1Storable(3) User Contributed Perl Documentation Storable(3)
2
6 Storable - persistence for Perl data structures
7
9 use Storable;
10 store \%table, 'file';
11 $hashref = retrieve('file');
12 use Storable qw(nstore store_fd nstore_fd freeze thaw dclone);
14 # Network order
16 nstore \%table, 'file';
17 $hashref = retrieve('file'); # There is NO nretrieve()
18 # Storing to and retrieving from an already opened file
20 store_fd \@array, \*STDOUT;
21 nstore_fd \%table, \*STDOUT;
22 $aryref = fd_retrieve(\*SOCKET);
23 $hashref = fd_retrieve(\*SOCKET);
24 # Serializing to memory
26 $serialized = freeze \%table;
27 %table_clone = %{ thaw($serialized) };
28 # Deep (recursive) cloning
30 $cloneref = dclone($ref);
31 # Advisory locking
33 use Storable qw(lock_store lock_nstore lock_retrieve)
34 lock_store \%table, 'file';
35 lock_nstore \%table, 'file';
36 $hashref = lock_retrieve('file');
37
39 The Storable package brings persistence to your Perl data structures
40 containing SCALAR, ARRAY, HASH or REF objects, i.e. anything that can
41 be conveniently stored to disk and retrieved at a later time.
42 It can be used in the regular procedural way by calling "store" with a
44 reference to the object to be stored, along with the file name where
45 the image should be written.
46 The routine returns "undef" for I/O problems or other internal error, a
48 true value otherwise. Serious errors are propagated as a "die"
49 exception.
50 To retrieve data stored to disk, use "retrieve" with a file name. The
52 objects stored into that file are recreated into memory for you, and a
53 reference to the root object is returned. In case an I/O error occurs
54 while reading, "undef" is returned instead. Other serious errors are
55 propagated via "die".
56 Since storage is performed recursively, you might want to stuff
58 references to objects that share a lot of common data into a single
59 array or hash table, and then store that object. That way, when you
60 retrieve back the whole thing, the objects will continue to share what
61 they originally shared.
62 At the cost of a slight header overhead, you may store to an already
64 opened file descriptor using the "store_fd" routine, and retrieve from
65 a file via "fd_retrieve". Those names aren't imported by default, so
66 you will have to do that explicitly if you need those routines. The
67 file descriptor you supply must be already opened, for read if you're
68 going to retrieve and for write if you wish to store.
69 store_fd(\%table, *STDOUT) || die "can't store to stdout\n";
71 $hashref = fd_retrieve(*STDIN);
72 You can also store data in network order to allow easy sharing across
74 multiple platforms, or when storing on a socket known to be remotely
75 connected. The routines to call have an initial "n" prefix for network,
76 as in "nstore" and "nstore_fd". At retrieval time, your data will be
77 correctly restored so you don't have to know whether you're restoring
78 from native or network ordered data. Double values are stored
79 stringified to ensure portability as well, at the slight risk of
80 loosing some precision in the last decimals.
81 When using "fd_retrieve", objects are retrieved in sequence, one object
83 (i.e. one recursive tree) per associated "store_fd".
84 If you're more from the object-oriented camp, you can inherit from
86 Storable and directly store your objects by invoking "store" as a
87 method. The fact that the root of the to-be-stored tree is a blessed
88 reference (i.e. an object) is special-cased so that the retrieve does
89 not provide a reference to that object but rather the blessed object
90 reference itself. (Otherwise, you'd get a reference to that blessed
91 object).
92
94 The Storable engine can also store data into a Perl scalar instead, to
95 later retrieve them. This is mainly used to freeze a complex structure
96 in some safe compact memory place (where it can possibly be sent to
97 another process via some IPC, since freezing the structure also
98 serializes it in effect). Later on, and maybe somewhere else, you can
99 thaw the Perl scalar out and recreate the original complex structure in
100 memory.
101 Surprisingly, the routines to be called are named "freeze" and "thaw".
103 If you wish to send out the frozen scalar to another machine, use
104 "nfreeze" instead to get a portable image.
105 Note that freezing an object structure and immediately thawing it
107 actually achieves a deep cloning of that structure:
108 dclone(.) = thaw(freeze(.))
110 Storable provides you with a "dclone" interface which does not create
112 that intermediary scalar but instead freezes the structure in some
113 internal memory space and then immediately thaws it out.
114
116 The "lock_store" and "lock_nstore" routine are equivalent to "store"
117 and "nstore", except that they get an exclusive lock on the file before
118 writing. Likewise, "lock_retrieve" does the same as "retrieve", but
119 also gets a shared lock on the file before reading.
120 As with any advisory locking scheme, the protection only works if you
122 systematically use "lock_store" and "lock_retrieve". If one side of
123 your application uses "store" whilst the other uses "lock_retrieve",
124 you will get no protection at all.
125 The internal advisory locking is implemented using Perl's flock()
127 routine. If your system does not support any form of flock(), or if
128 you share your files across NFS, you might wish to use other forms of
129 locking by using modules such as LockFile::Simple which lock a file
130 using a filesystem entry, instead of locking the file descriptor.
131
133 The heart of Storable is written in C for decent speed. Extra low-level
134 optimizations have been made when manipulating perl internals, to
135 sacrifice encapsulation for the benefit of greater speed.
136
138 Normally, Storable stores elements of hashes in the order they are
139 stored internally by Perl, i.e. pseudo-randomly. If you set
140 $Storable::canonical to some "TRUE" value, Storable will store hashes
141 with the elements sorted by their key. This allows you to compare data
142 structures by comparing their frozen representations (or even the
143 compressed frozen representations), which can be useful for creating
144 lookup tables for complicated queries.
145 Canonical order does not imply network order; those are two orthogonal
147 settings.
148
150 Since Storable version 2.05, CODE references may be serialized with the
151 help of B::Deparse. To enable this feature, set $Storable::Deparse to a
152 true value. To enable deserialization, $Storable::Eval should be set to
153 a true value. Be aware that deserialization is done through "eval",
154 which is dangerous if the Storable file contains malicious data. You
155 can set $Storable::Eval to a subroutine reference which would be used
156 instead of "eval". See below for an example using a Safe compartment
157 for deserialization of CODE references.
158 If $Storable::Deparse and/or $Storable::Eval are set to false values,
160 then the value of $Storable::forgive_me (see below) is respected while
161 serializing and deserializing.
162
164 This release of Storable can be used on a newer version of Perl to
165 serialize data which is not supported by earlier Perls. By default,
166 Storable will attempt to do the right thing, by "croak()"ing if it
167 encounters data that it cannot deserialize. However, the defaults can
168 be changed as follows:
169 utf8 data
171 Perl 5.6 added support for Unicode characters with code points >
172 255, and Perl 5.8 has full support for Unicode characters in hash
173 keys. Perl internally encodes strings with these characters using
174 utf8, and Storable serializes them as utf8. By default, if an
175 older version of Perl encounters a utf8 value it cannot represent,
176 it will "croak()". To change this behaviour so that Storable
177 deserializes utf8 encoded values as the string of bytes
178 (effectively dropping the is_utf8 flag) set $Storable::drop_utf8 to
179 some "TRUE" value. This is a form of data loss, because with
180 $drop_utf8 true, it becomes impossible to tell whether the original
181 data was the Unicode string, or a series of bytes that happen to be
182 valid utf8.
183 restricted hashes
185 Perl 5.8 adds support for restricted hashes, which have keys
186 restricted to a given set, and can have values locked to be read
187 only. By default, when Storable encounters a restricted hash on a
188 perl that doesn't support them, it will deserialize it as a normal
189 hash, silently discarding any placeholder keys and leaving the keys
190 and all values unlocked. To make Storable "croak()" instead, set
191 $Storable::downgrade_restricted to a "FALSE" value. To restore the
192 default set it back to some "TRUE" value.
193 The cperl PERL_PERTURB_KEYS_TOP hash strategy has a known problem
195 with restricted hashes.
196 huge objects
198 On 64bit systems some data structures may exceed the 2G (i.e.
199 I32_MAX) limit. On 32bit systems also strings between I32 and U32
200 (2G-4G). Since Storable 3.00 (not in perl5 core) we are able to
201 store and retrieve these objects, even if perl5 itself is not able
202 to handle them. These are strings longer then 4G, arrays with more
203 then 2G elements and hashes with more then 2G elements. cperl
204 forbids hashes with more than 2G elements, but this fail in cperl
205 then. perl5 itself at least until 5.26 allows it, but cannot
206 iterate over them. Note that creating those objects might cause
207 out of memory exceptions by the operating system before perl has a
208 chance to abort.
209 files from future versions of Storable
211 Earlier versions of Storable would immediately croak if they
212 encountered a file with a higher internal version number than the
213 reading Storable knew about. Internal version numbers are
214 increased each time new data types (such as restricted hashes) are
215 added to the vocabulary of the file format. This meant that a
216 newer Storable module had no way of writing a file readable by an
217 older Storable, even if the writer didn't store newer data types.
218 This version of Storable will defer croaking until it encounters a
220 data type in the file that it does not recognize. This means that
221 it will continue to read files generated by newer Storable modules
222 which are careful in what they write out, making it easier to
223 upgrade Storable modules in a mixed environment.
224 The old behaviour of immediate croaking can be re-instated by
226 setting $Storable::accept_future_minor to some "FALSE" value.
227 All these variables have no effect on a newer Perl which supports the
229 relevant feature.
230
232 Storable uses the "exception" paradigm, in that it does not try to
233 workaround failures: if something bad happens, an exception is
234 generated from the caller's perspective (see Carp and "croak()"). Use
235 eval {} to trap those exceptions.
236 When Storable croaks, it tries to report the error via the "logcroak()"
238 routine from the "Log::Agent" package, if it is available.
239 Normal errors are reported by having store() or retrieve() return
241 "undef". Such errors are usually I/O errors (or truncated stream
242 errors at retrieval).
243 When Storable throws the "Max. recursion depth with nested structures
245 exceeded" error we are already out of stack space. Unfortunately on
246 some earlier perl versions cleaning up a recursive data structure
247 recurses into the free calls, which will lead to stack overflows in the
248 cleanup. This data structure is not properly cleaned up then, it will
249 only be destroyed during global destruction.
250
252 Hooks
253 Any class may define hooks that will be called during the serialization
254 and deserialization process on objects that are instances of that
255 class. Those hooks can redefine the way serialization is performed
256 (and therefore, how the symmetrical deserialization should be
257 conducted).
258 Since we said earlier:
260 dclone(.) = thaw(freeze(.))
262 everything we say about hooks should also hold for deep cloning.
264 However, hooks get to know whether the operation is a mere
265 serialization, or a cloning.
266 Therefore, when serializing hooks are involved,
268 dclone(.) <> thaw(freeze(.))
270 Well, you could keep them in sync, but there's no guarantee it will
272 always hold on classes somebody else wrote. Besides, there is little
273 to gain in doing so: a serializing hook could keep only one attribute
274 of an object, which is probably not what should happen during a deep
275 cloning of that same object.
276 Here is the hooking interface:
278 "STORABLE_freeze" obj, cloning
280 The serializing hook, called on the object during serialization.
281 It can be inherited, or defined in the class itself, like any other
282 method.
283 Arguments: obj is the object to serialize, cloning is a flag
285 indicating whether we're in a dclone() or a regular serialization
286 via store() or freeze().
287 Returned value: A LIST "($serialized, $ref1, $ref2, ...)" where
289 $serialized is the serialized form to be used, and the optional
290 $ref1, $ref2, etc... are extra references that you wish to let the
291 Storable engine serialize.
292 At deserialization time, you will be given back the same LIST, but
294 all the extra references will be pointing into the deserialized
295 structure.
296 The first time the hook is hit in a serialization flow, you may
298 have it return an empty list. That will signal the Storable engine
299 to further discard that hook for this class and to therefore revert
300 to the default serialization of the underlying Perl data. The hook
301 will again be normally processed in the next serialization.
302 Unless you know better, serializing hook should always say:
304 sub STORABLE_freeze {
306 my ($self, $cloning) = @_;
307 return if $cloning; # Regular default serialization
308 ....
309 }
310 in order to keep reasonable dclone() semantics.
312 "STORABLE_thaw" obj, cloning, serialized, ...
314 The deserializing hook called on the object during deserialization.
315 But wait: if we're deserializing, there's no object yet... right?
316 Wrong: the Storable engine creates an empty one for you. If you
318 know Eiffel, you can view "STORABLE_thaw" as an alternate creation
319 routine.
320 This means the hook can be inherited like any other method, and
322 that obj is your blessed reference for this particular instance.
323 The other arguments should look familiar if you know
325 "STORABLE_freeze": cloning is true when we're part of a deep clone
326 operation, serialized is the serialized string you returned to the
327 engine in "STORABLE_freeze", and there may be an optional list of
328 references, in the same order you gave them at serialization time,
329 pointing to the deserialized objects (which have been processed
330 courtesy of the Storable engine).
331 When the Storable engine does not find any "STORABLE_thaw" hook
333 routine, it tries to load the class by requiring the package
334 dynamically (using the blessed package name), and then re-attempts
335 the lookup. If at that time the hook cannot be located, the engine
336 croaks. Note that this mechanism will fail if you define several
337 classes in the same file, but perlmod warned you.
338 It is up to you to use this information to populate obj the way you
340 want.
341 Returned value: none.
343 "STORABLE_attach" class, cloning, serialized
345 While "STORABLE_freeze" and "STORABLE_thaw" are useful for classes
346 where each instance is independent, this mechanism has difficulty
347 (or is incompatible) with objects that exist as common process-
348 level or system-level resources, such as singleton objects,
349 database pools, caches or memoized objects.
350 The alternative "STORABLE_attach" method provides a solution for
352 these shared objects. Instead of "STORABLE_freeze" -->
353 "STORABLE_thaw", you implement "STORABLE_freeze" -->
354 "STORABLE_attach" instead.
355 Arguments: class is the class we are attaching to, cloning is a
357 flag indicating whether we're in a dclone() or a regular de-
358 serialization via thaw(), and serialized is the stored string for
359 the resource object.
360 Because these resource objects are considered to be owned by the
362 entire process/system, and not the "property" of whatever is being
363 serialized, no references underneath the object should be included
364 in the serialized string. Thus, in any class that implements
365 "STORABLE_attach", the "STORABLE_freeze" method cannot return any
366 references, and "Storable" will throw an error if "STORABLE_freeze"
367 tries to return references.
368 All information required to "attach" back to the shared resource
370 object must be contained only in the "STORABLE_freeze" return
371 string. Otherwise, "STORABLE_freeze" behaves as normal for
372 "STORABLE_attach" classes.
373 Because "STORABLE_attach" is passed the class (rather than an
375 object), it also returns the object directly, rather than modifying
376 the passed object.
377 Returned value: object of type "class"
379 Predicates
381 Predicates are not exportable. They must be called by explicitly
382 prefixing them with the Storable package name.
383 "Storable::last_op_in_netorder"
385 The "Storable::last_op_in_netorder()" predicate will tell you
386 whether network order was used in the last store or retrieve
387 operation. If you don't know how to use this, just forget about
388 it.
389 "Storable::is_storing"
391 Returns true if within a store operation (via STORABLE_freeze
392 hook).
393 "Storable::is_retrieving"
395 Returns true if within a retrieve operation (via STORABLE_thaw
396 hook).
397 Recursion
399 With hooks comes the ability to recurse back to the Storable engine.
400 Indeed, hooks are regular Perl code, and Storable is convenient when it
401 comes to serializing and deserializing things, so why not use it to
402 handle the serialization string?
403 There are a few things you need to know, however:
405 · From Storable 3.05 to 3.13 we probed for the stack recursion limit
407 for references, arrays and hashes to a maximal depth of
408 ~1200-35000, otherwise we might fall into a stack-overflow. On
409 JSON::XS this limit is 512 btw. With references not immediately
410 referencing each other there's no such limit yet, so you might fall
411 into such a stack-overflow segfault.
412 This probing and the checks we performed have some limitations:
414 · the stack size at build time might be different at run time,
416 eg. the stack size may have been modified with ulimit(1). If
417 it's larger at run time Storable may fail the freeze() or
418 thaw() unnecessarily. If it's larger at build time Storable
419 may segmentation fault when processing a deep structure at run
420 time.
421 · the stack size might be different in a thread.
423 · array and hash recursion limits are checked separately against
425 the same recursion depth, a frozen structure with a large
426 sequence of nested arrays within many nested hashes may exhaust
427 the processor stack without triggering Storable's recursion
428 protection.
429 So these now have simple defaults rather than probing at build-
431 time.
432 You can control the maximum array and hash recursion depths by
434 modifying $Storable::recursion_limit and
435 $Storable::recursion_limit_hash respectively. Either can be set to
436 "-1" to prevent any depth checks, though this isn't recommended.
437 · You can create endless loops if the things you serialize via
439 freeze() (for instance) point back to the object we're trying to
440 serialize in the hook.
441 · Shared references among objects will not stay shared: if we're
443 serializing the list of object [A, C] where both object A and C
444 refer to the SAME object B, and if there is a serializing hook in A
445 that says freeze(B), then when deserializing, we'll get [A', C']
446 where A' refers to B', but C' refers to D, a deep clone of B'. The
447 topology was not preserved.
448 · The maximal stack recursion limit for your system is returned by
450 "stack_depth()" and "stack_depth_hash()". The hash limit is usually
451 half the size of the array and ref limit, as the Perl hash API is
452 not optimal.
453 That's why "STORABLE_freeze" lets you provide a list of references to
455 serialize. The engine guarantees that those will be serialized in the
456 same context as the other objects, and therefore that shared objects
457 will stay shared.
458 In the above [A, C] example, the "STORABLE_freeze" hook could return:
460 ("something", $self->{B})
462 and the B part would be serialized by the engine. In "STORABLE_thaw",
464 you would get back the reference to the B' object, deserialized for
465 you.
466 Therefore, recursion should normally be avoided, but is nonetheless
468 supported.
469 Deep Cloning
471 There is a Clone module available on CPAN which implements deep cloning
472 natively, i.e. without freezing to memory and thawing the result. It
473 is aimed to replace Storable's dclone() some day. However, it does not
474 currently support Storable hooks to redefine the way deep cloning is
475 performed.
476
478 Yes, there's a lot of that :-) But more precisely, in UNIX systems
479 there's a utility called "file", which recognizes data files based on
480 their contents (usually their first few bytes). For this to work, a
481 certain file called magic needs to taught about the signature of the
482 data. Where that configuration file lives depends on the UNIX flavour;
483 often it's something like /usr/share/misc/magic or /etc/magic. Your
484 system administrator needs to do the updating of the magic file. The
485 necessary signature information is output to STDOUT by invoking
486 Storable::show_file_magic(). Note that the GNU implementation of the
487 "file" utility, version 3.38 or later, is expected to contain support
488 for recognising Storable files out-of-the-box, in addition to other
489 kinds of Perl files.
490 You can also use the following functions to extract the file header
492 information from Storable images:
493 $info = Storable::file_magic( $filename )
495 If the given file is a Storable image return a hash describing it.
496 If the file is readable, but not a Storable image return "undef".
497 If the file does not exist or is unreadable then croak.
498 The hash returned has the following elements:
500 "version"
502 This returns the file format version. It is a string like
503 "2.7".
504 Note that this version number is not the same as the version
506 number of the Storable module itself. For instance Storable
507 v0.7 create files in format v2.0 and Storable v2.15 create
508 files in format v2.7. The file format version number only
509 increment when additional features that would confuse older
510 versions of the module are added.
511 Files older than v2.0 will have the one of the version numbers
513 "-1", "0" or "1". No minor number was used at that time.
514 "version_nv"
516 This returns the file format version as number. It is a string
517 like "2.007". This value is suitable for numeric comparisons.
518 The constant function "Storable::BIN_VERSION_NV" returns a
520 comparable number that represents the highest file version
521 number that this version of Storable fully supports (but see
522 discussion of $Storable::accept_future_minor above). The
523 constant "Storable::BIN_WRITE_VERSION_NV" function returns what
524 file version is written and might be less than
525 "Storable::BIN_VERSION_NV" in some configurations.
526 "major", "minor"
528 This also returns the file format version. If the version is
529 "2.7" then major would be 2 and minor would be 7. The minor
530 element is missing for when major is less than 2.
531 "hdrsize"
533 The is the number of bytes that the Storable header occupies.
534 "netorder"
536 This is TRUE if the image store data in network order. This
537 means that it was created with nstore() or similar.
538 "byteorder"
540 This is only present when "netorder" is FALSE. It is the
541 $Config{byteorder} string of the perl that created this image.
542 It is a string like "1234" (32 bit little endian) or "87654321"
543 (64 bit big endian). This must match the current perl for the
544 image to be readable by Storable.
545 "intsize", "longsize", "ptrsize", "nvsize"
547 These are only present when "netorder" is FALSE. These are the
548 sizes of various C datatypes of the perl that created this
549 image. These must match the current perl for the image to be
550 readable by Storable.
551 The "nvsize" element is only present for file format v2.2 and
553 higher.
554 "file"
556 The name of the file.
557 $info = Storable::read_magic( $buffer )
559 $info = Storable::read_magic( $buffer, $must_be_file )
560 The $buffer should be a Storable image or the first few bytes of
561 it. If $buffer starts with a Storable header, then a hash
562 describing the image is returned, otherwise "undef" is returned.
563 The hash has the same structure as the one returned by
565 Storable::file_magic(). The "file" element is true if the image is
566 a file image.
567 If the $must_be_file argument is provided and is TRUE, then return
569 "undef" unless the image looks like it belongs to a file dump.
570 The maximum size of a Storable header is currently 21 bytes. If
572 the provided $buffer is only the first part of a Storable image it
573 should at least be this long to ensure that read_magic() will
574 recognize it as such.
575
577 Here are some code samples showing a possible usage of Storable:
578 use Storable qw(store retrieve freeze thaw dclone);
580 %color = ('Blue' => 0.1, 'Red' => 0.8, 'Black' => 0, 'White' => 1);
582 store(\%color, 'mycolors') or die "Can't store %a in mycolors!\n";
584 $colref = retrieve('mycolors');
586 die "Unable to retrieve from mycolors!\n" unless defined $colref;
587 printf "Blue is still %lf\n", $colref->{'Blue'};
588 $colref2 = dclone(\%color);
590 $str = freeze(\%color);
592 printf "Serialization of %%color is %d bytes long.\n", length($str);
593 $colref3 = thaw($str);
594 which prints (on my machine):
596 Blue is still 0.100000
598 Serialization of %color is 102 bytes long.
599 Serialization of CODE references and deserialization in a safe
601 compartment:
602 use Storable qw(freeze thaw);
604 use Safe;
605 use strict;
606 my $safe = new Safe;
607 # because of opcodes used in "use strict":
608 $safe->permit(qw(:default require));
609 local $Storable::Deparse = 1;
610 local $Storable::Eval = sub { $safe->reval($_[0]) };
611 my $serialized = freeze(sub { 42 });
612 my $code = thaw($serialized);
613 $code->() == 42;
614
616 Do not accept Storable documents from untrusted sources!
617 Some features of Storable can lead to security vulnerabilities if you
619 accept Storable documents from untrusted sources with the default
620 flags. Most obviously, the optional (off by default) CODE reference
621 serialization feature allows transfer of code to the deserializing
622 process. Furthermore, any serialized object will cause Storable to
623 helpfully load the module corresponding to the class of the object in
624 the deserializing module. For manipulated module names, this can load
625 almost arbitrary code. Finally, the deserialized object's destructors
626 will be invoked when the objects get destroyed in the deserializing
627 process. Maliciously crafted Storable documents may put such objects in
628 the value of a hash key that is overridden by another key/value pair in
629 the same hash, thus causing immediate destructor execution.
630 To disable blessing objects while thawing/retrieving remove the flag
632 "BLESS_OK" = 2 from $Storable::flags or set the 2nd argument for
633 thaw/retrieve to 0.
634 To disable tieing data while thawing/retrieving remove the flag
636 "TIE_OK" = 4 from $Storable::flags or set the 2nd argument for
637 thaw/retrieve to 0.
638 With the default setting of $Storable::flags = 6, creating or
640 destroying random objects, even renamed objects can be controlled by an
641 attacker. See CVE-2015-1592 and its metasploit module.
642 If your application requires accepting data from untrusted sources, you
644 are best off with a less powerful and more-likely safe serialization
645 format and implementation. If your data is sufficiently simple,
646 Cpanel::JSON::XS, Data::MessagePack or Serial are the best choices and
647 offers maximum interoperability, but note that Serial is unsafe by
648 default.
649
651 If you're using references as keys within your hash tables, you're
652 bound to be disappointed when retrieving your data. Indeed, Perl
653 stringifies references used as hash table keys. If you later wish to
654 access the items via another reference stringification (i.e. using the
655 same reference that was used for the key originally to record the value
656 into the hash table), it will work because both references stringify to
657 the same string.
658 It won't work across a sequence of "store" and "retrieve" operations,
660 however, because the addresses in the retrieved objects, which are part
661 of the stringified references, will probably differ from the original
662 addresses. The topology of your structure is preserved, but not hidden
663 semantics like those.
664 On platforms where it matters, be sure to call "binmode()" on the
666 descriptors that you pass to Storable functions.
667 Storing data canonically that contains large hashes can be
669 significantly slower than storing the same data normally, as temporary
670 arrays to hold the keys for each hash have to be allocated, populated,
671 sorted and freed. Some tests have shown a halving of the speed of
672 storing -- the exact penalty will depend on the complexity of your
673 data. There is no slowdown on retrieval.
674
676 Storable now has experimental support for storing regular expressions,
677 but there are significant limitations:
678 · perl 5.8 or later is required.
680 · regular expressions with code blocks, ie "/(?{ ... })/" or "/(??{
682 ... })/" will throw an exception when thawed.
683 · regular expression syntax and flags have changed over the history
685 of perl, so a regular expression that you freeze in one version of
686 perl may fail to thaw or behave differently in another version of
687 perl.
688 · depending on the version of perl, regular expressions can change in
690 behaviour depending on the context, but later perls will bake that
691 behaviour into the regexp.
692 Storable will throw an exception if a frozen regular expression cannot
694 be thawed.
695
697 You can't store GLOB, FORMLINE, etc.... If you can define semantics for
698 those operations, feel free to enhance Storable so that it can deal
699 with them.
700 The store functions will "croak" if they run into such references
702 unless you set $Storable::forgive_me to some "TRUE" value. In that
703 case, the fatal message is converted to a warning and some meaningless
704 string is stored instead.
705 Setting $Storable::canonical may not yield frozen strings that compare
707 equal due to possible stringification of numbers. When the string
708 version of a scalar exists, it is the form stored; therefore, if you
709 happen to use your numbers as strings between two freezing operations
710 on the same data structures, you will get different results.
711 When storing doubles in network order, their value is stored as text.
713 However, you should also not expect non-numeric floating-point values
714 such as infinity and "not a number" to pass successfully through a
715 nstore()/retrieve() pair.
716 As Storable neither knows nor cares about character sets (although it
718 does know that characters may be more than eight bits wide), any
719 difference in the interpretation of character codes between a host and
720 a target system is your problem. In particular, if host and target use
721 different code points to represent the characters used in the text
722 representation of floating-point numbers, you will not be able be able
723 to exchange floating-point data, even with nstore().
724 "Storable::drop_utf8" is a blunt tool. There is no facility either to
726 return all strings as utf8 sequences, or to attempt to convert utf8
727 data back to 8 bit and "croak()" if the conversion fails.
728 Prior to Storable 2.01, no distinction was made between signed and
730 unsigned integers on storing. By default Storable prefers to store a
731 scalars string representation (if it has one) so this would only cause
732 problems when storing large unsigned integers that had never been
733 converted to string or floating point. In other words values that had
734 been generated by integer operations such as logic ops and then not
735 used in any string or arithmetic context before storing.
736 64 bit data in perl 5.6.0 and 5.6.1
738 This section only applies to you if you have existing data written out
739 by Storable 2.02 or earlier on perl 5.6.0 or 5.6.1 on Unix or Linux
740 which has been configured with 64 bit integer support (not the default)
741 If you got a precompiled perl, rather than running Configure to build
742 your own perl from source, then it almost certainly does not affect
743 you, and you can stop reading now (unless you're curious). If you're
744 using perl on Windows it does not affect you.
745 Storable writes a file header which contains the sizes of various C
747 language types for the C compiler that built Storable (when not writing
748 in network order), and will refuse to load files written by a Storable
749 not on the same (or compatible) architecture. This check and a check
750 on machine byteorder is needed because the size of various fields in
751 the file are given by the sizes of the C language types, and so files
752 written on different architectures are incompatible. This is done for
753 increased speed. (When writing in network order, all fields are
754 written out as standard lengths, which allows full interworking, but
755 takes longer to read and write)
756 Perl 5.6.x introduced the ability to optional configure the perl
758 interpreter to use C's "long long" type to allow scalars to store 64
759 bit integers on 32 bit systems. However, due to the way the Perl
760 configuration system generated the C configuration files on non-Windows
761 platforms, and the way Storable generates its header, nothing in the
762 Storable file header reflected whether the perl writing was using 32 or
763 64 bit integers, despite the fact that Storable was storing some data
764 differently in the file. Hence Storable running on perl with 64 bit
765 integers will read the header from a file written by a 32 bit perl, not
766 realise that the data is actually in a subtly incompatible format, and
767 then go horribly wrong (possibly crashing) if it encountered a stored
768 integer. This is a design failure.
769 Storable has now been changed to write out and read in a file header
771 with information about the size of integers. It's impossible to detect
772 whether an old file being read in was written with 32 or 64 bit
773 integers (they have the same header) so it's impossible to
774 automatically switch to a correct backwards compatibility mode. Hence
775 this Storable defaults to the new, correct behaviour.
776 What this means is that if you have data written by Storable 1.x
778 running on perl 5.6.0 or 5.6.1 configured with 64 bit integers on Unix
779 or Linux then by default this Storable will refuse to read it, giving
780 the error Byte order is not compatible. If you have such data then you
781 should set $Storable::interwork_56_64bit to a true value to make this
782 Storable read and write files with the old header. You should also
783 migrate your data, or any older perl you are communicating with, to
784 this current version of Storable.
785 If you don't have data written with specific configuration of perl
787 described above, then you do not and should not do anything. Don't set
788 the flag - not only will Storable on an identically configured perl
789 refuse to load them, but Storable a differently configured perl will
790 load them believing them to be correct for it, and then may well fail
791 or crash part way through reading them.
792
794 Thank you to (in chronological order):
795 Jarkko Hietaniemi <jhi@iki.fi>
797 Ulrich Pfeifer <pfeifer@charly.informatik.uni-dortmund.de>
798 Benjamin A. Holzman <bholzman@earthlink.net>
799 Andrew Ford <A.Ford@ford-mason.co.uk>
800 Gisle Aas <gisle@aas.no>
801 Jeff Gresham <gresham_jeffrey@jpmorgan.com>
802 Murray Nesbitt <murray@activestate.com>
803 Marc Lehmann <pcg@opengroup.org>
804 Justin Banks <justinb@wamnet.com>
805 Jarkko Hietaniemi <jhi@iki.fi> (AGAIN, as perl 5.7.0 Pumpkin!)
806 Salvador Ortiz Garcia <sog@msg.com.mx>
807 Dominic Dunlop <domo@computer.org>
808 Erik Haugan <erik@solbors.no>
809 Benjamin A. Holzman <ben.holzman@grantstreet.com>
810 Reini Urban <rurban@cpan.org>
811 Todd Rinaldo <toddr@cpanel.net>
812 Aaron Crane <arc@cpan.org>
813 for their bug reports, suggestions and contributions.
815 Benjamin Holzman contributed the tied variable support, Andrew Ford
817 contributed the canonical order for hashes, and Gisle Aas fixed a few
818 misunderstandings of mine regarding the perl internals, and optimized
819 the emission of "tags" in the output streams by simply counting the
820 objects instead of tagging them (leading to a binary incompatibility
821 for the Storable image starting at version 0.6--older images are, of
822 course, still properly understood). Murray Nesbitt made Storable
823 thread-safe. Marc Lehmann added overloading and references to tied
824 items support. Benjamin Holzman added a performance improvement for
825 overloaded classes; thanks to Grant Street Group for footing the bill.
826 Reini Urban took over maintainance from p5p, and added security fixes
827 and huge object support.
828
830 Storable was written by Raphael Manfredi <Raphael_Manfredi@pobox.com>
831 Maintenance is now done by cperl <http://perl11.org/cperl>
832 Please e-mail us with problems, bug fixes, comments and complaints,
834 although if you have compliments you should send them to Raphael.
835 Please don't e-mail Raphael with problems, as he no longer works on
836 Storable, and your message will be delayed while he forwards it to us.
837
839 Clone.
840perl v5.30.1 2019-11-25 Storable(3)