This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Mention known sfio+linux buglet.
[perl5.git] / ext / Storable / Storable.pm
CommitLineData
f0ffaed8 1;# $Id: Storable.pm,v 0.7.1.3 2000/08/23 22:49:25 ram Exp $
7a6a85bf
RG
2;#
3;# Copyright (c) 1995-2000, Raphael Manfredi
4;#
5;# You may redistribute only under the terms of the Artistic License,
6;# as specified in the README file that comes with the distribution.
7;#
8;# $Log: Storable.pm,v $
f0ffaed8
JH
9;# Revision 0.7.1.3 2000/08/23 22:49:25 ram
10;# patch3: updated version number
11;#
7a6a85bf
RG
12;# Revision 0.7.1.2 2000/08/14 07:18:40 ram
13;# patch2: increased version number
14;#
15;# Revision 0.7.1.1 2000/08/13 20:08:58 ram
16;# patch1: mention new Clone(3) extension in SEE ALSO
17;# patch1: contributor Marc Lehmann added overloading and ref to tied items
18;# patch1: updated e-mail from Benjamin Holzman
19;#
20;# Revision 0.7 2000/08/03 22:04:44 ram
21;# Baseline for second beta release.
22;#
23
24require DynaLoader;
25require Exporter;
26package Storable; @ISA = qw(Exporter DynaLoader);
27
28@EXPORT = qw(store retrieve);
29@EXPORT_OK = qw(
30 nstore store_fd nstore_fd retrieve_fd
31 freeze nfreeze thaw
32 dclone
33);
34
35use AutoLoader;
36use vars qw($forgive_me $VERSION);
37
f0ffaed8 38$VERSION = '0.703';
7a6a85bf
RG
39*AUTOLOAD = \&AutoLoader::AUTOLOAD; # Grrr...
40
41#
42# Use of Log::Agent is optional
43#
44
45eval "use Log::Agent";
46
47unless (defined @Log::Agent::EXPORT) {
48 eval q{
49 sub logcroak {
50 require Carp;
51 Carp::croak(@_);
52 }
53 };
54}
55
56sub logcroak;
57
cb3d9de5
JH
58# 8.3 limitation avoidance trickery. --mjtguy
59sub retrieve_fd { goto &fdretrieve };
60
7a6a85bf
RG
61bootstrap Storable;
621;
63__END__
64
65#
66# store
67#
68# Store target object hierarchy, identified by a reference to its root.
69# The stored object tree may later be retrieved to memory via retrieve.
70# Returns undef if an I/O error occurred, in which case the file is
71# removed.
72#
73sub store {
74 return _store(\&pstore, @_);
75}
76
77#
78# nstore
79#
80# Same as store, but in network order.
81#
82sub nstore {
83 return _store(\&net_pstore, @_);
84}
85
86# Internal store to file routine
87sub _store {
88 my $xsptr = shift;
89 my $self = shift;
90 my ($file) = @_;
91 logcroak "not a reference" unless ref($self);
92 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
93 local *FILE;
94 open(FILE, ">$file") || logcroak "can't create $file: $!";
95 binmode FILE; # Archaic systems...
96 my $da = $@; # Don't mess if called from exception handler
97 my $ret;
98 # Call C routine nstore or pstore, depending on network order
99 eval { $ret = &$xsptr(*FILE, $self) };
100 close(FILE) or $ret = undef;
101 unlink($file) or warn "Can't unlink $file: $!\n" if $@ || !defined $ret;
102 logcroak $@ if $@ =~ s/\.?\n$/,/;
103 $@ = $da;
104 return $ret ? $ret : undef;
105}
106
107#
108# store_fd
109#
110# Same as store, but perform on an already opened file descriptor instead.
111# Returns undef if an I/O error occurred.
112#
113sub store_fd {
114 return _store_fd(\&pstore, @_);
115}
116
117#
118# nstore_fd
119#
120# Same as store_fd, but in network order.
121#
122sub nstore_fd {
123 my ($self, $file) = @_;
124 return _store_fd(\&net_pstore, @_);
125}
126
127# Internal store routine on opened file descriptor
128sub _store_fd {
129 my $xsptr = shift;
130 my $self = shift;
131 my ($file) = @_;
132 logcroak "not a reference" unless ref($self);
133 logcroak "too many arguments" unless @_ == 1; # No @foo in arglist
134 my $fd = fileno($file);
135 logcroak "not a valid file descriptor" unless defined $fd;
136 my $da = $@; # Don't mess if called from exception handler
137 my $ret;
138 # Call C routine nstore or pstore, depending on network order
139 eval { $ret = &$xsptr($file, $self) };
140 logcroak $@ if $@ =~ s/\.?\n$/,/;
141 $@ = $da;
142 return $ret ? $ret : undef;
143}
144
145#
146# freeze
147#
148# Store oject and its hierarchy in memory and return a scalar
149# containing the result.
150#
151sub freeze {
152 _freeze(\&mstore, @_);
153}
154
155#
156# nfreeze
157#
158# Same as freeze but in network order.
159#
160sub nfreeze {
161 _freeze(\&net_mstore, @_);
162}
163
164# Internal freeze routine
165sub _freeze {
166 my $xsptr = shift;
167 my $self = shift;
168 logcroak "not a reference" unless ref($self);
169 logcroak "too many arguments" unless @_ == 0; # No @foo in arglist
170 my $da = $@; # Don't mess if called from exception handler
171 my $ret;
172 # Call C routine mstore or net_mstore, depending on network order
173 eval { $ret = &$xsptr($self) };
174 logcroak $@ if $@ =~ s/\.?\n$/,/;
175 $@ = $da;
176 return $ret ? $ret : undef;
177}
178
179#
180# retrieve
181#
182# Retrieve object hierarchy from disk, returning a reference to the root
183# object of that tree.
184#
185sub retrieve {
186 my ($file) = @_;
187 local *FILE;
188 open(FILE, "$file") || logcroak "can't open $file: $!";
189 binmode FILE; # Archaic systems...
190 my $self;
191 my $da = $@; # Could be from exception handler
192 eval { $self = pretrieve(*FILE) }; # Call C routine
193 close(FILE);
194 logcroak $@ if $@ =~ s/\.?\n$/,/;
195 $@ = $da;
196 return $self;
197}
198
199#
cb3d9de5 200# fdretrieve
7a6a85bf
RG
201#
202# Same as retrieve, but perform from an already opened file descriptor instead.
203#
cb3d9de5 204sub fdretrieve {
7a6a85bf
RG
205 my ($file) = @_;
206 my $fd = fileno($file);
207 logcroak "not a valid file descriptor" unless defined $fd;
208 my $self;
209 my $da = $@; # Could be from exception handler
210 eval { $self = pretrieve($file) }; # Call C routine
211 logcroak $@ if $@ =~ s/\.?\n$/,/;
212 $@ = $da;
213 return $self;
214}
215
216#
217# thaw
218#
219# Recreate objects in memory from an existing frozen image created
220# by freeze. If the frozen image passed is undef, return undef.
221#
222sub thaw {
223 my ($frozen) = @_;
224 return undef unless defined $frozen;
225 my $self;
226 my $da = $@; # Could be from exception handler
227 eval { $self = mretrieve($frozen) }; # Call C routine
228 logcroak $@ if $@ =~ s/\.?\n$/,/;
229 $@ = $da;
230 return $self;
231}
232
233=head1 NAME
234
235Storable - persistency for perl data structures
236
237=head1 SYNOPSIS
238
239 use Storable;
240 store \%table, 'file';
241 $hashref = retrieve('file');
242
243 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
244
245 # Network order
246 nstore \%table, 'file';
247 $hashref = retrieve('file'); # There is NO nretrieve()
248
249 # Storing to and retrieving from an already opened file
250 store_fd \@array, \*STDOUT;
251 nstore_fd \%table, \*STDOUT;
252 $aryref = retrieve_fd(\*SOCKET);
253 $hashref = retrieve_fd(\*SOCKET);
254
255 # Serializing to memory
256 $serialized = freeze \%table;
257 %table_clone = %{ thaw($serialized) };
258
259 # Deep (recursive) cloning
260 $cloneref = dclone($ref);
261
262=head1 DESCRIPTION
263
264The Storable package brings persistency to your perl data structures
265containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can be
266convenientely stored to disk and retrieved at a later time.
267
268It can be used in the regular procedural way by calling C<store> with
269a reference to the object to be stored, along with the file name where
270the image should be written.
271The routine returns C<undef> for I/O problems or other internal error,
272a true value otherwise. Serious errors are propagated as a C<die> exception.
273
274To retrieve data stored to disk, use C<retrieve> with a file name,
275and the objects stored into that file are recreated into memory for you,
276a I<reference> to the root object being returned. In case an I/O error
277occurs while reading, C<undef> is returned instead. Other serious
278errors are propagated via C<die>.
279
280Since storage is performed recursively, you might want to stuff references
281to objects that share a lot of common data into a single array or hash
282table, and then store that object. That way, when you retrieve back the
283whole thing, the objects will continue to share what they originally shared.
284
285At the cost of a slight header overhead, you may store to an already
286opened file descriptor using the C<store_fd> routine, and retrieve
287from a file via C<retrieve_fd>. Those names aren't imported by default,
288so you will have to do that explicitely if you need those routines.
289The file descriptor you supply must be already opened, for read
290if you're going to retrieve and for write if you wish to store.
291
292 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
293 $hashref = retrieve_fd(*STDIN);
294
295You can also store data in network order to allow easy sharing across
296multiple platforms, or when storing on a socket known to be remotely
297connected. The routines to call have an initial C<n> prefix for I<network>,
298as in C<nstore> and C<nstore_fd>. At retrieval time, your data will be
299correctly restored so you don't have to know whether you're restoring
300from native or network ordered data.
301
302When using C<retrieve_fd>, objects are retrieved in sequence, one
303object (i.e. one recursive tree) per associated C<store_fd>.
304
305If you're more from the object-oriented camp, you can inherit from
306Storable and directly store your objects by invoking C<store> as
307a method. The fact that the root of the to-be-stored tree is a
308blessed reference (i.e. an object) is special-cased so that the
309retrieve does not provide a reference to that object but rather the
310blessed object reference itself. (Otherwise, you'd get a reference
311to that blessed object).
312
313=head1 MEMORY STORE
314
315The Storable engine can also store data into a Perl scalar instead, to
316later retrieve them. This is mainly used to freeze a complex structure in
317some safe compact memory place (where it can possibly be sent to another
318process via some IPC, since freezing the structure also serializes it in
319effect). Later on, and maybe somewhere else, you can thaw the Perl scalar
320out and recreate the original complex structure in memory.
321
322Surprisingly, the routines to be called are named C<freeze> and C<thaw>.
323If you wish to send out the frozen scalar to another machine, use
324C<nfreeze> instead to get a portable image.
325
326Note that freezing an object structure and immediately thawing it
327actually achieves a deep cloning of that structure:
328
329 dclone(.) = thaw(freeze(.))
330
331Storable provides you with a C<dclone> interface which does not create
332that intermediary scalar but instead freezes the structure in some
333internal memory space and then immediatly thaws it out.
334
335=head1 SPEED
336
337The heart of Storable is written in C for decent speed. Extra low-level
338optimization have been made when manipulating perl internals, to
339sacrifice encapsulation for the benefit of a greater speed.
340
341=head1 CANONICAL REPRESENTATION
342
343Normally Storable stores elements of hashes in the order they are
344stored internally by Perl, i.e. pseudo-randomly. If you set
345C<$Storable::canonical> to some C<TRUE> value, Storable will store
346hashes with the elements sorted by their key. This allows you to
347compare data structures by comparing their frozen representations (or
348even the compressed frozen representations), which can be useful for
349creating lookup tables for complicated queries.
350
351Canonical order does not imply network order, those are two orthogonal
352settings.
353
354=head1 ERROR REPORTING
355
356Storable uses the "exception" paradigm, in that it does not try to workaround
357failures: if something bad happens, an exception is generated from the
358caller's perspective (see L<Carp> and C<croak()>). Use eval {} to trap
359those exceptions.
360
361When Storable croaks, it tries to report the error via the C<logcroak()>
362routine from the C<Log::Agent> package, if it is available.
363
364=head1 WIZARDS ONLY
365
366=head2 Hooks
367
368Any class may define hooks that will be called during the serialization
369and deserialization process on objects that are instances of that class.
370Those hooks can redefine the way serialization is performed (and therefore,
371how the symetrical deserialization should be conducted).
372
373Since we said earlier:
374
375 dclone(.) = thaw(freeze(.))
376
377everything we say about hooks should also hold for deep cloning. However,
378hooks get to know whether the operation is a mere serialization, or a cloning.
379
380Therefore, when serializing hooks are involved,
381
382 dclone(.) <> thaw(freeze(.))
383
384Well, you could keep them in sync, but there's no guarantee it will always
385hold on classes somebody else wrote. Besides, there is little to gain in
386doing so: a serializing hook could only keep one attribute of an object,
387which is probably not what should happen during a deep cloning of that
388same object.
389
390Here is the hooking interface:
391
392=over
393
394=item C<STORABLE_freeze> I<obj>, I<cloning>
395
396The serializing hook, called on the object during serialization. It can be
397inherited, or defined in the class itself, like any other method.
398
399Arguments: I<obj> is the object to serialize, I<cloning> is a flag indicating
400whether we're in a dclone() or a regular serialization via store() or freeze().
401
402Returned value: A LIST C<($serialized, $ref1, $ref2, ...)> where $serialized
403is the serialized form to be used, and the optional $ref1, $ref2, etc... are
404extra references that you wish to let the Storable engine serialize.
405
406At deserialization time, you will be given back the same LIST, but all the
407extra references will be pointing into the deserialized structure.
408
409The B<first time> the hook is hit in a serialization flow, you may have it
410return an empty list. That will signal the Storable engine to further
411discard that hook for this class and to therefore revert to the default
412serialization of the underlying Perl data. The hook will again be normally
413processed in the next serialization.
414
415Unless you know better, serializing hook should always say:
416
417 sub STORABLE_freeze {
418 my ($self, $cloning) = @_;
419 return if $cloning; # Regular default serialization
420 ....
421 }
422
423in order to keep reasonable dclone() semantics.
424
425=item C<STORABLE_thaw> I<obj>, I<cloning>, I<serialized>, ...
426
427The deserializing hook called on the object during deserialization.
428But wait. If we're deserializing, there's no object yet... right?
429
430Wrong: the Storable engine creates an empty one for you. If you know Eiffel,
431you can view C<STORABLE_thaw> as an alternate creation routine.
432
433This means the hook can be inherited like any other method, and that
434I<obj> is your blessed reference for this particular instance.
435
436The other arguments should look familiar if you know C<STORABLE_freeze>:
437I<cloning> is true when we're part of a deep clone operation, I<serialized>
438is the serialized string you returned to the engine in C<STORABLE_freeze>,
439and there may be an optional list of references, in the same order you gave
440them at serialization time, pointing to the deserialized objects (which
441have been processed courtesy of the Storable engine).
442
443It is up to you to use these information to populate I<obj> the way you want.
444
445Returned value: none.
446
447=back
448
449=head2 Predicates
450
451Predicates are not exportable. They must be called by explicitely prefixing
452them with the Storable package name.
453
454=over
455
456=item C<Storable::last_op_in_netorder>
457
458The C<Storable::last_op_in_netorder()> predicate will tell you whether
459network order was used in the last store or retrieve operation. If you
460don't know how to use this, just forget about it.
461
462=item C<Storable::is_storing>
463
464Returns true if within a store operation (via STORABLE_freeze hook).
465
466=item C<Storable::is_retrieving>
467
468Returns true if within a retrieve operation, (via STORABLE_thaw hook).
469
470=back
471
472=head2 Recursion
473
474With hooks comes the ability to recurse back to the Storable engine. Indeed,
475hooks are regular Perl code, and Storable is convenient when it comes to
476serialize and deserialize things, so why not use it to handle the
477serialization string?
478
479There are a few things you need to know however:
480
481=over
482
483=item *
484
485You can create endless loops if the things you serialize via freeze()
486(for instance) point back to the object we're trying to serialize in the hook.
487
488=item *
489
490Shared references among objects will not stay shared: if we're serializing
491the list of object [A, C] where both object A and C refer to the SAME object
492B, and if there is a serializing hook in A that says freeze(B), then when
493deserializing, we'll get [A', C'] where A' refers to B', but C' refers to D,
494a deep clone of B'. The topology was not preserved.
495
496=back
497
498That's why C<STORABLE_freeze> lets you provide a list of references
499to serialize. The engine guarantees that those will be serialized in the
500same context as the other objects, and therefore that shared objects will
501stay shared.
502
503In the above [A, C] example, the C<STORABLE_freeze> hook could return:
504
505 ("something", $self->{B})
506
507and the B part would be serialized by the engine. In C<STORABLE_thaw>, you
508would get back the reference to the B' object, deserialized for you.
509
510Therefore, recursion should normally be avoided, but is nonetheless supported.
511
512=head2 Deep Cloning
513
514There is a new Clone module available on CPAN which implements deep cloning
515natively, i.e. without freezing to memory and thawing the result. It is
516aimed to replace Storable's dclone() some day. However, it does not currently
517support Storable hooks to redefine the way deep cloning is performed.
518
519=head1 EXAMPLES
520
521Here are some code samples showing a possible usage of Storable:
522
523 use Storable qw(store retrieve freeze thaw dclone);
524
525 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
526
527 store(\%color, '/tmp/colors') or die "Can't store %a in /tmp/colors!\n";
528
529 $colref = retrieve('/tmp/colors');
530 die "Unable to retrieve from /tmp/colors!\n" unless defined $colref;
531 printf "Blue is still %lf\n", $colref->{'Blue'};
532
533 $colref2 = dclone(\%color);
534
535 $str = freeze(\%color);
536 printf "Serialization of %%color is %d bytes long.\n", length($str);
537 $colref3 = thaw($str);
538
539which prints (on my machine):
540
541 Blue is still 0.100000
542 Serialization of %color is 102 bytes long.
543
544=head1 WARNING
545
546If you're using references as keys within your hash tables, you're bound
547to disapointment when retrieving your data. Indeed, Perl stringifies
548references used as hash table keys. If you later wish to access the
549items via another reference stringification (i.e. using the same
550reference that was used for the key originally to record the value into
551the hash table), it will work because both references stringify to the
552same string.
553
554It won't work across a C<store> and C<retrieve> operations however, because
555the addresses in the retrieved objects, which are part of the stringified
556references, will probably differ from the original addresses. The
557topology of your structure is preserved, but not hidden semantics
558like those.
559
560On platforms where it matters, be sure to call C<binmode()> on the
561descriptors that you pass to Storable functions.
562
563Storing data canonically that contains large hashes can be
564significantly slower than storing the same data normally, as
565temprorary arrays to hold the keys for each hash have to be allocated,
566populated, sorted and freed. Some tests have shown a halving of the
567speed of storing -- the exact penalty will depend on the complexity of
568your data. There is no slowdown on retrieval.
569
570=head1 BUGS
571
572You can't store GLOB, CODE, FORMLINE, etc... If you can define
573semantics for those operations, feel free to enhance Storable so that
574it can deal with them.
575
576The store functions will C<croak> if they run into such references
577unless you set C<$Storable::forgive_me> to some C<TRUE> value. In that
578case, the fatal message is turned in a warning and some
579meaningless string is stored instead.
580
581Setting C<$Storable::canonical> may not yield frozen strings that
582compare equal due to possible stringification of numbers. When the
583string version of a scalar exists, it is the form stored, therefore
584if you happen to use your numbers as strings between two freezing
585operations on the same data structures, you will get different
586results.
587
588Due to the aforementionned optimizations, Storable is at the mercy
589of perl's internal redesign or structure changes. If that bothers
590you, you can try convincing Larry that what is used in Storable
591should be documented and consistently kept in future revisions.
592
593=head1 CREDITS
594
595Thank you to (in chronological order):
596
597 Jarkko Hietaniemi <jhi@iki.fi>
598 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
599 Benjamin A. Holzman <bah@ecnvantage.com>
600 Andrew Ford <A.Ford@ford-mason.co.uk>
601 Gisle Aas <gisle@aas.no>
602 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
603 Murray Nesbitt <murray@activestate.com>
604 Marc Lehmann <pcg@opengroup.org>
605
606for their bug reports, suggestions and contributions.
607
608Benjamin Holzman contributed the tied variable support, Andrew Ford
609contributed the canonical order for hashes, and Gisle Aas fixed
610a few misunderstandings of mine regarding the Perl internals,
611and optimized the emission of "tags" in the output streams by
612simply counting the objects instead of tagging them (leading to
613a binary incompatibility for the Storable image starting at version
6140.6--older images are of course still properly understood).
615Murray Nesbitt made Storable thread-safe. Marc Lehmann added overloading
616and reference to tied items support.
617
618=head1 TRANSLATIONS
619
620There is a Japanese translation of this man page available at
621http://member.nifty.ne.jp/hippo2000/perltips/storable.htm ,
622courtesy of Kawai, Takanori <kawai@nippon-rad.co.jp>.
623
624=head1 AUTHOR
625
626Raphael Manfredi F<E<lt>Raphael_Manfredi@pobox.comE<gt>>
627
628=head1 SEE ALSO
629
630Clone(3).
631
632=cut
633