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
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
|
<!-- $Id$ -->
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="GENERATOR" CONTENT="Mozilla/4.04 [en] (X11; I; Linux 2.0.32 i486) [Netscape]">
<META NAME="Author" CONTENT="James CE Johnson">
<META NAME="Description" CONTENT="A first step towards using ACE productively">
<TITLE>ACE Tutorial 007</TITLE>
</HEAD>
<BODY TEXT="#000000" BGCOLOR="#FFFFFF" LINK="#000FFF" VLINK="#FF0F0F">
<CENTER><B><FONT SIZE=+2>ACE Tutorial 007</FONT></B></CENTER>
<CENTER><B><FONT SIZE=+2>Creating a thread-pool server</FONT></B></CENTER>
<HR>
<P>As you might expect, <A HREF="client_handler.h">client_handler.h</A>
is next.
<P>
<HR WIDTH="100%">
<PRE>
<font color=red>// $Id$</font>
<font color=blue>#ifndef</font> <font color=purple>CLIENT_HANDLER_H</font>
<font color=blue>#define</font> <font color=purple>CLIENT_HANDLER_H</font>
<font color=red>/* Our client handler must exist somewhere in the ACE_Event_Handler
object hierarchy. This is a requirement of the ACE_Reactor because
it maintains ACE_Event_Handler pointers for each registered event
handler. You could derive our Client_Handler directly from
ACE_Event_Handler but you still have to have an ACE_SOCK_Stream for
the actually connection. With a direct derivative of
ACE_Event_Handler, you'll have to contain and maintain an
ACE_SOCK_Stream instance yourself. With ACE_Svc_Handler (which is
a derivative of ACE_Event_Handler) some of those details are
handled for you. */</font>
<font color=blue>#include</font> "<A HREF="../../../ace/Svc_Handler.h">ace/Svc_Handler.h</A>"
<font color=blue>#if !defined</font> (<font color=purple>ACE_LACKS_PRAGMA_ONCE</font>)
<font color=blue># pragma</font> <font color=purple>once</font>
<font color=blue>#endif</font> <font color=red>/* ACE_LACKS_PRAGMA_ONCE */</font>
<font color=blue>#include</font> "<A HREF="../../../ace/SOCK_Stream.h">ace/SOCK_Stream.h</A>"
class Client_Acceptor;
class Thread_Pool;
<font color=red>/* Another feature of ACE_Svc_Handler is it's ability to present the
ACE_Task<> interface as well. That's what the ACE_NULL_SYNCH
parameter below is all about. That's beyond our scope here but
we'll come back to it in the next tutorial when we start looking at
concurrency options. */</font>
class Client_Handler : public ACE_Svc_Handler <ACE_SOCK_STREAM, ACE_NULL_SYNCH>
{
public:
typedef ACE_Svc_Handler <ACE_SOCK_STREAM, ACE_NULL_SYNCH> inherited;
<font color=red>// Constructor...</font>
Client_Handler (void);
<font color=red>/* The destroy() method is our preferred method of destruction. We
could have overloaded the delete operator but that is neither easy
nor intuitive (at least to me). Instead, we provide a new method
of destruction and we make our destructor protected so that only
ourselves, our derivatives and our friends can delete us. It's a
nice compromise. */</font>
void destroy (void);
<font color=red>/* Most ACE objects have an open() method. That's how you make them
ready to do work. ACE_Event_Handler has a virtual open() method
which allows us to create this overrride. ACE_Acceptor<> will
invoke this method after creating a new Client_Handler when a
client connects. Notice that the parameter to open() is a void*.
It just so happens that the pointer points to the acceptor which
created us. You would like for the parameter to be an
ACE_Acceptor<>* but since ACE_Event_Handler is generic, that would
tie it too closely to the ACE_Acceptor<> set of objects. In our
definition of open() you'll see how we get around that. */</font>
int open (void *acceptor);
<font color=red>/* When an ACE_Task<> object falls out of the svc() method, the
framework will call the close() method. That's where we want to
cleanup ourselves if we're running in either thread-per-connection
or thread-pool mode. */</font>
int close (u_long flags = 0);
<font color=red>/* When there is activity on a registered handler, the
handle_input() method of the handler will be invoked. If that
method returns an error code (eg -- -1) then the reactor will
invoke handle_close() to allow the object to clean itself
up. Since an event handler can be registered for more than one
type of callback, the callback mask is provided to inform
handle_close() exactly which method failed. That way, you don't
have to maintain state information between your handle_* method
calls. The <handle> parameter is explained below... As a
side-effect, the reactor will also invoke remove_handler() for the
object on the mask that caused the -1 return. This means that we
don't have to do that ourselves! */</font>
int handle_close (ACE_HANDLE handle,
ACE_Reactor_Mask mask);
<font color=red>/* When we register with the reactor, we're going to tell it that we
want to be notified of READ events. When the reactor sees that
there is read activity for us, our handle_input() will be
invoked. The _handleg provided is the handle (file descriptor in
Unix) of the actual connection causing the activity. Since we're
derived from ACE_Svc_Handler<> and it maintains it's own peer
(ACE_SOCK_Stream) object, this is redundant for us. However, if
we had been derived directly from ACE_Event_Handler, we may have
chosen not to contain the peer. In that case, the <handle> would
be important to us for reading the client's data. */</font>
int handle_input (ACE_HANDLE handle);
protected:
<font color=red>/* If the Client_Acceptor which created us has chosen a
thread-per-connection strategy then our open() method will
activate us into a dedicate thread. The svc() method will then
execute in that thread performing some of the functions we used to
leave up to the reactor. */</font>
int svc (void);
<font color=red>/* This has nothing at all to do with ACE. I've added this here as
a worker function which I will call from handle_input(). That
allows me to introduce concurrencly in later tutorials with a no
changes to the worker function. You can think of process() as
application-level code and everything elase as
application-framework code. */</font>
int process (char *rdbuf, int rdbuf_len);
<font color=red>/* We don't really do anything in our destructor but we've declared
it to be protected to prevent casual deletion of this object. As
I said above, I really would prefer that everyone goes through the
destroy() method to get rid of us. */</font>
~Client_Handler (void);
<font color=red>/* When we get to the definition of Client_Handler we'll see that
there are several places where we go back to the Client_Acceptor
for information. It is generally a good idea to do that through
an accesor rather than using the member variable directly. */</font>
Client_Acceptor *client_acceptor (void)
{
return this->client_acceptor_;
}
<font color=red>/* And since you shouldn't access a member variable directly,
neither should you set (mutate) it. Although it might seem silly
to do it this way, you'll thank yourself for it later. */</font>
void client_acceptor (Client_Acceptor *_client_acceptor)
{
this->client_acceptor_ = _client_acceptor;
}
<font color=red>/* The concurrency() accessor tells us the current concurrency
strategy. It actually queries the Client_Acceptor for it but by
having the accessor in place, we could change our implementation
without affecting everything that needs to know. */</font>
int concurrency (void);
<font color=red>/* Likewise for access to the Thread_Pool that we belong to. */</font>
Thread_Pool * thread_pool (void);
Client_Acceptor *client_acceptor_;
<font color=red>/* For some reason I didn't create accessor/mutator methods for
this. So much for consistency....
This variable is used to remember the thread in which we were
created: the "<font color=green>creator</font>" thread in other words. handle_input()
needs to know if it is operating in the main reactor thread (which
is the one that created us) or if it is operating in one of the
thread pool threads. More on this when we get to handle_input(). */</font>
ACE_thread_t creator_;
};
<font color=blue>#endif</font> <font color=red>/* CLIENT_HANDLER_H */</font>
</PRE>
<HR WIDTH="100%">
<P>Still, we're just not seeing a lot of changes due to introduction of
the thread pool. That's a good thing! You don't want to go turning
your application upside down just because you changed thread models.
<P><HR WIDTH="100%">
<CENTER>[<A HREF="../online-tutorials.html">Tutorial Index</A>] [<A HREF="page06.html">Continue This Tutorial</A>]</CENTER>
|
