Understanding the Unix file descriptors

Question

From Beej's socket programming book:

When Unix programs do any sort of I/O, they do it by reading or writing to a file descriptor. A file descriptor is simply an integer associated with an open file.

Does it say that programs read and "write" to a "integer"? If yes, how is it possible and what does it mean?

Answer 1

It means that in order to read from a file or write to it, you need to perform a system call, and your file descriptor (the integer) is a parameter you use to tell the kernel which file you are talking about. The fact that they are integers is relatively uninteresting, they could be anything; you only ever use them transparently, that is, you open a file, store the descriptor, then you pass the descriptor back to the kernel for any operations on that file. Its value is only interesting to the kernel. One notable exception are the special file descriptors 0, 1, and 2 (stdin, stdout, and stderr), but even for those, you usually use the predefined constants rather than literal integer values.

Answer 2

I strugled with file descriptors and beejs book too. The moment of enlightement was when I understood how c dup function works.

PS:This is late answer but maybe it can help someone

Answer 3

It's saying that file descriptors are implemented as an integer. That is each open file is given a unique way of referencing it, a file handle which is actually just a unique number.

When you do:

int fd = open("filename", flags);

fd is an integer, which is returned by the implementation. It's unique to the file you opened within your program and given to you so that later you can refer to that same file, e.g. for read/write/close etc. It's nothing more than a token used to associate a sequence of operations.