/[svn]/libgig/trunk/src/Serialization.h
ViewVC logotype

Contents of /libgig/trunk/src/Serialization.h

Parent Directory Parent Directory | Revision Log Revision Log


Revision 3198 - (show annotations) (download) (as text)
Sun May 21 12:46:05 2017 UTC (6 years, 10 months ago) by schoenebeck
File MIME type: text/x-c++hdr
File size: 51863 byte(s)
* RIFF/DLS/gig/Serialization: Exception classes now have a variadic
  constructor which allows to add textual format specifiers like
  with printf().
* gig.cpp: On unknown leverage controller exception: show precise unknown
  leverage controller number found.

1 /***************************************************************************
2 * *
3 * Copyright (C) 2017 Christian Schoenebeck *
4 * <cuse@users.sourceforge.net> *
5 * *
6 * This library is part of libgig. *
7 * *
8 * This library is free software; you can redistribute it and/or modify *
9 * it under the terms of the GNU General Public License as published by *
10 * the Free Software Foundation; either version 2 of the License, or *
11 * (at your option) any later version. *
12 * *
13 * This library is distributed in the hope that it will be useful, *
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of *
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
16 * GNU General Public License for more details. *
17 * *
18 * You should have received a copy of the GNU General Public License *
19 * along with this library; if not, write to the Free Software *
20 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, *
21 * MA 02111-1307 USA *
22 ***************************************************************************/
23
24 #ifndef LIBGIG_SERIALIZATION_H
25 #define LIBGIG_SERIALIZATION_H
26
27 #ifdef HAVE_CONFIG_H
28 # include <config.h>
29 #endif
30
31 #include <stdint.h>
32 #include <stdio.h>
33 #include <typeinfo>
34 #include <string>
35 #include <vector>
36 #include <map>
37 #include <time.h>
38 #include <stdarg.h>
39
40 #ifndef __has_extension
41 # define __has_extension(x) 0
42 #endif
43
44 #ifndef HAS_BUILTIN_TYPE_TRAITS
45 # if __cplusplus >= 201103L
46 # define HAS_BUILTIN_TYPE_TRAITS 1
47 # elif ( __has_extension(is_class) && __has_extension(is_enum) )
48 # define HAS_BUILTIN_TYPE_TRAITS 1
49 # elif ( __GNUC__ > 4 || ( __GNUC__ == 4 && __GNUC_MINOR__ >= 3 ) )
50 # define HAS_BUILTIN_TYPE_TRAITS 1
51 # elif _MSC_VER >= 1400 /* MS Visual C++ 8.0 (Visual Studio 2005) */
52 # define HAS_BUILTIN_TYPE_TRAITS 1
53 # elif __INTEL_COMPILER >= 1100
54 # define HAS_BUILTIN_TYPE_TRAITS 1
55 # else
56 # define HAS_BUILTIN_TYPE_TRAITS 0
57 # endif
58 #endif
59
60 #if !HAS_BUILTIN_TYPE_TRAITS
61 # include <tr1/type_traits>
62 # define LIBGIG_IS_CLASS(type) std::tr1::__is_union_or_class<type>::value //NOTE: without compiler support we cannot distinguish union from class
63 #else
64 # define LIBGIG_IS_CLASS(type) __is_class(type)
65 #endif
66
67 /** @brief Serialization / deserialization framework.
68 *
69 * See class Archive as starting point for how to implement serialization and
70 * deserialization with your application.
71 *
72 * The classes in this namespace allow to serialize and deserialize native
73 * C++ objects in a portable, easy and flexible way. Serialization is a
74 * technique that allows to transform the current state and data of native
75 * (in this case C++) objects into a data stream (including all other objects
76 * the "serialized" objects relate to); the data stream may then be sent over
77 * "wire" (for example via network connection to another computer, which might
78 * also have a different OS, CPU architecture, native memory word size and
79 * endian type); and finally the data stream would be "deserialized" on that
80 * receiver side, that is transformed again to modify all objects and data
81 * structures on receiver side to resemble the objects' state and data as it
82 * was originally on sender side.
83 *
84 * In contrast to many other already existing serialization frameworks, this
85 * implementation has a strong focus on robustness regarding long-term changes
86 * to the serialized C++ classes of the serialized objects. So even if sender
87 * and receiver are using different versions of their serialized/deserialized
88 * C++ classes, structures and data types (thus having different data structure
89 * layout to a certain extent), this framework aims trying to automatically
90 * adapt its serialization and deserialization process in that case so that
91 * the deserialized objects on receiver side would still reflect the overall
92 * expected states and overall data as intended by the sender. For being able to
93 * do so, this framework stores all kind of additional information about each
94 * serialized object and each data structure member (for example name of each
95 * data structure member, but also the offset of each member within its
96 * containing data structure, precise data types, and more).
97 *
98 * Like most other serialization frameworks, this frameworks does not require a
99 * tree-structured layout of the serialized data structures. So it automatically
100 * handles also cyclic dependencies between serialized data structures
101 * correctly, without i.e. causing endless recursion or redundancy.
102 *
103 * Additionally this framework also allows partial deserialization. Which means
104 * the receiver side may for example decide that it wants to restrict
105 * deserialization so that it would only modify certain objects or certain
106 * members by the deserialization process, leaving all other ones untouched.
107 * So this partial deserialization technique for example allows to implement
108 * flexible preset features for applications in a powerful and easy way.
109 */
110 namespace Serialization {
111
112 // just symbol prototyping
113 class DataType;
114 class Object;
115 class Member;
116 class Archive;
117 class ObjectPool;
118 class Exception;
119
120 typedef std::string String;
121
122 /** @brief Raw data stream of serialized C++ objects.
123 *
124 * This data type is used for the data stream as a result of serializing
125 * your C++ objects with Archive::serialize(), and for native raw data
126 * representation of individual serialized C/C++ objects, members and variables.
127 *
128 * @see Archive::rawData(), Object::rawData()
129 */
130 typedef std::vector<uint8_t> RawData;
131
132 /** @brief Abstract identifier for serialized C++ objects.
133 *
134 * This data type is used for identifying serialized C++ objects and members
135 * of your C++ objects. It is important to know that such an ID might not
136 * necessarily be unique. For example the ID of one C++ object might often
137 * be identical to the ID of the first member of that particular C++ object.
138 * That's why there is additionally the concept of an UID in this framework.
139 *
140 * @see UID
141 */
142 typedef void* ID;
143
144 /** @brief Version number data type.
145 *
146 * This data type is used for maintaining version number information of
147 * your C++ class implementations.
148 *
149 * @see Archive::setVersion() and Archive::setMinVersion()
150 */
151 typedef uint32_t Version;
152
153 /** @brief To which time zone a certain timing information relates to.
154 *
155 * The constants in this enum type are used to define to which precise time
156 * zone a time stamp relates to.
157 */
158 enum time_base_t {
159 LOCAL_TIME, ///< The time stamp relates to the machine's local time zone. Request a time stamp in local time if you want to present that time stamp to the end user.
160 UTC_TIME ///< The time stamp relates to "Greenwhich Mean Time" zone, also known as "Coordinated Universal Time". Request time stamp with UTC if you want to compare that time stamp with other time stamps.
161 };
162
163 /** @brief Check whether data is a C/C++ @c enum type.
164 *
165 * Returns true if the supplied C++ variable or object is of a C/C++ @c enum
166 * type.
167 *
168 * @param data - the variable or object whose data type shall be checked
169 */
170 template<typename T>
171 bool IsEnum(const T& data) {
172 #if !HAS_BUILTIN_TYPE_TRAITS
173 return std::tr1::is_enum<T>::value;
174 #else
175 return __is_enum(T);
176 #endif
177 }
178
179 /** @brief Check whether data is a C++ @c union type.
180 *
181 * Returns true if the supplied C++ variable or object is of a C/C++ @c union
182 * type. Note that the result of this function is only reliable if the C++
183 * compiler you are using has support for built-in type traits. If your C++
184 * compiler does not have built-in type traits support, then this function
185 * will simply return @c false on all your calls.
186 *
187 * @param data - the variable or object whose data type shall be checked
188 */
189 template<typename T>
190 bool IsUnion(const T& data) {
191 #if !HAS_BUILTIN_TYPE_TRAITS
192 return false; // without compiler support we cannot distinguish union from class
193 #else
194 return __is_union(T);
195 #endif
196 }
197
198 /** @brief Check whether data is a C/C++ @c struct or C++ @c class type.
199 *
200 * Returns true if the supplied C++ variable or object is of C/C++ @c struct
201 * or C++ @c class type. Note that if you are using a C++ compiler which
202 * does have built-in type traits support, then this function will also
203 * return @c true on C/C++ @c union types.
204 *
205 * @param data - the variable or object whose data type shall be checked
206 */
207 template<typename T>
208 bool IsClass(const T& data) {
209 #if !HAS_BUILTIN_TYPE_TRAITS
210 return std::tr1::__is_union_or_class<T>::value; // without compiler support we cannot distinguish union from class
211 #else
212 return __is_class(T);
213 #endif
214 }
215
216 /*template<typename T>
217 bool IsTrivial(T data) {
218 return __is_trivial(T);
219 }*/
220
221 /*template<typename T>
222 bool IsPOD(T data) {
223 return __is_pod(T);
224 }*/
225
226 /** @brief Unique identifier referring to one specific native C++ object, member, fundamental variable, or any other native C++ data.
227 *
228 * Reflects a unique identifier for one specific serialized C++ data, i.e.
229 * C++ class instance, C/C++ struct instance, member, primitive pointer,
230 * fundamental variables, or any other native C/C++ data originally being
231 * serialized.
232 *
233 * A unique identifier is composed of an id (an identifier which is not
234 * necessarily unique) and a size. Since the underlying ID is derived from
235 * the original C++ object's memory location, such an ID is not sufficient
236 * to distinguish a particular C++ object from the first member of that C++
237 * object, since both typically share the same memory address. So
238 * additionally the memory size of the respective object or member is
239 * bundled with UID objects to make them unique and distinguishable.
240 */
241 class UID {
242 public:
243 ID id; ///< Abstract non-unique ID of the object or member in question.
244 size_t size; ///< Memory size of the object or member in question.
245
246 bool isValid() const;
247 operator bool() const { return isValid(); } ///< Same as calling isValid().
248 //bool operator()() const { return isValid(); }
249 bool operator==(const UID& other) const { return id == other.id && size == other.size; }
250 bool operator!=(const UID& other) const { return id != other.id || size != other.size; }
251 bool operator<(const UID& other) const { return id < other.id || (id == other.id && size < other.size); }
252 bool operator>(const UID& other) const { return id > other.id || (id == other.id && size > other.size); }
253
254 /** @brief Create an unique indentifier for a native C++ object/member/variable.
255 *
256 * Creates and returns an unique identifier for the passed native C++
257 * object, object member or variable. For the same C++ object/member/variable
258 * this function will always return the same UID. For all other ones,
259 * this function is guaranteed to return a different UID.
260 */
261 template<typename T>
262 static UID from(const T& obj) {
263 return Resolver<T>::resolve(obj);
264 }
265
266 protected:
267 // UID resolver for non-pointer types
268 template<typename T>
269 struct Resolver {
270 static UID resolve(const T& obj) {
271 const UID uid = { (ID) &obj, sizeof(obj) };
272 return uid;
273 }
274 };
275
276 // UID resolver for pointer types (of 1st degree)
277 template<typename T>
278 struct Resolver<T*> {
279 static UID resolve(const T* const & obj) {
280 const UID uid = { (ID) obj, sizeof(*obj) };
281 return uid;
282 }
283 };
284 };
285
286 /**
287 * Reflects an invalid UID and behaves similar to NULL as invalid value for
288 * pointer types. All UID objects are first initialized with this value,
289 * and it essentially an all zero object.
290 */
291 extern const UID NO_UID;
292
293 /** @brief Chain of UIDs.
294 *
295 * This data type is used for native C++ pointers. The first member of the
296 * UID chain is the unique identifier of the C++ pointer itself, then the
297 * following UIDs are the respective objects or variables the pointer is
298 * pointing to. The size (the amount of elements) of the UIDChain depends
299 * solely on the degree of the pointer type. For example the following C/C++
300 * pointer:
301 * @code
302 * int* pNumber;
303 * @endcode
304 * is an integer pointer of first degree. Such a pointer would have a
305 * UIDChain with 2 members: the first element would be the UID of the
306 * pointer itself, the second element of the chain would be the integer data
307 * that pointer is pointing to. In the following example:
308 * @code
309 * bool*** pppSomeFlag;
310 * @endcode
311 * That boolean pointer would be of third degree, and thus its UIDChain
312 * would have a size of 4 (elements).
313 *
314 * Accordingly a non pointer type like:
315 * @code
316 * float f;
317 * @endcode
318 * would yield in a UIDChain of size 1.
319 *
320 * Since however this serialization framework currently only supports
321 * pointers of first degree yet, all UIDChains are currently either of
322 * size 1 or 2, which might change in future though.
323 */
324 typedef std::vector<UID> UIDChain;
325
326 // prototyping of private internal friend functions
327 static String _encodePrimitiveValue(const Object& obj);
328 static DataType _popDataTypeBlob(const char*& p, const char* end);
329 static Member _popMemberBlob(const char*& p, const char* end);
330 static Object _popObjectBlob(const char*& p, const char* end);
331 static void _popPrimitiveValue(const char*& p, const char* end, Object& obj);
332 static String _primitiveObjectValueToString(const Object& obj);
333 // |
334 template<typename T>
335 static T _primitiveObjectValueToNumber(const Object& obj);
336
337 /** @brief Abstract reflection of a native C++ data type.
338 *
339 * Provides detailed information about a serialized C++ data type, whether
340 * it is a fundamental C/C++ data type (like @c int, @c float, @c char,
341 * etc.) or custom defined data types like a C++ @c class, C/C++ @c struct,
342 * @c enum, as well as other features of the respective data type like its
343 * native memory size and more.
344 *
345 * All informations provided by this class are retrieved from the
346 * respective individual C++ objects, their members and other data when
347 * they are serialized, and all those information are stored with the
348 * serialized archive and its resulting data stream. Due to the availability
349 * of these extensive data type information within serialized archives, this
350 * framework is capable to use them in order to adapt its deserialization
351 * process upon subsequent changes to your individual C++ classes.
352 */
353 class DataType {
354 public:
355 DataType();
356 size_t size() const { return m_size; } ///< Returns native memory size of the respective C++ object or variable.
357 bool isValid() const;
358 bool isPointer() const;
359 bool isClass() const;
360 bool isPrimitive() const;
361 bool isInteger() const;
362 bool isReal() const;
363 bool isBool() const;
364 bool isEnum() const;
365 bool isSigned() const;
366 operator bool() const { return isValid(); } ///< Same as calling isValid().
367 //bool operator()() const { return isValid(); }
368 bool operator==(const DataType& other) const;
369 bool operator!=(const DataType& other) const;
370 bool operator<(const DataType& other) const;
371 bool operator>(const DataType& other) const;
372 String asLongDescr() const;
373 String baseTypeName() const;
374 String customTypeName(bool demangle = false) const;
375
376 /** @brief Construct a DataType object for the given native C++ data.
377 *
378 * Use this function to create corresponding DataType objects for
379 * native C/C++ objects, members and variables.
380 *
381 * @param data - native C/C++ object/member/variable a DataType object
382 * shall be created for
383 * @returns corresponding DataType object for the supplied native C/C++
384 * object/member/variable
385 */
386 template<typename T>
387 static DataType dataTypeOf(const T& data) {
388 return Resolver<T>::resolve(data);
389 }
390
391 protected:
392 DataType(bool isPointer, int size, String baseType, String customType = "");
393
394 template<typename T, bool T_isPointer>
395 struct ResolverBase {
396 static DataType resolve(const T& data) {
397 const std::type_info& type = typeid(data);
398 const int sz = sizeof(data);
399
400 // for primitive types we are using our own type names instead of
401 // using std:::type_info::name(), because the precise output of the
402 // latter may vary between compilers
403 if (type == typeid(int8_t)) return DataType(T_isPointer, sz, "int8");
404 if (type == typeid(uint8_t)) return DataType(T_isPointer, sz, "uint8");
405 if (type == typeid(int16_t)) return DataType(T_isPointer, sz, "int16");
406 if (type == typeid(uint16_t)) return DataType(T_isPointer, sz, "uint16");
407 if (type == typeid(int32_t)) return DataType(T_isPointer, sz, "int32");
408 if (type == typeid(uint32_t)) return DataType(T_isPointer, sz, "uint32");
409 if (type == typeid(int64_t)) return DataType(T_isPointer, sz, "int64");
410 if (type == typeid(uint64_t)) return DataType(T_isPointer, sz, "uint64");
411 if (type == typeid(bool)) return DataType(T_isPointer, sz, "bool");
412 if (type == typeid(float)) return DataType(T_isPointer, sz, "real32");
413 if (type == typeid(double)) return DataType(T_isPointer, sz, "real64");
414
415 if (IsEnum(data)) return DataType(T_isPointer, sz, "enum", rawCppTypeNameOf(data));
416 if (IsUnion(data)) return DataType(T_isPointer, sz, "union", rawCppTypeNameOf(data));
417 if (IsClass(data)) return DataType(T_isPointer, sz, "class", rawCppTypeNameOf(data));
418
419 return DataType();
420 }
421 };
422
423 // DataType resolver for non-pointer types
424 template<typename T>
425 struct Resolver : ResolverBase<T,false> {
426 static DataType resolve(const T& data) {
427 return ResolverBase<T,false>::resolve(data);
428 }
429 };
430
431 // DataType resolver for pointer types (of 1st degree)
432 template<typename T>
433 struct Resolver<T*> : ResolverBase<T,true> {
434 static DataType resolve(const T*& data) {
435 return ResolverBase<T,true>::resolve(*data);
436 }
437 };
438
439 template<typename T>
440 static String rawCppTypeNameOf(const T& data) {
441 #if defined _MSC_VER // Microsoft compiler ...
442 # warning type_info::raw_name() demangling has not been tested yet with Microsoft compiler! Feedback appreciated!
443 String name = typeid(data).raw_name(); //NOTE: I haven't checked yet what MSC actually outputs here exactly
444 #else // i.e. especially GCC and clang ...
445 String name = typeid(data).name();
446 #endif
447 //while (!name.empty() && name[0] >= 0 && name[0] <= 9)
448 // name = name.substr(1);
449 return name;
450 }
451
452 private:
453 String m_baseTypeName;
454 String m_customTypeName;
455 int m_size;
456 bool m_isPointer;
457
458 friend DataType _popDataTypeBlob(const char*& p, const char* end);
459 friend class Archive;
460 };
461
462 /** @brief Abstract reflection of a native C++ class/struct's member variable.
463 *
464 * Provides detailed information about a specific C++ member variable of
465 * serialized C++ object, like its C++ data type, offset of this member
466 * within its containing data structure/class, its C++ member variable name
467 * and more.
468 *
469 * Consider you defined the following user defined C/C++ @c struct type in
470 * your application:
471 * @code
472 * struct Foo {
473 * int a;
474 * bool b;
475 * double someValue;
476 * };
477 * @endcode
478 * Then @c a, @c b and @c someValue are "members" of @c struct @c Foo for
479 * instance. So that @c struct would have 3 members in the latter example.
480 *
481 * @see Object::members()
482 */
483 class Member {
484 public:
485 Member();
486 UID uid() const;
487 String name() const;
488 size_t offset() const;
489 const DataType& type() const;
490 bool isValid() const;
491 operator bool() const { return isValid(); } ///< Same as calling isValid().
492 //bool operator()() const { return isValid(); }
493 bool operator==(const Member& other) const;
494 bool operator!=(const Member& other) const;
495 bool operator<(const Member& other) const;
496 bool operator>(const Member& other) const;
497
498 protected:
499 Member(String name, UID uid, size_t offset, DataType type);
500 friend class Archive;
501
502 private:
503 UID m_uid;
504 size_t m_offset;
505 String m_name;
506 DataType m_type;
507
508 friend Member _popMemberBlob(const char*& p, const char* end);
509 };
510
511 /** @brief Abstract reflection of some native serialized C/C++ data.
512 *
513 * When your native C++ objects are serialized, all native data is
514 * translated and reflected by such an Object reflection. So each instance
515 * of your serialized native C++ class objects become available as an
516 * Object, but also each member variable of your C++ objects is translated
517 * into an Object, and any other native C/C++ data. So essentially every
518 * native data is turned into its own Object and accessible by this API.
519 *
520 * For each one of those Object reflections, this class provides detailed
521 * information about their native origin. For example if an Object
522 * represents a native C++ class instante, then it provides access to its
523 * C++ class/struct name, to its C++ member variables, its native memory
524 * size and much more.
525 *
526 * Even though this framework allows you to adjust abstract Object instances
527 * to a certain extent, most of the methods of this Object class are
528 * read-only though and the actual modifyable methods are made available
529 * not as part of this Object class, but as part of the Archive class
530 * instead. This design decision was made for performance and safety
531 * reasons.
532 *
533 * @see Archive::setIntValue() as an example for modifying Object instances.
534 */
535 class Object {
536 public:
537 Object();
538 Object(UIDChain uidChain, DataType type);
539
540 UID uid(int index = 0) const;
541 const UIDChain& uidChain() const;
542 const DataType& type() const;
543 const RawData& rawData() const;
544 Version version() const;
545 Version minVersion() const;
546 bool isVersionCompatibleTo(const Object& other) const;
547 std::vector<Member>& members();
548 const std::vector<Member>& members() const;
549 Member memberNamed(String name) const;
550 Member memberByUID(const UID& uid) const;
551 std::vector<Member> membersOfType(const DataType& type) const;
552 int sequenceIndexOf(const Member& member) const;
553 bool isValid() const;
554 operator bool() const { return isValid(); } ///< Same as calling isValid().
555 //bool operator()() const { return isValid(); }
556 bool operator==(const Object& other) const;
557 bool operator!=(const Object& other) const;
558 bool operator<(const Object& other) const;
559 bool operator>(const Object& other) const;
560
561 protected:
562 void remove(const Member& member);
563 void setVersion(Version v);
564 void setMinVersion(Version v);
565
566 private:
567 DataType m_type;
568 UIDChain m_uid;
569 Version m_version;
570 Version m_minVersion;
571 RawData m_data;
572 std::vector<Member> m_members;
573
574 friend String _encodePrimitiveValue(const Object& obj);
575 friend Object _popObjectBlob(const char*& p, const char* end);
576 friend void _popPrimitiveValue(const char*& p, const char* end, Object& obj);
577 friend String _primitiveObjectValueToString(const Object& obj);
578
579 template<typename T>
580 friend T _primitiveObjectValueToNumber(const Object& obj);
581
582 friend class Archive;
583 };
584
585 /** @brief Destination container for serialization, and source container for deserialization.
586 *
587 * This is the main class for implementing serialization and deserialization
588 * with your C++ application. This framework does not require a a tree
589 * structured layout of your C++ objects being serialized/deserialized, it
590 * uses a concept of a "root" object though. So to start serialization
591 * construct an empty Archive object and then instruct it to serialize your
592 * C++ objects by pointing it to your "root" object:
593 * @code
594 * Archive a;
595 * a.serialize(&myRootObject);
596 * @endcode
597 * Or if you prefer the look of operator based code:
598 * @code
599 * Archive a;
600 * a << myRootObject;
601 * @endcode
602 * The Archive object will then serialize all members of the passed C++
603 * object, and will recursively serialize all other C++ objects which it
604 * contains or points to. So the root object is the starting point for the
605 * overall serialization. After the serialize() method returned, you can
606 * then access the serialized data stream by calling rawData() and send that
607 * data stream over "wire", or store it on disk or whatever you may intend
608 * to do with it.
609 *
610 * Then on receiver side likewise, you create a new Archive object, pass the
611 * received data stream i.e. via constructor to the Archive object and call
612 * deserialize() by pointing it to the root object on receiver side:
613 * @code
614 * Archive a(rawDataStream);
615 * a.deserialize(&myRootObject);
616 * @endcode
617 * Or with operator instead:
618 * @code
619 * Archive a(rawDataStream);
620 * a >> myRootObject;
621 * @endcode
622 * Now this framework automatically handles serialization and
623 * deserialization of fundamental data types automatically for you (like
624 * i.e. char, int, long int, float, double, etc.). However for your own
625 * custom C++ classes and structs you must implement one method which
626 * defines which members of your class should actually be serialized and
627 * deserialized. That method to be added must have the following signature:
628 * @code
629 * void serialize(Serialization::Archive* archive);
630 * @endcode
631 * So let's say you have the following simple data structures:
632 * @code
633 * struct Foo {
634 * int a;
635 * bool b;
636 * double c;
637 * };
638 *
639 * struct Bar {
640 * char one;
641 * float two;
642 * Foo foo1;
643 * Foo* pFoo2;
644 * Foo* pFoo3DontTouchMe; // shall not be serialized/deserialized
645 * };
646 * @endcode
647 * So in order to be able to serialize and deserialize objects of those two
648 * structures you would first add the mentioned method to each struct
649 * definition (i.e. in your header file):
650 * @code
651 * struct Foo {
652 * int a;
653 * bool b;
654 * double c;
655 *
656 * void serialize(Serialization::Archive* archive);
657 * };
658 *
659 * struct Bar {
660 * char one;
661 * float two;
662 * Foo foo1;
663 * Foo* pFoo2;
664 * Foo* pFoo3DontTouchMe; // shall not be serialized/deserialized
665 *
666 * void serialize(Serialization::Archive* archive);
667 * };
668 * @endcode
669 * And then you would implement those two new methods like this (i.e. in
670 * your .cpp file):
671 * @code
672 * #define SRLZ(member) \
673 * archive->serializeMember(*this, member, #member);
674 *
675 * void Foo::serialize(Serialization::Archive* archive) {
676 * SRLZ(a);
677 * SRLZ(b);
678 * SRLZ(c);
679 * }
680 *
681 * void Bar::serialize(Serialization::Archive* archive) {
682 * SRLZ(one);
683 * SRLZ(two);
684 * SRLZ(foo1);
685 * SRLZ(pFoo2);
686 * // leaving out pFoo3DontTouchMe here
687 * }
688 * @endcode
689 * Now when you serialize such a Bar object, this framework will also
690 * automatically serialize the respective Foo object(s) accordingly, also
691 * for the pFoo2 pointer for instance (as long as it is not a NULL pointer
692 * that is).
693 *
694 * Note that there is only one method that you need to implement. So the
695 * respective serialize() method implementation of your classes/structs are
696 * both called for serialization, as well as for deserialization!
697 *
698 * In case you need to enforce backward incompatiblity for one of your C++
699 * classes, you can do so by setting a version and minimum version for your
700 * class (see @c setVersion() and @c setMinVersion() for details).
701 */
702 class Archive {
703 public:
704 Archive();
705 Archive(const RawData& data);
706 Archive(const uint8_t* data, size_t size);
707 virtual ~Archive();
708
709 /** @brief Initiate serialization.
710 *
711 * Initiates serialization of all native C++ objects, which means
712 * capturing and storing the current data of all your C++ objects as
713 * content of this Archive.
714 *
715 * This framework has a concept of a "root" object which you must pass
716 * to this method. The root object is the starting point for
717 * serialization of your C++ objects. The framework will then
718 * recursively serialize all members of that C++ object an continue to
719 * serialize all other C++ objects that it might contain or point to.
720 *
721 * After this method returned, you might traverse all serialized objects
722 * by walking them starting from the rootObject(). You might then modify
723 * that abstract reflection of your C++ objects and finally you might
724 * call rawData() to get an encoded raw data stream which you might use
725 * for sending it "over wire" to somewhere where it is going to be
726 * deserialized later on.
727 *
728 * Note that whenever you call this method, the previous content of this
729 * Archive will first be cleared.
730 *
731 * @param obj - native C++ root object where serialization shall start
732 * @see Archive::operator<<()
733 */
734 template<typename T>
735 void serialize(const T* obj) {
736 m_operation = OPERATION_SERIALIZE;
737 m_allObjects.clear();
738 m_rawData.clear();
739 m_root = UID::from(obj);
740 const_cast<T*>(obj)->serialize(this);
741 encode();
742 m_operation = OPERATION_NONE;
743 }
744
745 /** @brief Initiate deserialization.
746 *
747 * Initiates deserialization of all native C++ objects, which means all
748 * your C++ objects will be restored with the values contained in this
749 * Archive. So that also means calling deserialize() only makes sense if
750 * this a non-empty Archive, which i.e. is the case if you either called
751 * serialize() with this Archive object before or if you passed a
752 * previously serialized raw data stream to the constructor of this
753 * Archive object.
754 *
755 * This framework has a concept of a "root" object which you must pass
756 * to this method. The root object is the starting point for
757 * deserialization of your C++ objects. The framework will then
758 * recursively deserialize all members of that C++ object an continue to
759 * deserialize all other C++ objects that it might contain or point to,
760 * according to the values stored in this Archive.
761 *
762 * @param obj - native C++ root object where deserialization shall start
763 * @see Archive::operator>>()
764 *
765 * @throws Exception if the data stored in this Archive cannot be
766 * restored to the C++ objects passed to this method, i.e.
767 * because of version or type incompatibilities.
768 */
769 template<typename T>
770 void deserialize(T* obj) {
771 Archive a;
772 m_operation = OPERATION_DESERIALIZE;
773 obj->serialize(&a);
774 a.m_root = UID::from(obj);
775 Syncer s(a, *this);
776 m_operation = OPERATION_NONE;
777 }
778
779 /** @brief Initiate serialization of your C++ objects.
780 *
781 * Same as calling @c serialize(), this is just meant if you prefer
782 * to use operator based code instead, which you might find to be more
783 * intuitive.
784 *
785 * Example:
786 * @code
787 * Archive a;
788 * a << myRootObject;
789 * @endcode
790 *
791 * @see Archive::serialize() for more details.
792 */
793 template<typename T>
794 void operator<<(const T& obj) {
795 serialize(&obj);
796 }
797
798 /** @brief Initiate deserialization of your C++ objects.
799 *
800 * Same as calling @c deserialize(), this is just meant if you prefer
801 * to use operator based code instead, which you might find to be more
802 * intuitive.
803 *
804 * Example:
805 * @code
806 * Archive a(rawDataStream);
807 * a >> myRootObject;
808 * @endcode
809 *
810 * @throws Exception if the data stored in this Archive cannot be
811 * restored to the C++ objects passed to this method, i.e.
812 * because of version or type incompatibilities.
813 *
814 * @see Archive::deserialize() for more details.
815 */
816 template<typename T>
817 void operator>>(T& obj) {
818 deserialize(&obj);
819 }
820
821 const RawData& rawData();
822 virtual String rawDataFormat() const;
823
824 /** @brief Serialize a native C/C++ member variable.
825 *
826 * This method is usually called by the serialize() method
827 * implementation of your C/C++ structs and classes, for each of the
828 * member variables that shall be serialized and deserialized
829 * automatically with this framework. It is recommend that you are not
830 * using this method name directly, but rather define a short hand C
831 * macro in your .cpp file like:
832 * @code
833 * #define SRLZ(member) \
834 * archive->serializeMember(*this, member, #member);
835 *
836 * void Foo::serialize(Serialization::Archive* archive) {
837 * SRLZ(a);
838 * SRLZ(b);
839 * SRLZ(c);
840 * }
841 * @endcode
842 * As you can see, using such a macro makes your code more readable and
843 * less error prone.
844 *
845 * It is completely up to you to decide which ones of your member
846 * variables shall automatically be serialized and deserialized with
847 * this framework. Only those member variables which are registered by
848 * calling this method will be serialized and deserialized. It does not
849 * really matter in which order you register your individiual member
850 * variables by calling this method, but the sequence is actually stored
851 * as meta information with the resulting archive and the resulting raw
852 * data stream. That meta information might then be used by this
853 * framework to automatically correct and adapt deserializing that
854 * archive later on for a future (or older) and potentially heavily
855 * modified version of your software. So it is recommended, even though
856 * also not required, that you may retain the sequence of your
857 * serializeMember() calls for your individual C++ classes' members over
858 * all your software versions, to retain backward compatibility of older
859 * archives as much as possible.
860 *
861 * @param nativeObject - native C++ object to be registered for
862 * serialization / deserialization
863 * @param nativeMember - native C++ member variable of @a nativeObject
864 * to be registered for serialization /
865 * deserialization
866 * @param memberName - name of @a nativeMember to be stored with this
867 * archive
868 */
869 template<typename T_classType, typename T_memberType>
870 void serializeMember(const T_classType& nativeObject, const T_memberType& nativeMember, const char* memberName) {
871 const size_t offset =
872 ((const uint8_t*)(const void*)&nativeMember) -
873 ((const uint8_t*)(const void*)&nativeObject);
874 const UIDChain uids = UIDChainResolver<T_memberType>(nativeMember);
875 const DataType type = DataType::dataTypeOf(nativeMember);
876 const Member member(memberName, uids[0], offset, type);
877 const UID parentUID = UID::from(nativeObject);
878 Object& parent = m_allObjects[parentUID];
879 if (!parent) {
880 const UIDChain uids = UIDChainResolver<T_classType>(nativeObject);
881 const DataType type = DataType::dataTypeOf(nativeObject);
882 parent = Object(uids, type);
883 }
884 parent.members().push_back(member);
885 const Object obj(uids, type);
886 const bool bExistsAlready = m_allObjects.count(uids[0]);
887 const bool isValidObject = obj;
888 const bool bExistingObjectIsInvalid = !m_allObjects[uids[0]];
889 if (!bExistsAlready || (bExistingObjectIsInvalid && isValidObject)) {
890 m_allObjects[uids[0]] = obj;
891 // recurse serialization for all members of this member
892 // (only for struct/class types, noop for primitive types)
893 SerializationRecursion<T_memberType>::serializeObject(this, nativeMember);
894 }
895 }
896
897 /** @brief Set current version number for your C++ class.
898 *
899 * By calling this method you can define a version number for your
900 * current C++ class (that is a version for its current data structure
901 * layout and method implementations) that is going to be stored along
902 * with the serialized archive. Only call this method if you really want
903 * to constrain compatibility of your C++ class.
904 *
905 * Along with calling @c setMinVersion() this provides a way for you
906 * to constrain backward compatibility regarding serialization and
907 * deserialization of your C++ class which the Archive class will obey
908 * to. If required, then typically you might do so in your
909 * @c serialize() method implementation like:
910 * @code
911 * #define SRLZ(member) \
912 * archive->serializeMember(*this, member, #member);
913 *
914 * void Foo::serialize(Serialization::Archive* archive) {
915 * // when serializing: the current version of this class that is
916 * // going to be stored with the serialized archive
917 * archive->setVersion(*this, 6);
918 * // when deserializing: the minimum version this C++ class is
919 * // compatible with
920 * archive->setMinVersion(*this, 3);
921 * // actual data mebers to serialize / deserialize
922 * SRLZ(a);
923 * SRLZ(b);
924 * SRLZ(c);
925 * }
926 * @endcode
927 * In this example above, the C++ class "Foo" would be serialized along
928 * with the version number @c 6 and minimum version @c 3 as additional
929 * meta information in the resulting archive (and its raw data stream
930 * respectively).
931 *
932 * When deserializing archives with the example C++ class code above,
933 * the Archive object would check whether your originally serialized
934 * C++ "Foo" object had at least version number @c 3, if not the
935 * deserialization process would automatically be stopped with a
936 * @c Serialization::Exception, claiming that the classes are version
937 * incompatible.
938 *
939 * But also consider the other way around: you might have serialized
940 * your latest version of your C++ class, and might deserialize that
941 * archive with an older version of your C++ class. In that case it will
942 * likewise be checked whether the version of that old C++ class is at
943 * least as high as the minimum version set with the already seralized
944 * bleeding edge C++ class.
945 *
946 * Since this Serialization / deserialization framework is designed to
947 * be robust on changes to your C++ classes and aims trying to
948 * deserialize all your C++ objects correctly even if your C++ classes
949 * have seen substantial software changes in the meantime; you might
950 * sometimes see it as necessary to constrain backward compatibility
951 * this way. Because obviously there are certain things this framework
952 * can cope with, like for example that you renamed a data member while
953 * keeping the layout consistent, or that you have added new members to
954 * your C++ class or simply changed the order of your members in your
955 * C++ class. But what this framework cannot detect is for example if
956 * you changed the semantics of the values stored with your members, or
957 * even substantially changed the algorithms in your class methods such
958 * that they would not handle the data of your C++ members in the same
959 * and correct way anymore.
960 *
961 * @param nativeObject - your C++ object you want to set a version for
962 * @param v - the version number to set for your C++ class (by default,
963 * that is if you do not explicitly call this method, then
964 * your C++ object will be stored with version number @c 0 ).
965 */
966 template<typename T_classType>
967 void setVersion(const T_classType& nativeObject, Version v) {
968 const UID uid = UID::from(nativeObject);
969 Object& obj = m_allObjects[uid];
970 if (!obj) {
971 const UIDChain uids = UIDChainResolver<T_classType>(nativeObject);
972 const DataType type = DataType::dataTypeOf(nativeObject);
973 obj = Object(uids, type);
974 }
975 setVersion(obj, v);
976 }
977
978 /** @brief Set a minimum version number for your C++ class.
979 *
980 * Call this method to define a minimum version that your current C++
981 * class implementation would be compatible with when it comes to
982 * deserialization of an archive containing an object of your C++ class.
983 * Like the version information, the minimum version will also be stored
984 * for objects of your C++ class with the resulting archive (and its
985 * resulting raw data stream respectively).
986 *
987 * When you start to constrain version compatibility of your C++ class
988 * you usually start by using 1 as version and 1 as minimum version.
989 * So it is eligible to set the same number to both version and minimum
990 * version. However you must @b not set a minimum version higher than
991 * version. Doing so would not raise an exception, but the resulting
992 * behavior would be undefined.
993 *
994 * It is not relevant whether you first set version and then minimum
995 * version or vice versa. It is also not relevant when exactly you set
996 * those two numbers, even though usually you would set both in your
997 * serialize() method implementation.
998 *
999 * @see @c setVersion() for more details about this overall topic.
1000 *
1001 * @param nativeObject - your C++ object you want to set a version for
1002 * @param v - the minimum version you want to define for your C++ class
1003 * (by default, that is if you do not explicitly call this
1004 * method, then a minium version of @c 0 is assumed for your
1005 * C++ class instead).
1006 */
1007 template<typename T_classType>
1008 void setMinVersion(const T_classType& nativeObject, Version v) {
1009 const UID uid = UID::from(nativeObject);
1010 Object& obj = m_allObjects[uid];
1011 if (!obj) {
1012 const UIDChain uids = UIDChainResolver<T_classType>(nativeObject);
1013 const DataType type = DataType::dataTypeOf(nativeObject);
1014 obj = Object(uids, type);
1015 }
1016 setMinVersion(obj, v);
1017 }
1018
1019 virtual void decode(const RawData& data);
1020 virtual void decode(const uint8_t* data, size_t size);
1021 void clear();
1022 bool isModified() const;
1023 void removeMember(Object& parent, const Member& member);
1024 void remove(const Object& obj);
1025 Object& rootObject();
1026 Object& objectByUID(const UID& uid);
1027 void setAutoValue(Object& object, String value);
1028 void setIntValue(Object& object, int64_t value);
1029 void setRealValue(Object& object, double value);
1030 void setBoolValue(Object& object, bool value);
1031 void setEnumValue(Object& object, uint64_t value);
1032 String valueAsString(const Object& object);
1033 int64_t valueAsInt(const Object& object);
1034 double valueAsReal(const Object& object);
1035 bool valueAsBool(const Object& object);
1036 void setVersion(Object& object, Version v);
1037 void setMinVersion(Object& object, Version v);
1038 String name() const;
1039 void setName(String name);
1040 String comment() const;
1041 void setComment(String comment);
1042 time_t timeStampCreated() const;
1043 time_t timeStampModified() const;
1044 tm dateTimeCreated(time_base_t base = LOCAL_TIME) const;
1045 tm dateTimeModified(time_base_t base = LOCAL_TIME) const;
1046
1047 protected:
1048 // UID resolver for non-pointer types
1049 template<typename T>
1050 class UIDChainResolver {
1051 public:
1052 UIDChainResolver(const T& data) {
1053 m_uid.push_back(UID::from(data));
1054 }
1055
1056 operator UIDChain() const { return m_uid; }
1057 UIDChain operator()() const { return m_uid; }
1058 private:
1059 UIDChain m_uid;
1060 };
1061
1062 // UID resolver for pointer types (of 1st degree)
1063 template<typename T>
1064 class UIDChainResolver<T*> {
1065 public:
1066 UIDChainResolver(const T*& data) {
1067 const UID uids[2] = {
1068 { &data, sizeof(data) },
1069 { data, sizeof(*data) }
1070 };
1071 m_uid.push_back(uids[0]);
1072 m_uid.push_back(uids[1]);
1073 }
1074
1075 operator UIDChain() const { return m_uid; }
1076 UIDChain operator()() const { return m_uid; }
1077 private:
1078 UIDChain m_uid;
1079 };
1080
1081 // SerializationRecursion for non-pointer class/struct types.
1082 template<typename T, bool T_isRecursive>
1083 struct SerializationRecursionImpl {
1084 static void serializeObject(Archive* archive, const T& obj) {
1085 const_cast<T&>(obj).serialize(archive);
1086 }
1087 };
1088
1089 // SerializationRecursion for pointers (of 1st degree) to class/structs.
1090 template<typename T, bool T_isRecursive>
1091 struct SerializationRecursionImpl<T*,T_isRecursive> {
1092 static void serializeObject(Archive* archive, const T*& obj) {
1093 if (!obj) return;
1094 const_cast<T*&>(obj)->serialize(archive);
1095 }
1096 };
1097
1098 // NOOP SerializationRecursion for primitive types.
1099 template<typename T>
1100 struct SerializationRecursionImpl<T,false> {
1101 static void serializeObject(Archive* archive, const T& obj) {}
1102 };
1103
1104 // NOOP SerializationRecursion for pointers (of 1st degree) to primitive types.
1105 template<typename T>
1106 struct SerializationRecursionImpl<T*,false> {
1107 static void serializeObject(Archive* archive, const T*& obj) {}
1108 };
1109
1110 // Automatically handles recursion for class/struct types, while ignoring all primitive types.
1111 template<typename T>
1112 struct SerializationRecursion : SerializationRecursionImpl<T, LIBGIG_IS_CLASS(T)> {
1113 };
1114
1115 class ObjectPool : public std::map<UID,Object> {
1116 public:
1117 // prevent passing obvious invalid UID values from creating a new pair entry
1118 Object& operator[](const UID& k) {
1119 static Object invalid;
1120 if (!k.isValid()) {
1121 invalid = Object();
1122 return invalid;
1123 }
1124 return std::map<UID,Object>::operator[](k);
1125 }
1126 };
1127
1128 friend String _encode(const ObjectPool& objects);
1129
1130 private:
1131 String _encodeRootBlob();
1132 void _popRootBlob(const char*& p, const char* end);
1133 void _popObjectsBlob(const char*& p, const char* end);
1134
1135 protected:
1136 class Syncer {
1137 public:
1138 Syncer(Archive& dst, Archive& src);
1139 protected:
1140 void syncObject(const Object& dst, const Object& src);
1141 void syncPrimitive(const Object& dst, const Object& src);
1142 void syncPointer(const Object& dst, const Object& src);
1143 void syncMember(const Member& dstMember, const Member& srcMember);
1144 static Member dstMemberMatching(const Object& dstObj, const Object& srcObj, const Member& srcMember);
1145 private:
1146 Archive& m_dst;
1147 Archive& m_src;
1148 };
1149
1150 enum operation_t {
1151 OPERATION_NONE,
1152 OPERATION_SERIALIZE,
1153 OPERATION_DESERIALIZE
1154 };
1155
1156 virtual void encode();
1157
1158 ObjectPool m_allObjects;
1159 operation_t m_operation;
1160 UID m_root;
1161 RawData m_rawData;
1162 bool m_isModified;
1163 String m_name;
1164 String m_comment;
1165 time_t m_timeCreated;
1166 time_t m_timeModified;
1167 };
1168
1169 /**
1170 * Will be thrown whenever an error occurs during an serialization or
1171 * deserialization process.
1172 */
1173 class Exception {
1174 public:
1175 String Message;
1176
1177 Exception(String format, ...);
1178 Exception(String format, va_list arg);
1179 void PrintMessage();
1180 virtual ~Exception() {}
1181
1182 protected:
1183 Exception();
1184 static String assemble(String format, va_list arg);
1185 };
1186
1187 } // namespace Serialization
1188
1189 #endif // LIBGIG_SERIALIZATION_H

  ViewVC Help
Powered by ViewVC