1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
|
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Author" CONTENT="James CE Johnson">
<TITLE>ACE Tutorial 010</TITLE>
</HEAD>
<BODY TEXT="#000000" BGCOLOR="#FFFFFF" LINK="#000FFF" VLINK="#FF0F0F">
<CENTER><B><FONT SIZE=+2>ACE Tutorial 010</FONT></B></CENTER>
<CENTER><B><FONT SIZE=+2>Passing chunks of data through an ACE_Message_Queue</FONT></B></CENTER>
<P>
<HR WIDTH="100%">
We'll look first at <A HREF="message_queue.cpp">main()</A>.
<P>
<HR WIDTH="100%">
<PRE>
<font color=red>// $Id$</font>
<font color=red>/* To illustrate the ACE_Message_Queue, we use a derivative of
ACE_Task<>. We also derive from ACE_Message_Block to show that we
don't have memory leaks. */</font>
<font color=blue>#include</font> "<font color=green>task.h</font>"
<font color=blue>#include</font> "<font color=green>block.h</font>"
int
run_test (int iterations,
int threads)
{
<font color=red>/* Create and start an instance of our Task object. */</font>
Task task (threads);
if (task.open () == -1)
ACE_ERROR_RETURN ((LM_ERROR,
"<font color=green>%p\n</font>",
"<font color=green>open</font>"),
-1);
<font color=red>/* Give the threads a moment to open. This isn't really necessary
but if we don't we find that all of our blocks are constructed and
enqueued before any of the threads get created. Basically, the
sleep() makes the output look more interesting. */</font>
<font color=#008888>ACE_OS::sleep</font> (ACE_Time_Value (1));
int i;
for (i = 0; i < iterations; ++i)
{
<font color=red>/* Create a new message block to hold our data. Here, we ask
for a block that has 128 bytes of data space. */</font>
Block *message;
ACE_NEW_RETURN (message,
Block (128),
-1);
<font color=red>/* Grab the "<font color=green>write pointer</font>". This is a pointer into the data
area where we can write our data. After writting the data you
have to increment the wr_ptr() so that subsequent writes won't
clobber what you've put there. */</font>
<font color=#008888>ACE_OS::sprintf</font> (message->wr_ptr (),
"<font color=green>This is message %d.</font>",
i);
message->wr_ptr (<font color=#008888>ACE_OS::strlen</font> (message->rd_ptr ()));
<font color=red>/* Put the message block into the queue. One of the threads in
the Task object will pick up the block and "<font color=green>do work</font>" on it. */</font>
if (task.putq (message) == -1)
break;
}
<font color=red>/* Once we're done, we have to signal the Task objects to shut
down. There are several choices including: - Send a message of
zero length - Send a message with a special content I don't like
these choices because they're likely to interfere with application
logic. Instead, I use the message type feature to send a message
of type "<font color=green>hangup</font>". The default type is MB_DATA, so when the tasks
get a MB_HANGUP type, they know to go away. */</font>
Block *message;
ACE_NEW_RETURN (message,
Block (),
-1);
message->msg_type (<font color=#008888>ACE_Message_Block::MB_HANGUP</font>);
task.putq (message);
<font color=red>/* Wait for the threads in our task object to go away. */</font>
task.wait ();
return 0;
}
int
main (int argc, char *argv[])
{
<font color=red>/* Set the number of iterations through our putq() loop and the
number of threads to use in our Task<> derivative. */</font>
int iterations = argc > 1 ? atoi (argv[1]) : 9;
int threads = argc > 2 ? atoi (argv[2]) : 2;
run_test (iterations, threads);
ACE_DEBUG ((LM_DEBUG,
"<font color=green>(%P|%t) Application exiting\n</font>"));
return 0;
}
</PRE>
<HR WIDTH="100%">
<P>
This looks a lot like our thread-pool server and it even does some things
better. In particular, I've scoped the Task object so that it's destructor
will have a chance to get called before the application exits.
Notice how we write actual data into the message block though. In the thread-pool
server we just provided a pointer. Writting the data is actually a more correct
way of doing things since you don't get into strange pointer casting situations.
What if you want to put complex objects into the message block though? We'll do
that in the next tutorial, let's stick with the basics first.
<P>
On the next page we'll take a look at our Block object...
<P>
<P><HR WIDTH="100%">
<CENTER>[<A HREF="../online-tutorials.html">Tutorial Index</A>] [<A HREF="page03.html">Continue This Tutorial</A>]</CENTER>
|
