters may be more than eight bits wide), any difference
701 in the interpretation of character codes between a host and a target
702 system is your problem. In particular, if host and target use different
703 code points to represent the characters used in the text representation
704 of floating-point numbers, you will not be able be able to exchange
705 floating-point data, even with nstore().
709 Thank you to (in chronological order):
711 Jarkko Hietaniemi <jhi@iki.fi>
712 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
713 Benjamin A. Holzman <bah@ecnvantage.com>
714 Andrew Ford <A.Ford@ford-mason.co.uk>
715 Gisle Aas <gisle@aas.no>
716 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
717 Murray Nesbitt <murray@activestate.com>
718 Marc Lehmann <pcg@opengroup.org>
719 Justin Banks <justinb@wamnet.com>
720 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
721 Salvador Ortiz Garcia <sog@msg.com.mx>
722 Dominic Dunlop <domo@computer.org>
723 Erik Haugan <erik@solbors.no>
725 for their bug reports, suggestions and contributions.
727 Benjamin Holzman contributed the tied variable support, Andrew Ford
728 contributed the canonical order for hashes, and Gisle Aas fixed
729 a few misunderstandings of mine regarding the Perl internals,
730 and optimized the emission of "tags" in the output streams by
731 simply counting the objects instead of tagging them (leading to
732 a binary incompatibility for the Storable image starting at version
733 0.6--older images are of course still properly understood).
734 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
735 and reference to tied items support.
739 There is a Japanese translation of this man page available at
740 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
741 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
745 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>