1 ;# $Id: Storable.pm,v 1.0 2000/09/01 19:40:41 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 2000/09/01 19:40:41 ram
10 ;# Baseline for first official release.
15 package Storable; @ISA = qw(Exporter DynaLoader);
17 @EXPORT = qw(store retrieve);
19 nstore store_fd nstore_fd fd_retrieve
23 lock_store lock_nstore lock_retrieve
27 use vars qw($forgive_me $VERSION);
30 *AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
33 # Use of Log::Agent is optional
36 eval "use Log::Agent";
38 unless (defined @Log::Agent::EXPORT) {
48 # They might miss :flock in Fcntl
53 if (exists $Fcntl::EXPORT_TAGS{'flock'}) {
54 Fcntl->import(':flock');
65 sub retrieve_fd { &fd_retrieve } # Backward compatibility
74 # Store target object hierarchy, identified by a reference to its root.
75 # The stored object tree may later be retrieved to memory via retrieve.
76 # Returns undef if an I/O error occurred, in which case the file is
80 return _store(\&pstore, @_, 0);
86 # Same as store, but in network order.
89 return _store(\&net_pstore, @_, 0);
95 # Same as store, but flock the file first (advisory locking).
98 return _store(\&pstore, @_, 1);
104 # Same as nstore, but flock the file first (advisory locking).
107 return _store(\&net_pstore, @_, 1);
110 # Internal store to file routine
114 my ($file, $use_locking) = @_;
115 logcroak "not a reference" unless ref($self);
116 logcroak "too many arguments" unless @_ == 2; # No @foo in arglist
118 open(FILE, ">$file") || logcroak "can't create $file: $!";
119 binmode FILE; # Archaic systems...
121 flock(FILE, LOCK_EX) ||
122 logcroak "can't get exclusive lock on $file: $!";
124 # Unlocking will happen when FILE is closed
126 my $da = $@; # Don't mess if called from exception handler
128 # Call C routine nstore or pstore, depending on network order
129 eval { $ret = &$xsptr(*FILE, $self) };
130 close(FILE) or $ret = undef;
131 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
132 logcroak $@ if $@ =~ s/\.?\n$/,/;
134 return $ret ? $ret : undef;
140 # Same as store, but perform on an already opened file descriptor instead.
141 # Returns undef if an I/O error occurred.
144 return _store_fd(\&pstore, @_);
150 # Same as store_fd, but in network order.
153 my ($self, $file) = @_;
154 return _store_fd(\&net_pstore, @_);
157 # Internal store routine on opened file descriptor
162 logcroak "not a reference" unless ref($self);
163 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
164 my $fd = fileno($file);
165 logcroak "not a valid file descriptor" unless defined $fd;
166 my $da = $@; # Don't mess if called from exception handler
168 # Call C routine nstore or pstore, depending on network order
169 eval { $ret = &$xsptr($file, $self) };
170 logcroak $@ if $@ =~ s/\.?\n$/,/;
172 return $ret ? $ret : undef;
178 # Store oject and its hierarchy in memory and return a scalar
179 # containing the result.
182 _freeze(\&mstore, @_);
188 # Same as freeze but in network order.
191 _freeze(\&net_mstore, @_);
194 # Internal freeze routine
198 logcroak "not a reference" unless ref($self);
199 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
200 my $da = $@; # Don't mess if called from exception handler
202 # Call C routine mstore or net_mstore, depending on network order
203 eval { $ret = &$xsptr($self) };
204 logcroak $@ if $@ =~ s/\.?\n$/,/;
206 return $ret ? $ret : undef;
212 # Retrieve object hierarchy from disk, returning a reference to the root
213 # object of that tree.
222 # Same as retrieve, but with advisory locking.
228 # Internal retrieve routine
230 my ($file, $use_locking) = @_;
232 open(FILE, $file) || logcroak "can't open $file: $!";
233 binmode FILE; # Archaic systems...
235 my $da = $@; # Could be from exception handler
237 flock(FILE, LOCK_SH) || logcroak "can't get shared lock on $file: $!";
238 # Unlocking will happen when FILE is closed
240 eval { $self = pretrieve(*FILE) }; # Call C routine
242 logcroak $@ if $@ =~ s/\.?\n$/,/;
250 # Same as retrieve, but perform from an already opened file descriptor instead.
254 my $fd = fileno($file);
255 logcroak "not a valid file descriptor" unless defined $fd;
257 my $da = $@; # Could be from exception handler
258 eval { $self = pretrieve($file) }; # Call C routine
259 logcroak $@ if $@ =~ s/\.?\n$/,/;
267 # Recreate objects in memory from an existing frozen image created
268 # by freeze. If the frozen image passed is undef, return undef.
272 return undef unless defined $frozen;
274 my $da = $@; # Could be from exception handler
275 eval { $self = mretrieve($frozen) }; # Call C routine
276 logcroak $@ if $@ =~ s/\.?\n$/,/;
283 Storable - persistency for perl data structures
288 store \%table, 'file';
289 $hashref = retrieve('file');
291 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
294 nstore \%table, 'file';
295 $hashref = retrieve('file'); # There is NO nretrieve()
297 # Storing to and retrieving from an already opened file
298 store_fd \@array, \*STDOUT;
299 nstore_fd \%table, \*STDOUT;
300 $aryref = fd_retrieve(\*SOCKET);
301 $hashref = fd_retrieve(\*SOCKET);
303 # Serializing to memory
304 $serialized = freeze \%table;
305 %table_clone = %{ thaw($serialized) };
307 # Deep (recursive) cloning
308 $cloneref = dclone($ref);
311 use Storable qw(lock_store lock_nstore lock_retrieve)
312 lock_store \%table, 'file';
313 lock_nstore \%table, 'file';
314 $hashref = lock_retrieve('file');
318 The Storable package brings persistency to your perl data structures
319 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
320 convenientely stored to disk and retrieved at a later time.
322 It can be used in the regular procedural way by calling C<store> with
323 a reference to the object to be stored, along with the file name where
324 the image should be written.
325 The routine returns C<undef> for I/O problems or other internal error,
326 a true value otherwise. Serious errors are propagated as a C<die> exception.
328 To retrieve data stored to disk, use C<retrieve> with a file name,
329 and the objects stored into that file are recreated into memory for you,
330 a I<reference> to the root object being returned. In case an I/O error
331 occurs while reading, C<undef> is returned instead. Other serious
332 errors are propagated via C<die>.
334 Since storage is performed recursively, you might want to stuff references
335 to objects that share a lot of common data into a single array or hash
336 table, and then store that object. That way, when you retrieve back the
337 whole thing, the objects will continue to share what they originally shared.
339 At the cost of a slight header overhead, you may store to an already
340 opened file descriptor using the C<store_fd> routine, and retrieve
341 from a file via C<fd_retrieve>. Those names aren't imported by default,
342 so you will have to do that explicitely if you need those routines.
343 The file descriptor you supply must be already opened, for read
344 if you're going to retrieve and for write if you wish to store.
346 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
347 $hashref = fd_retrieve(*STDIN);
349 You can also store data in network order to allow easy sharing across
350 multiple platforms, or when storing on a socket known to be remotely
351 connected. The routines to call have an initial C<n> prefix for I<network>,
352 as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
353 correctly restored so you don't have to know whether you're restoring
354 from native or network ordered data. Double values are stored stringified
355 to ensure portability as well, at the slight risk of loosing some precision
356 in the last decimals.
358 When using C<fd_retrieve>, objects are retrieved in sequence, one
359 object (i.e. one recursive tree) per associated C<store_fd>.
361 If you're more from the object-oriented camp, you can inherit from
362 Storable and directly store your objects by invoking C<store> as
363 a method. The fact that the root of the to-be-stored tree is a
364 blessed reference (i.e. an object) is special-cased so that the
365 retrieve does not provide a reference to that object but rather the
366 blessed object reference itself. (Otherwise, you'd get a reference
367 to that blessed object).
371 The Storable engine can also store data into a Perl scalar instead, to
372 later retrieve them. This is mainly used to freeze a complex structure in
373 some safe compact memory place (where it can possibly be sent to another
374 process via some IPC, since freezing the structure also serializes it in
375 effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
376 out and recreate the original complex structure in memory.
378 Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
379 If you wish to send out the frozen scalar to another machine, use
380 C<nfreeze> instead to get a portable image.
382 Note that freezing an object structure and immediately thawing it
383 actually achieves a deep cloning of that structure:
385 dclone(.) = thaw(freeze(.))
387 Storable provides you with a C<dclone> interface which does not create
388 that intermediary scalar but instead freezes the structure in some
389 internal memory space and then immediatly thaws it out.
391 =head1 ADVISORY LOCKING
393 The C<lock_store> and C<lock_nstore> routine are equivalent to C<store>
394 and C<nstore>, only they get an exclusive lock on the file before
395 writing. Likewise, C<lock_retrieve> performs as C<retrieve>, but also
396 gets a shared lock on the file before reading.
398 Like with any advisory locking scheme, the protection only works if
399 you systematically use C<lock_store> and C<lock_retrieve>. If one
400 side of your application uses C<store> whilst the other uses C<lock_retrieve>,
401 you will get no protection at all.
403 The internal advisory locking is implemented using Perl's flock() routine.
404 If your system does not support any form of flock(), or if you share
405 your files across NFS, you might wish to use other forms of locking by
406 using modules like LockFile::Simple which lock a file using a filesystem
407 entry, instead of locking the file descriptor.
411 The heart of Storable is written in C for decent speed. Extra low-level
412 optimization have been made when manipulating perl internals, to
413 sacrifice encapsulation for the benefit of a greater speed.
415 =head1 CANONICAL REPRESENTATION
417 Normally Storable stores elements of hashes in the order they are
418 stored internally by Perl, i.e. pseudo-randomly. If you set
419 C<$Storable::canonical> to some C<TRUE> value, Storable will store
420 hashes with the elements sorted by their key. This allows you to
421 compare data structures by comparing their frozen representations (or
422 even the compressed frozen representations), which can be useful for
423 creating lookup tables for complicated queries.
425 Canonical order does not imply network order, those are two orthogonal
428 =head1 ERROR REPORTING
430 Storable uses the "exception" paradigm, in that it does not try to workaround
431 failures: if something bad happens, an exception is generated from the
432 caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
435 When Storable croaks, it tries to report the error via the C<logcroak()>
436 routine from the C<Log::Agent> package, if it is available.
442 Any class may define hooks that will be called during the serialization
443 and deserialization process on objects that are instances of that class.
444 Those hooks can redefine the way serialization is performed (and therefore,
445 how the symetrical deserialization should be conducted).
447 Since we said earlier:
449 dclone(.) = thaw(freeze(.))
451 everything we say about hooks should also hold for deep cloning. However,
452 hooks get to know whether the operation is a mere serialization, or a cloning.
454 Therefore, when serializing hooks are involved,
456 dclone(.) <> thaw(freeze(.))
458 Well, you could keep them in sync, but there's no guarantee it will always
459 hold on classes somebody else wrote. Besides, there is little to gain in
460 doing so: a serializing hook could only keep one attribute of an object,
461 which is probably not what should happen during a deep cloning of that
464 Here is the hooking interface:
468 =item C<STORABLE_freeze> I<obj>, I<cloning>
470 The serializing hook, called on the object during serialization. It can be
471 inherited, or defined in the class itself, like any other method.
473 Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
474 whether we're in a dclone() or a regular serialization via store() or freeze().
476 Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
477 is the serialized form to be used, and the optional $ref1, $ref2, etc... are
478 extra references that you wish to let the Storable engine serialize.
480 At deserialization time, you will be given back the same LIST, but all the
481 extra references will be pointing into the deserialized structure.
483 The B<first time> the hook is hit in a serialization flow, you may have it
484 return an empty list. That will signal the Storable engine to further
485 discard that hook for this class and to therefore revert to the default
486 serialization of the underlying Perl data. The hook will again be normally
487 processed in the next serialization.
489 Unless you know better, serializing hook should always say:
491 sub STORABLE_freeze {
492 my ($self, $cloning) = @_;
493 return if $cloning; # Regular default serialization
497 in order to keep reasonable dclone() semantics.
499 =item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
501 The deserializing hook called on the object during deserialization.
502 But wait. If we're deserializing, there's no object yet... right?
504 Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
505 you can view C<STORABLE_thaw> as an alternate creation routine.
507 This means the hook can be inherited like any other method, and that
508 I<obj> is your blessed reference for this particular instance.
510 The other arguments should look familiar if you know C<STORABLE_freeze>:
511 I<cloning> is true when we're part of a deep clone operation, I<serialized>
512 is the serialized string you returned to the engine in C<STORABLE_freeze>,
513 and there may be an optional list of references, in the same order you gave
514 them at serialization time, pointing to the deserialized objects (which
515 have been processed courtesy of the Storable engine).
517 It is up to you to use these information to populate I<obj> the way you want.
519 Returned value: none.
525 Predicates are not exportable. They must be called by explicitely prefixing
526 them with the Storable package name.
530 =item C<Storable::last_op_in_netorder>
532 The C<Storable::last_op_in_netorder()> predicate will tell you whether
533 network order was used in the last store or retrieve operation. If you
534 don't know how to use this, just forget about it.
536 =item C<Storable::is_storing>
538 Returns true if within a store operation (via STORABLE_freeze hook).
540 =item C<Storable::is_retrieving>
542 Returns true if within a retrieve operation, (via STORABLE_thaw hook).
548 With hooks comes the ability to recurse back to the Storable engine. Indeed,
549 hooks are regular Perl code, and Storable is convenient when it comes to
550 serialize and deserialize things, so why not use it to handle the
551 serialization string?
553 There are a few things you need to know however:
559 You can create endless loops if the things you serialize via freeze()
560 (for instance) point back to the object we're trying to serialize in the hook.
564 Shared references among objects will not stay shared: if we're serializing
565 the list of object [A, C] where both object A and C refer to the SAME object
566 B, and if there is a serializing hook in A that says freeze(B), then when
567 deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
568 a deep clone of B'. The topology was not preserved.
572 That's why C<STORABLE_freeze> lets you provide a list of references
573 to serialize. The engine guarantees that those will be serialized in the
574 same context as the other objects, and therefore that shared objects will
577 In the above [A, C] example, the C<STORABLE_freeze> hook could return:
579 ("something", $self->{B})
581 and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
582 would get back the reference to the B' object, deserialized for you.
584 Therefore, recursion should normally be avoided, but is nonetheless supported.
588 There is a new Clone module available on CPAN which implements deep cloning
589 natively, i.e. without freezing to memory and thawing the result. It is
590 aimed to replace Storable's dclone() some day. However, it does not currently
591 support Storable hooks to redefine the way deep cloning is performed.
595 Here are some code samples showing a possible usage of Storable:
597 use Storable qw(store retrieve freeze thaw dclone);
599 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
601 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
603 $colref = retrieve('/tmp/colors');
604 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
605 printf "Blue is still %lf\n", $colref->{'Blue'};
607 $colref2 = dclone(\%color);
609 $str = freeze(\%color);
610 printf "Serialization of %%color is %d bytes long.\n", length($str);
611 $colref3 = thaw($str);
613 which prints (on my machine):
615 Blue is still 0.100000
616 Serialization of %color is 102 bytes long.
620 If you're using references as keys within your hash tables, you're bound
621 to disapointment when retrieving your data. Indeed, Perl stringifies
622 references used as hash table keys. If you later wish to access the
623 items via another reference stringification (i.e. using the same
624 reference that was used for the key originally to record the value into
625 the hash table), it will work because both references stringify to the
628 It won't work across a C<store> and C<retrieve> operations however, because
629 the addresses in the retrieved objects, which are part of the stringified
630 references, will probably differ from the original addresses. The
631 topology of your structure is preserved, but not hidden semantics
634 On platforms where it matters, be sure to call C<binmode()> on the
635 descriptors that you pass to Storable functions.
637 Storing data canonically that contains large hashes can be
638 significantly slower than storing the same data normally, as
639 temprorary arrays to hold the keys for each hash have to be allocated,
640 populated, sorted and freed. Some tests have shown a halving of the
641 speed of storing -- the exact penalty will depend on the complexity of
642 your data. There is no slowdown on retrieval.
646 You can't store GLOB, CODE, FORMLINE, etc... If you can define
647 semantics for those operations, feel free to enhance Storable so that
648 it can deal with them.
650 The store functions will C<croak> if they run into such references
651 unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
652 case, the fatal message is turned in a warning and some
653 meaningless string is stored instead.
655 Setting C<$Storable::canonical> may not yield frozen strings that
656 compare equal due to possible stringification of numbers. When the
657 string version of a scalar exists, it is the form stored, therefore
658 if you happen to use your numbers as strings between two freezing
659 operations on the same data structures, you will get different
662 When storing doubles in network order, their value is stored as text.
663 However, you should also not expect non-numeric floating-point values
664 such as infinity and "not a number" to pass successfully through a
665 nstore()/retrieve() pair.
667 As Storable neither knows nor cares about character sets (although it
668 does know that characters may be more than eight bits wide), any difference
669 in the interpretation of character codes between a host and a target
670 system is your problem. In particular, if host and target use different
671 code points to represent the characters used in the text representation
672 of floating-point numbers, you will not be able be able to exchange
673 floating-point data, even with nstore().
677 Thank you to (in chronological order):
679 Jarkko Hietaniemi <jhi@iki.fi>
680 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
681 Benjamin A. Holzman <bah@ecnvantage.com>
682 Andrew Ford <A.Ford@ford-mason.co.uk>
683 Gisle Aas <gisle@aas.no>
684 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
685 Murray Nesbitt <murray@activestate.com>
686 Marc Lehmann <pcg@opengroup.org>
687 Justin Banks <justinb@wamnet.com>
688 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
689 Salvador Ortiz Garcia <sog@msg.com.mx>
690 Dominic Dunlop <domo@computer.org>
691 Erik Haugan <erik@solbors.no>
693 for their bug reports, suggestions and contributions.
695 Benjamin Holzman contributed the tied variable support, Andrew Ford
696 contributed the canonical order for hashes, and Gisle Aas fixed
697 a few misunderstandings of mine regarding the Perl internals,
698 and optimized the emission of "tags" in the output streams by
699 simply counting the objects instead of tagging them (leading to
700 a binary incompatibility for the Storable image starting at version
701 0.6--older images are of course still properly understood).
702 Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
703 and reference to tied items support.
707 There is a Japanese translation of this man page available at
708 http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
709 courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
713 Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>