Here’s an interesting bit I ran into few days ago. I got curious how is that less (or more) can read file contents from standard input and yet it is able to process input that comes from user. Both of them come from standard input, yet these are quiet heterogeneous streams of information. So, how can it be?

At first, I thought less reads entire input first. This would make standard input stream free to process key presses from the user. So I decided to check this out. I created a 1Gb long text file and ran less on it. I expected less to take some time to show file contents, but it showed first lines of the file instantly. Also, it didn’t consume 1Gb of RAM as I expected.

The conclusion is obvious. less does not read entire input buffer before letting user to interact with itself. Then how?

Here’s what less does. It separates input file stream from user input stream. Both inputs initially come via standard input stream, so less separates between them. First it duplicates the standard input stream. This allocates a new file descriptor. Then it closes old file descriptor, freeing file descriptor 0 for use. Then it opens /dev/tty. When opening a file, Linux uses next available file descriptor. File descriptor used as standard input is 0, so when less opens /dev/tty again, file descriptor of the newly opened file has value 0.

Eventually, it ends up with the new input coming via standard input stream file descriptor (0), and old input still available via file descriptor that it duplicated in the beginning. It reads the input file from the duplicated file descriptor, and uses curses on standard input stream.

You may be wondering what /dev/tty is and what it has to do with standard input streams. This is really fascinating stuff.

As you know, in Linux, everything is a file. So is terminal. Linux uses device files to represent various system devices. /dev/tty is a file that represents terminal of the current process. When process reads from /dev/tty it becomes its input. When program writes to /dev/tty it becomes its standard output or standard error stream.

So, when less reopens /dev/tty, it actually recreates standard input. Older descriptor, one that has been duplicated, can no longer accept new input, but less still can use it to read what it has been written into it already.