Simple examples that show how to do socket programming in C++. Use
make
to compile the code in each directory.
The files simple-echo/server.cc and simple-echo/client.cc contain
a basic echo server and client. The server handles only one client at
a time and simply sends back to the client whatever it receives.
Run server with client and see what
happens.
Caution: both the client and the server do not properly check the return values of send() and recv(). This means that long messages may not be echoed properly. Better code is in subsequent examples. This code is here just to show the basics.
To illustrate the poor quality of the code in the simple echo server
and client, the file simple-echo/slow-server.cc is an echo server
that sends replies only one character at a time. While this is not how
any server is written, it illustrates what can happen when sending
larger messages across the network. Those messages can be split into
smaller pieces by TCP, so the a developer cannot assume that a single
call to recv() will get the entire message.
Run slow-server with client and see what happens.
This code implements a more functional and complete echo server and client. The client sends messages terminated by a newline, and the server parses until it receives a newline. Both client and server check the return value of send() and recv() and use loops to be sure the entire message is sent or received in each case.
This code is located in the echo directory. The main programs are in
echo-server.cc and echo-client.cc. The server uses
inet-server.cc and inet-server.h to create Internet sockets using
IPv4 addresses. These files in turn rely on server.cc and server.h
for the protocol logic and processing. The client uses
inet-client.cc and inet-client.h to create the client version of
an Internet socket. These files likewise rely on client.cc and
client.h for the protocol processing.
This code converts the echo client and server to use UNIX sockets
instead of Internet sockets. UNIX sockets use the file system to identify
the socket, rather than IP addresses and ports. In this code, we use
/tmp/unix-socket as the name of the socket.
By using a UNIX socket, this code is restricted to communication with local processes only. In addition, it is not portable to operating systems that do not support UNIX sockets. This code is tested only on Linux, but most of it should be portable to BSD and Mac OS X, with some small changes.
This code is located in the unix-echo directory. The server uses
unix-echo-server.cc, unix-server.cc, and unix-server.h. These
latter two files derive a class based on socket.cc and
socket.h. The client uses unix-echo-client.cc, unix-client.cc,
and unix-client.h. Likewise, these latter two files derive a class
based on client.cc and client.h.
Note that this code reuses the same base classes in socket.h and client.h,
showing how you can swap out a different type of socket, but the protocol
logic and processing remains the same.