ACE_Message_Block-----ACE消息的存储

winston · 发表于 2012-2-29 00:19:06

ACE_Message_Block在Ace中用来表示消息的存放空间，可用做网络通信中的消息缓冲区，使用非常频繁，下面将在如下方简单的介绍一下ACE_Message_Block相关功能。使用案例：[cpp] view plain copy print ?

ACE_Message_Block * head = NULL;
ACE_NEW_RETURN(head, ACE_Message_Block( sizeof( void * ) ), -1 );
ACE_Message_Block *msg = head;
ACE_OS::memcpy( msg->wr_ptr(), &pMsg, sizeof( void *) );
msg->wr_ptr(sizeof( void *));

ACE_Message_Block * head = NULL; ACE_NEW_RETURN(head, ACE_Message_Block( sizeof( void * ) ), -1 ); ACE_Message_Block *msg = head; ACE_OS::memcpy( msg->wr_ptr(), &pMsg, sizeof( void *) ); msg->wr_ptr(sizeof( void *)); 下面是转自网络：1。创建消息块 创建消息块的方式比较灵活，常用的有以下几种方式： 1。直接给消息块分配内存空间创建。 ACE_Message_Block *mb =new ACE_Message_Block (30); 2。共享底层数据块创建。 char buffer[100];
ACE_Message_Block *mb = new ACE_Message_Block (buffer,30);这种方式共享底层的数据块，被创建的消息块并不拷贝该数据，也不假定自己拥有它的所有权。在消息块mb被销毁时，相关联的数据缓冲区data将不会被销毁。这是有意义的：消息块没有拷贝数据，因此内存也不是它分配的，这样它也不应该负责销毁它。3。通过duplicate()函数从已有的消息块中创建副本。 ACE_Message_Block *mb =new ACE_Message_Block (30);
ACE_Message_Block *mb2 = mb->duplicate(); 这种方式下，mb2和mb共享同一数据空间，使用的是ACE_Message_Block的引用计数机制。它返回指向要被复制的消息块的指针，并在内部增加内部引用计数。4。通过clone()函数从已有的消息块中复制。 ACE_Message_Block *mb =new ACE_Message_Block (30);
ACE_Message_Block *mb2 = mb->clone(); clone()方法实际地创建整个消息块的新副本，包括它的数据块和附加部分；也就是说，这是一次"深拷贝"。2。释放消息块 一旦使用完消息块，程序员可以调用它的release()方法来释放它。

如果消息数据内存是由该消息块分配的，调用release()方法就也会释放此内存。
如果消息块是引用计数的，release()就会减少计数，直到到达０为止；之后消息块和与它相关联的数据块才从内存中被移除。
如果消息块是通过共享已分配的底层数据块创建的，底层数据块不会被释放。

无论消息块是哪种方式创建的，只要在使用完后及时调用release()函数，就能确保相应的内存能正确的释放。 3。从消息块中读写数据 ACE_Message_Block提供了两个指针函数以供程序员进行读写操作，rd_ptr()指向可读的数据块地址，wr_ptr()指向可写的数据块地址，默认情况下都执行数据块的首地址。下面的例子简单了演示它的使用方法。#include"ace/Message_Queue.h"
#include "ace/OS.h"

int main(int argc,char *argv[])
{
ACE_Message_Block *mb = new ACE_Message_Block (30);
ACE_OS::sprintf(mb->wr_ptr(),"%s","hello");
ACE_OS::printf("%s\n",mb->rd_ptr ());
mb->release();
return 0;
} 注意：这两个指针所指向的位置并不会自动移动，在上面的例子中，函数执行完毕后，执行的位置仍然是最开始的0，而不是最新的可写位置5，程序员需要通过wr_ptr(5)函数手动移动写指针的位置。4。数据的拷贝 一般的数据的拷贝可以通过函数来实现数据的拷贝，copy()还会保证wr_ptr()的更新，使其指向缓冲区的新末尾处。下面的例子演示了copy()函数的用法。 mb->copy("hello");
mb->copy("123",4); 注意：由于c++是以'\0'作为字符串结束标志的，对于上面的例子，底层数据块中保存的是"hello\0123\0"，而用ACE_OS::printf("%s\n",mb->rd_ptr ());打印出来的结果是"hello"，使用copy函数进行字符串连接的时候需要注意。5。其它常用函数

length() 返回当前的数据长度
next() 获取和设置下一个ACE_Message_Block的链接。（用来建立消息队列非常有用）
space() 获取剩余可用空间大小
size() 获取和设置数据存储空间大小。

下面是源代码，可以直接调用，对数据块操作：
[cpp] view plain copy print ?

// -*- C++ -*-
//==========================================================================
/**
* @file Message_Block.h
*
* $Id: Message_Block.h 93359 2011-02-11 11:33:12Z mcorino $
*
* @author Douglas C. Schmidt <schmidt@cs.wustl.edu>
*/
//==========================================================================
#ifndef ACE_MESSAGE_BLOCK_H
#define ACE_MESSAGE_BLOCK_H
#include /**/ "ace/pre.h"
#include "ace/config-lite.h"
#include /**/ "ace/ACE_export.h"
#if !defined (ACE_LACKS_PRAGMA_ONCE)
# pragma once
#endif /* ACE_LACKS_PRAGMA_ONCE */
#include "ace/Default_Constants.h"
#include "ace/Global_Macros.h"
#include "ace/Time_Value.h"
ACE_BEGIN_VERSIONED_NAMESPACE_DECL
// Forward declaration.
class ACE_Allocator;
class ACE_Data_Block;
class ACE_Lock;
/**
* @class ACE_Message_Block
*
* @brief Stores messages for use throughout ACE (particularly
* in an ACE_Message_Queue).
*
* An ACE_Message_Block is modeled after the message data
* structures used in System V STREAMS. Its purpose is to
* enable efficient manipulation of arbitrarily large messages
* without incurring much memory copying overhead. Here are the
* main characteristics of an ACE_Message_Block:
* - Contains a pointer to a reference-counted
* ACE_Data_Block, which in turn points to the actual data
* buffer. This allows very flexible and efficient sharing of
* data by multiple ACE_Message_Block objects.
* - One or more ACE_Message_Blocks can be linked to form a
* ``fragment chain.''
* - ACE_Message_Blocks can be linked together in a doubly linked fashion
* to form a queue of messages (this is how ACE_Message_Queue works).
*
* @see C++NPv1, section 4.2; APG, section 12.3.2.
*/
class ACE_Export ACE_Message_Block
{
public:
friend class ACE_Data_Block;
enum
{
// = Data and proto
/// Undifferentiated data message
MB_DATA = 0x01,
/// Undifferentiated protocol control
MB_PROTO = 0x02,
// = Control messages
/// Line break (regular and priority)
MB_BREAK = 0x03,
/// Pass file pointer
MB_PASSFP = 0x04,
/// Post an event to an event queue
MB_EVENT = 0x05,
/// Generate process signal
MB_SIG = 0x06,
/// ioctl; set/get params
MB_IOCTL = 0x07,
/// Set various stream head options
MB_SETOPTS = 0x08,
// = Control messages
/// Acknowledge ioctl (high priority; go to head of queue)
MB_IOCACK = 0x81,
/// Negative ioctl acknowledge
MB_IOCNAK = 0x82,
/// Priority proto message
MB_PCPROTO = 0x83,
/// Generate process signal
MB_PCSIG = 0x84,
/// Generate read notification
MB_READ = 0x85,
/// Flush your queues
MB_FLUSH = 0x86,
/// Stop transmission immediately
MB_STOP = 0x87,
/// Restart transmission after stop
MB_START = 0x88,
/// Line disconnect
MB_HANGUP = 0x89,
/// Fatal error used to set u.u_error
MB_ERROR = 0x8a,
/// Post an event to an event queue
MB_PCEVENT = 0x8b,
// = Message class masks
/// Normal priority message mask
MB_NORMAL = 0x00,
/// High priority control message mask
MB_PRIORITY = 0x80,
/// User-defined message mask
MB_USER = 0x200
};
typedef int ACE_Message_Type;
typedef unsigned long Message_Flags;
enum
{
/// Don't delete the data on exit since we don't own it.
DONT_DELETE = 01,
/// user defined flags start here
USER_FLAGS = 0x1000
};
// = Initialization and termination.
/// Create an empty message.
ACE_Message_Block (ACE_Allocator *message_block_allocator = 0);
/**
* Create an ACE_Message_Block that owns the specified ACE_Data_Block
* without copying it. If the @a flags is set to @c DONT_DELETE we
* don't delete the ACE_Data_Block. It is left to the client's
* responsibility to take care of the memory allocated for the
* data_block
*/
ACE_Message_Block (ACE_Data_Block *data_block,
Message_Flags flags = 0,
ACE_Allocator *message_block_allocator = 0);
/**
* Create an ACE_Message_Block that refers to @a data without
* copying it. The @a data memory will not be freed when this block is
* destroyed; memory management of @a data is left to the caller.
* Note that the @c size of the new ACE_Message_Block will be @a size, but
* the @c length will be 0 until the write pointer is changed.
*/
ACE_Message_Block (const char *data,
size_t size = 0,
unsigned long priority = ACE_DEFAULT_MESSAGE_BLOCK_PRIORITY);
/**
* Create an initialized message of type @a type containing @a size
* bytes. The @a cont argument initializes the continuation field in
* the ACE_Message_Block. If @a data == 0 then this block allocates and
* owns the block's memory, using @a allocator to get the data if it's
* non-0. If @a data != 0 then this block refers to that memory until
* this this block ceases to exist; this object will not free @a data on
* destruction. If @a locking_strategy is non-0 then this is used
* to protect regions of code that access shared state (e.g.,
* reference counting) from race conditions. Note that the @c size
* of the ACE_Message_Block will be @a size, but the @c length will be 0
* until the write pointer is set. The @a data_block_allocator is used to
* allocate the data blocks while the @a allocator_strategy is used
* to allocate the buffers contained by those. The
* @a message_block_allocator is used to allocate new ACE_Message_Block
* objects when the duplicate() method is called. If a
* @a message_block_allocator is given, this ACE_Message_Block and
* future ACE_Message_Block objects created by duplicate() will be
* freed using this allocator when they are released.
* @note If you use this allocator, the ACE_Message_Block you created
* should have been created using this allocator because it will be
* released to the same allocator.
*/
ACE_Message_Block (size_t size,
ACE_Message_Type type = MB_DATA,
ACE_Message_Block *cont = 0,
const char *data = 0,
ACE_Allocator *allocator_strategy = 0,
ACE_Lock *locking_strategy = 0,
unsigned long priority = ACE_DEFAULT_MESSAGE_BLOCK_PRIORITY,
const ACE_Time_Value &execution_time = ACE_Time_Value::zero,
const ACE_Time_Value &deadline_time = ACE_Time_Value::max_time,
ACE_Allocator *data_block_allocator = 0,
ACE_Allocator *message_block_allocator = 0);
/**
* A copy constructor. This constructor is a bit different. If the
* incoming Message Block has a data block from the stack this
* constructor does a deep copy ie. allocates a new data block on
* the heap and does a copy of the data from the incoming message
* block. As a final note, the alignment information is used to
* align the data block if it is created afresh. If the incoming
* @a mb has a data block has a data block allocated from the heap,
* then this constructor just duplicates (ie. a shallow copy) the
* data block of the incoming @a mb.
*/
ACE_Message_Block (const ACE_Message_Block &mb,
size_t align);
/**
* Create a Message Block that assumes it has ownership of @a data,
* but in reality it doesnt (i.e., cannot delete it since it didn't
* malloc it!). Note that the @c size of the Message_Block will
* be @a size, but the @a length will be 0 until <wr_ptr> is set.
*/
int init (const char *data,
size_t size = 0);
/**
* Create an initialized message of type @a type containing @a size
* bytes. The @a cont argument initializes the continuation field in
* the Message_Block. If @a data == 0 then we create and own the
* @a data, using @a allocator_strategy to get the data if it's non-0. If
* @a data != 0 we assume that we have ownership of the @a data till
* this object ceases to exist (and don't delete it during
* destruction). If @a locking_strategy is non-0 then this is used
* to protect regions of code that access shared state (e.g.,
* reference counting) from race conditions. Note that the @a size
* of the Message_Block will be @a size, but the @a length will be 0
* until <wr_ptr> is set. The @a data_block_allocator is use to
* allocate the data blocks while the @a allocator_strategy is used
* to allocate the buffers contained by those.
*/
int init (size_t size,
ACE_Message_Type type = MB_DATA,
ACE_Message_Block *cont = 0,
const char *data = 0,
ACE_Allocator *allocator_strategy = 0,
ACE_Lock *locking_strategy = 0,
unsigned long priority = ACE_DEFAULT_MESSAGE_BLOCK_PRIORITY,
const ACE_Time_Value &execution_time = ACE_Time_Value::zero,
const ACE_Time_Value &deadline_time = ACE_Time_Value::max_time,
ACE_Allocator *data_block_allocator = 0,
ACE_Allocator *message_block_allocator = 0);
/**
* Delete all the resources held in the message.
*
* @note Note that release() is designed to release the continuation
* chain; the destructor is not. See release() for details.
*/
virtual ~ACE_Message_Block (void);
// = Message Type accessors and mutators.
/// Get type of the message.
ACE_Message_Type msg_type (void) const;
/// Set type of the message.
void msg_type (ACE_Message_Type type);
/// Find out what type of message this is.
int is_data_msg (void) const;
/// Find out what class of message this is (there are two classes,
/// @c normal messages and @c high-priority messages).
ACE_Message_Type msg_class (void) const;
// = Message flag accessors and mutators.
/// Bitwise-or the @a more_flags into the existing message flags and
/// return the new value.
Message_Flags set_flags (Message_Flags more_flags);
/// Clear the message flag bits specified in @a less_flags and return
/// the new value.
Message_Flags clr_flags (Message_Flags less_flags);
/// Get the current message flags.
Message_Flags flags (void) const;
// = Data Block flag accessors and mutators.
/// Bitwise-or the @a more_flags into the existing message flags and
/// return the new value.
/* @todo: I think the following set of methods could not be used at
* all. May be they are useless. Let us have it so that we dont
* mess up memory management of the Message_Block. Somebody correct
* me if I am totally totally wrong..
*/
Message_Flags set_self_flags (ACE_Message_Block::Message_Flags more_flags);
/// Clear the message flag bits specified in @a less_flags and return
/// the new value.
Message_Flags clr_self_flags (ACE_Message_Block::Message_Flags less_flags);
/// Get the current message flags.
Message_Flags self_flags (void) const;
/// Get priority of the message.
unsigned long msg_priority (void) const;
/// Set priority of the message.
void msg_priority (unsigned long priority);
/// Get execution time associated with the message.
const ACE_Time_Value &msg_execution_time (void) const;
/// Set execution time associated with the message.
void msg_execution_time (const ACE_Time_Value &et);
/// Get absolute time of deadline associated with the message.
const ACE_Time_Value &msg_deadline_time (void) const;
/// Set absolute time of deadline associated with the message.
void msg_deadline_time (const ACE_Time_Value &dt);
// = Deep copy and shallow copy methods.
/// Return an exact "deep copy" of the message, i.e., create fresh
/// new copies of all the Data_Blocks and continuations.
virtual ACE_Message_Block *clone (Message_Flags mask = 0) const;
/// Return a "shallow" copy that increments our reference count by 1.
virtual ACE_Message_Block *duplicate (void) const;
/**
* Return a "shallow" copy that increments our reference count by 1.
* This is similar to CORBA's <_duplicate> method, which is useful
* if you want to eliminate lots of checks for NULL @a mb pointers
* before calling <_duplicate> on them.
*/
static ACE_Message_Block *duplicate (const ACE_Message_Block *mb);
/**
* Decrease the shared ACE_Data_Block's reference count by 1. If the
* ACE_Data_Block's reference count goes to 0, it is deleted.
* In all cases, this ACE_Message_Block is deleted - it must have come
* from the heap, or there will be trouble.
*
* release() is designed to release the continuation chain; the
* destructor is not. If we make the destructor release the
* continuation chain by calling release() or delete on the message
* blocks in the continuation chain, the following code will not
* work since the message block in the continuation chain is not off
* the heap:
*
* ACE_Message_Block mb1 (1024);
* ACE_Message_Block mb2 (1024);
*
* mb1.cont (&mb2);
*
* And hence, call release() on a dynamically allocated message
* block. This will release all the message blocks in the
* continuation chain. If you call delete or let the message block
* fall off the stack, cleanup of the message blocks in the
* continuation chain becomes the responsibility of the user.
*
* @retval 0, always, and the object this method was invoked on is no
* longer valid.
*/
virtual ACE_Message_Block *release (void);
/**
* This behaves like the non-static method <release>, except that it
* checks if @a mb is 0. This is similar to <CORBA::release>, which
* is useful if you want to eliminate lots of checks for NULL
* pointers before calling <release> on them. Returns @a mb.
*/
static ACE_Message_Block *release (ACE_Message_Block *mb);
// = Operations on Message data
/**
* Copies data into this ACE_Message_Block. Data is copied into the
* block starting at the current write pointer.
*
* @param buf Pointer to the buffer to copy from.
* @param n The number of bytes to copy.
*
* @retval 0 on success; the write pointer is advanced by @arg n.
* @retval -1 if the amount of free space following the write pointer
* in the block is less than @arg n. Free space can be checked
* by calling space().
*/
int copy (const char *buf, size_t n);
/**
* Copies a 0-terminated character string into this ACE_Message_Block.
* The string is copied into the block starting at the current write
* pointer. The 0-terminator is included in the copied data.
*
* @param buf Pointer to the character string to copy from.
*
* @retval 0 on success; the write pointer is advanced by the string's
* length, including the 0 terminator.
* @retval -1 if the amount of free space following the write pointer
* in the block is less than required to hold the entire string.
* Free space can be checked by calling space().
*/
int copy (const char *buf);
/// Normalizes data in the top-level Message_Block to align with the base,
/// i.e., it "shifts" the data pointed to by <rd_ptr> down to the <base> and
/// then readjusts <rd_ptr> to point to <base> and <wr_ptr> to point
/// to <base> + the length of the moved data. Returns -1 and does
/// nothing if the <rd_ptr> is > <wr_ptr>, else 0 on success.
int crunch (void);
/// Resets the Message Block data to contain nothing, i.e., sets the
/// read and write pointers to align with the base.
void reset (void);
/// Access all the allocators in the message block.
/// @todo Not sure whether we would need finer control while
/// trying to access allocators ie. a method for every allocator.
/**
* This method returns the allocators only from the first message
* block in the chain.
*
* @param allocator_strategy Strategy used to allocate the
* underlying buffer
*
* @param data_block_allocator Strategy used to allocate the
* underlying data block
*
* @param message_block_allocator Strategy used to allocate the
* message block
*/
void access_allocators (ACE_Allocator *&allocator_strategy,
ACE_Allocator *&data_block_allocator,
ACE_Allocator *&message_block_allocator);
/// Reset all the allocators in the message block.
/// @todo Not sure whether we would need finer control while
/// trying to reset allocators ie. a method for every allocator.
/**
* This method resets the allocators in all the message blocks in
* the chain.
*/
void reset_allocators (ACE_Allocator *allocator_strategy = 0,
ACE_Allocator *data_block_allocator = 0,
ACE_Allocator *message_block_allocator = 0);
/// Get message data.
char *base (void) const;
/// Set message data (doesn't reallocate).
void base (char *data,
size_t size,
Message_Flags = DONT_DELETE);
/// Return a pointer to 1 past the end of the allocated data in a message.
char *end (void) const;
/**
* Return a pointer to 1 past the end of the allotted data in a message.
* Allotted data may be less than allocated data if a value smaller than
* capacity() to is passed to size().
*/
char *mark (void) const;
/// Get the read pointer.
char *rd_ptr (void) const;
/// Set the read pointer to @a ptr.
void rd_ptr (char *ptr);
/// Set the read pointer ahead @a n bytes.
void rd_ptr (size_t n);
/// Get the write pointer.
char *wr_ptr (void) const;
/// Set the write pointer to @a ptr.
void wr_ptr (char *ptr);
/// Set the write pointer ahead @a n bytes. This is used to compute
/// the <length> of a message.
void wr_ptr (size_t n);
/** @name Message length and size operations
*
* Message length is (wr_ptr - rd_ptr).
*
* Message size is capacity of the message, including data outside
* the [rd_ptr,wr_ptr] range.
*/
//@{
/// Get the length of the message
size_t length (void) const;
/// Set the length of the message
void length (size_t n);
/// Get the length of the Message_Blocks, including chained
/// Message_Blocks.
size_t total_length (void) const;
/// Get the total number of bytes in all Message_Blocks, including
/// chained Message_Blocks.
size_t total_size (void) const;
/// Get the total number of bytes and total length in all
/// Message_Blocks, including chained Message_Blocks.
void total_size_and_length (size_t &mb_size,
size_t &mb_length) const;
/// Get the number of bytes in the top-level Message_Block (i.e.,
/// does not consider the bytes in chained Message_Blocks).
size_t size (void) const;
/**
* Set the number of bytes in the top-level Message_Block,
* reallocating space if necessary. However, the @c rd_ptr_ and
* @c wr_ptr_ remain at the original offsets into the buffer, even if
* it is reallocated. Returns 0 if successful, else -1.
*/
int size (size_t length);
/// Get the number of allocated bytes in all Message_Block, including
/// chained Message_Blocks.
size_t total_capacity (void) const;
/// Get the number of allocated bytes in the top-level Message_Block.
size_t capacity (void) const;
/// Get the number of bytes available after the <wr_ptr_> in the
/// top-level Message_Block.
size_t space (void) const;
//@}
// = ACE_Data_Block methods.
/**
* Get a pointer to the data block. Note that the ACE_Message_Block
* still references the block; this call does not change the reference
* count.
*/
ACE_Data_Block *data_block (void) const;
/**
* Set a new data block pointer. The original ACE_Data_Block is released
* as a result of this call. If you need to keep the original block, call
* <replace_data_block> instead. Upon return, this ACE_Message_Block
* holds a pointer to the new ACE_Data_Block, taking over the reference
* you held on it prior to the call.
*/
void data_block (ACE_Data_Block *);
/// Set a new data block pointer. A pointer to the original ACE_Data_Block
/// is returned, and not released (as it is with <data_block>).
ACE_Data_Block *replace_data_block (ACE_Data_Block*);
// = The continuation field chains together composite messages.
/// Get the continuation field.
ACE_Message_Block *cont (void) const;
/// Set the continuation field.
void cont (ACE_Message_Block *);
// = Pointer to the Message_Block directly ahead in the ACE_Message_Queue.
/// Get link to next message.
ACE_Message_Block *next (void) const;
/// Set link to next message.
void next (ACE_Message_Block *);
// = Pointer to the Message_Block directly behind in the ACE_Message_Queue.
/// Get link to prev message.
ACE_Message_Block *prev (void) const;
/// Set link to prev message.
void prev (ACE_Message_Block *);
// = The locking strategy prevents race conditions.
/// Get the locking strategy.
ACE_Lock *locking_strategy (void);
/// Set a new locking strategy and return the hold one.
ACE_Lock *locking_strategy (ACE_Lock *);
/// Get the current reference count.
int reference_count (void) const;
/// Dump the state of an object.
void dump (void) const;
/// Declare the dynamic allocation hooks.
ACE_ALLOC_HOOK_DECLARE;
protected:
// = Internal initialization methods.
/// Perform the actual initialization.
ACE_Message_Block (size_t size,
ACE_Message_Type type,
ACE_Message_Block *cont,
const char *data,
ACE_Allocator *allocator_strategy,
ACE_Lock *locking_strategy,
Message_Flags flags,
unsigned long priority,
const ACE_Time_Value &execution_time,
const ACE_Time_Value &deadline_time,
ACE_Data_Block *db,
ACE_Allocator *data_block_allocator,
ACE_Allocator *message_block_allocator);
/// Internal release implementation
/// Returns 1 if the data block has to be destroyed.
int release_i (ACE_Lock *lock);
/// Perform the actual initialization.
int init_i (size_t size,
ACE_Message_Type type,
ACE_Message_Block *cont,
const char *data,
ACE_Allocator *allocator_strategy,
ACE_Lock *locking_strategy,
Message_Flags flags,
unsigned long priority,
const ACE_Time_Value &execution_time,
const ACE_Time_Value &deadline_time,
ACE_Data_Block *db,
ACE_Allocator *data_block_allocator,
ACE_Allocator *message_block_allocator);
/// Pointer to beginning of next read.
size_t rd_ptr_;
/// Pointer to beginning of next write.
size_t wr_ptr_;
/// Priority of message.
unsigned long priority_;
#if defined (ACE_HAS_TIMED_MESSAGE_BLOCKS)
/// Execution time associated with the message.
ACE_Time_Value execution_time_;
/// Absolute deadline time for message.
ACE_Time_Value deadline_time_;
#endif /* ACE_HAS_TIMED_MESSAGE_BLOCKS */
// = Links to other ACE_Message_Block *s.
/// Pointer to next message block in the chain.
ACE_Message_Block *cont_;
/// Pointer to next message in the list.
ACE_Message_Block *next_;
/// Pointer to previous message in the list.
ACE_Message_Block *prev_;
/// Misc flags (e.g., DONT_DELETE and USER_FLAGS).
ACE_Message_Block::Message_Flags flags_;
/// Pointer to the reference counted data structure that contains the
/// actual memory buffer.
ACE_Data_Block *data_block_;
/// The allocator used to destroy ourselves when release is called
/// and create new message blocks on duplicate.
ACE_Allocator *message_block_allocator_;
private:
// = Disallow these operations for now (use <clone> instead).
ACE_Message_Block &operator= (const ACE_Message_Block &);
ACE_Message_Block (const ACE_Message_Block &);
};
/**
* @class ACE_Data_Block
*
* @brief Stores the data payload that is accessed via one or more
* ACE_Message_Block's.
*
* This data structure is reference counted to maximize
* sharing. It also contains the <locking_strategy_> (which
* protects the reference count from race conditions in
* concurrent programs) and the <allocation_strategy_> (which
* determines what memory pool is used to allocate the memory).
*/
class ACE_Export ACE_Data_Block
{
public:
// = Initialization and termination methods.
/// Default "do-nothing" constructor.
ACE_Data_Block (void);
/// Initialize.
ACE_Data_Block (size_t size,
ACE_Message_Block::ACE_Message_Type msg_type,
const char *msg_data,
ACE_Allocator *allocator_strategy,
ACE_Lock *locking_strategy,
ACE_Message_Block::Message_Flags flags,
ACE_Allocator *data_block_allocator);
/// Delete all the resources held in the message.
virtual ~ACE_Data_Block (void);
/// Get type of the message.
ACE_Message_Block::ACE_Message_Type msg_type (void) const;
/// Set type of the message.
void msg_type (ACE_Message_Block::ACE_Message_Type type);
/// Get message data pointer
char *base (void) const;
/// Set message data pointer (doesn't reallocate).
void base (char *data,
size_t size,
ACE_Message_Block::Message_Flags mflags = ACE_Message_Block::DONT_DELETE);
/// Return a pointer to 1 past the end of the allocated data in a message.
char *end (void) const;
/**
* Return a pointer to 1 past the end of the allotted data in a message.
* The allotted data may be less than allocated data if <size()> is passed
* an argument less than <capacity()>.
*/
char *mark (void) const;
// = Message size is the total amount of space allotred.
/// Get the total amount of allotted space in the message. The amount of
/// allotted space may be less than allocated space.
size_t size (void) const;
/// Set the total amount of space in the message. Returns 0 if
/// successful, else -1.
int size (size_t length);
/// Get the total amount of allocated space.
size_t capacity (void) const;
/**
* Return an exact "deep copy" of the message, i.e., create fresh
* new copies of all the Data_Blocks and continuations.
* Notice that Data_Blocks can act as "Prototypes", i.e. derived
* classes can override this method and create instances of
* themselves.
*/
virtual ACE_Data_Block *clone (ACE_Message_Block::Message_Flags mask = 0) const;
/**
* As clone above, but it does not copy the contents of the buffer,
* i.e., create a new Data_Block of the same dynamic type, with the
* same allocator, locking_strategy, and with the same amount of
* storage available (if @a max_size is zero) but the buffer is unitialized.
* If @a max_size is specified other than zero, it will be used when
* creating the new data block.
*/
virtual ACE_Data_Block *clone_nocopy (ACE_Message_Block::Message_Flags mask = 0,
size_t max_size = 0) const;
/// Return a "shallow" copy that increments our reference count by 1.
ACE_Data_Block *duplicate (void);
/**
* Decrease the shared reference count by 1. If the reference count
* is > 0 then return this; else if reference count == 0 then delete
* @c this and @a mb and return 0. Behavior is undefined if reference
* count < 0.
*/
ACE_Data_Block *release (ACE_Lock *lock = 0);
// = Message flag accessors and mutators.
/// Bitwise-or the @a more_flags into the existing message flags and
/// return the new value.
ACE_Message_Block::Message_Flags set_flags (ACE_Message_Block::Message_Flags more_flags);
/// Clear the message flag bits specified in @a less_flags and return
/// the new value.
ACE_Message_Block::Message_Flags clr_flags (ACE_Message_Block::Message_Flags less_flags);
/// Get the current message flags.
ACE_Message_Block::Message_Flags flags (void) const;
/// Obtain the allocator strategy.
ACE_Allocator *allocator_strategy (void) const;
// = The locking strategy prevents race conditions.
/// Get the locking strategy.
ACE_Lock *locking_strategy (void);
/// Set a new locking strategy and return the hold one.
ACE_Lock *locking_strategy (ACE_Lock *);
/// Dump the state of an object.
void dump (void) const;
/// Get the current reference count.
int reference_count (void) const;
/// Get the allocator used to create this object
ACE_Allocator *data_block_allocator (void) const;
protected:
/// Internal release implementation
virtual ACE_Data_Block *release_i (void);
/// Internal get the current reference count.
int reference_count_i (void) const;
/**
* Decrease the reference count, but don't delete the object.
* Returns 0 if the object should be removed.
* If @a lock is equal to the locking strategy then we assume that
* the lock is being held by the current thread; this is used to
* release all the data blocks in a chain while holding a single
* lock.
*/
friend class ACE_Message_Block;
ACE_Data_Block *release_no_delete (ACE_Lock *lock);
/// Type of message.
ACE_Message_Block::ACE_Message_Type type_;
/// Current size of message block.
size_t cur_size_;
/// Total size of buffer.
size_t max_size_;
/// Misc flags (e.g., DONT_DELETE and USER_FLAGS).
ACE_Message_Block::Message_Flags flags_;
/// Pointer To beginning of message payload.
char *base_;
// = Strategies.
/**
* Pointer to the allocator defined for this ACE_Data_Block. Note
* that this pointer is shared by all owners of this
* ACE_Data_Block.
*/
ACE_Allocator *allocator_strategy_;
/**
* Pointer to the locking strategy defined for this
* ACE_Data_Block. This is used to protect regions of code that
* access shared ACE_Data_Block state. Note that this lock is
* shared by all owners of the ACE_Data_Block's data.
*/
ACE_Lock *locking_strategy_;
/**
* Reference count for this ACE_Data_Block, which is used to avoid
* deep copies (i.e., clone()). Note that this pointer value is
* shared by all owners of the <Data_Block>'s data, i.e., all the
* ACE_Message_Blocks.
*/
int reference_count_;
/// The allocator use to destroy ourselves.
ACE_Allocator *data_block_allocator_;
private:
// = Disallow these operations.
ACE_Data_Block &operator= (const ACE_Data_Block &);
ACE_Data_Block (const ACE_Data_Block &);
};
ACE_END_VERSIONED_NAMESPACE_DECL
#if defined (__ACE_INLINE__)
#include "ace/Message_Block.inl"
#endif /* __ACE_INLINE__ */
#include "ace/Message_Block_T.h"
#include /**/ "ace/post.h"
#endif /* ACE_MESSAGE_BLOCK_H */

以上均就是所有的方法了
下面也就是分析容易混淆的函数ACE_Message_Block中有多个获取大小或者长度的函数,容易混淆.
下图是根据ACE_Message_Block(实际是ACE_Data_Block)空间的处理状况所绘,能比较清晰的反应出它们的异同.
需要注意,为了表现出多样性,下图是wr_ptr(),rd_ptr(),size()都调用过之后的情景.
红色表示是ACE_Message_Block独有的函数, 其余则ACE_Message_Block和ACE_Data_Block均有.
矩形纸上函数的返回值均为指针类型,之下的返回值均为size_t类型.

函数	说明
length()	有效数据的长度== wr_ptr() – rd_ptr()
size()	全部可用空间的长度,如果没有size()而变小,则等同capacity()== mark() – base()
space()	剩余可用空间的长度<= size() - length(),因为不含rd_ptr()移动过的空间== mark() – wr_ptr()
capacity()	最大空间的长度(ACE_Message_Block构造或初始化时所用参数值)== end() – base()
total_length()	复合消息(ACE_Message_Block内单向链 cont())的总长度
total_size()	复合消息(ACE_Message_Block内单向链 cont())的总大小
total_capacity()	复合消息(ACE_Message_Block内单向链 cont())的总空间大小

duplicate()浅拷贝函数,公用一个内部的ACE_Data_Block
ACE_Message_Block::duplicate() 与 ACE_Data_Block.duplicate()的实现是不同的.
ACE_Data_Block::duplicate()简单的只是将自身的reference加+1, 然后返回自身(this)
ACE_Message_Block:duplicate()则将自身copy了一份, 然后将自身的状态值赋给拷贝,注意它们公用同一个data_block.而且ACE_Message_Block::duplicate()支持复合消息,它会检查内部单向链,来依次调用其duplicate().ACE_Message_Block::clone()深拷贝, 不但拷贝自身,内部的ACE_Data_Block也一并拷贝了,并且支持复合消息.ACE_Data_Block.size(size_t len)函数, 动态的变化ACE_Data_Block持有的空间.
ACE_Message_Block.size(size_t len)函数是ACE_Data_Block.size(size_t len)的简单包裹.
如果len比现有的尺寸小, 简单的cur_size_ = length;
如果len比现有的尺寸大, 会申请新的空间并拷贝原所有数据.注意! 这里可能会发生空间控制权的转换! 即标志位DONT_DELETE的变化.若原ACE_Data_Block使用托管空间, 则此时会更替为自己申请的空间,从而拥有了控制权, 所以此时要注意原有空间的管理.
对ACE_Message_Block和ACE_Data_Block, 除非主动调用size(), 否则它们不会自动申请和扩大空间.ACE_Message_Block::crunch() 将现有数据移动到现有的缓冲的开始.ACE_Message_Block::reset()将现有读写指针赋为初始值(ACE_Data_Block.base())ACE_Message_Block::base()是对ACE_Data_Block.base()的简单包裹1)ACE_Message_Block的构造函数中,如果data为NULL, 则ACE_Message_Block会为其自动分配空间. 但如data非NULL,则ACE_Message_Block会直接引用data指向的空间, 并不会进行新的空间分配和拷贝.所以需要特别注意, 在ACE_Message_Block的实例没有销毁之前,不能释放data指向的空间. 2)虽然ACE_Message_Block会根据size的值来更改自己的size(),但wr_ptr不会根据data的长度进行设置, 造成length()的返回为0.需要特别注意, 当构造一个ACE_Message_Block实例后, 随之需要追加数据时,必须设置wr_ptr的值,否则原有数据将会被覆写.此时的含义是: ACE_Message_Block代管了data缓冲区,但不负责缓冲区的空间管理(因为也不是由它申请的).默认定义的flag: enum { [url=]DONT_DELETE[/url] = 01, [url=]USER_FLAGS[/url] = 0x1000 } 1) set_flags()、clr_flags()是对ACE_Message_Block中的数据指针(ACE_Data_Block*)进行设置.2) set_self_flags(),clr_self_flags()是对ACE_Message_Block本身进行设置.ACE_Message_Block::copy(const char* buf) 函数将字符串copy到ACE_Message_Block, 如果内在空间不足, 将会返回-1.需要特别注意, copy的数据将包括末尾的0, 也就是copy的数据长度为strlen(buf)+1.而且, 会自动进行wr_ptr()的设置ACE_Data_Block的析构函数是释放持有空间base_的惟一路径(size()的情况不讨论).ACE_Data_Block中通过duplicate()递增引用计数. ACE_Data_Block中通过release()递减引用计数, 当引用计数为0时,先调用ACE_Data_Block析构函数,然后释放ACE_Data_Block自身.
注意, ACE_Data_Block的构造和析构函数都不知道引用计数的存在. 在构造函数中, 只是设置了初始值1.ACE_Data_Block一个很奇怪的地方就是ACE_Data_Block::duplicate()的实现, 并没有创建新的拷贝, 而仅仅是返回了自身(return this). 这中实现方式带来了很多奇怪的问题.如下面的2,3.
release()-> release_no_delete()->release_i()->~ACE_Data_Block()如果在Stack上构造ACE_Data_Block,那么不能使用release()函数, 因为release()函数会试图删除this如果在stack上构造ACE_Data_Block, 那么不能使用duplicate()函数, 因为duplicate()返回的是this指针, 栈中的ACE_Data_Block析构后会导致问题.如果在heap上构造ACE_Data_Block,那么尽量使用release()来替代delete, 如果存在因为析构并不处理reference count, delete时不考虑其它会导致指针悬空.

更多文章，欢迎访问：http://blog.csdn.net/wallwind

此间足迹 · 发表于 2012-3-14 21:36:35

支持下辛苦了楼主

		自动登录	找回密码
密码			用户注册