Entry

What is the exact purpose of fileno()?

Sep 20th, 2000 03:56
unknown unknown, Donn Cave

The file object is an I/O buffer and a device.  The C library allocates
space for the buffer in your process memory, and when you read or write
the data comes from or goes to that buffer.  The stdio C library 
functions manage the buffer by reading or writing to the device.
Since physical device access is a very expensive operation for the 
computer, buffering like that basically allows you to work with small or 
random size I/O requests while economizing on system resources.  
However, there are times when you need to do something that the stdio 
library doesn't account for, like select(), or a filesystem lock, or 
some terminal driver function.
The fileno() function exposes the file object's file descriptor or unit
number, which on UNIX is the raw device.  Select() already knows about
fileno(), so you can give it a socket or whatever and it will call 
fileno() itself, but usually it's up to you to call fileno() and get the 
descriptor.
Of course the socket fileno() function is more or less the same idea as
the file object's fileno(), but the socket isn't a file object and 
doesn't have its own buffer.  (On UNIX you can make a file object from a 
socket, but it's difficult to make that work both properly and 
efficiently with select().)  The result from socket fileno() is a file 
descriptor on UNIX, but it isn't necessarily so on other operating 
systems.
You can get file descriptors directly if you use the posix.open() 
function, (A.K.A. os.open().)