What is the `tee` command in Linux?

Question

I am trying to interpret the following tee command:

cat colors.txt words.txt | tee colorsAndWords.txt | wc

Is my understanding, as follows, correct?

1. cat colors.txt words.txt: This command concatenates the contents of the colors.txt and words.txt files and sends the combined output to the standard output (the terminal).

2. | tee colorsAndWords.txt: The | (pipe) symbol takes the output of the previous command and passes it as input to the tee command. tee is used to both display the data on the standard output (usually the terminal) and write it to a file. In this case, it writes the concatenated output to a file named colorsAndWords.txt.

3. | wc: The final | wc takes the output of the tee command, which is still the concatenated content, and passes it to the wc command. wc is used to count the number of lines, words, and characters in the text it receives as input.

Answer 1

Your understanding is correct, and there is nothing that I can see that needs correcting with regard to the tee utility. It is used to duplicate a data stream, and the example you show does this, storing one copy in a single file (there could be many) while also passing it on to the next stage of the pipeline.

It's a standard utility, meaning it will work the same on any Unix system, not just Linux. Its standard specification is found here, and the implementations found on Linux and elsewhere should adhere to that description (but may add extensions, like new command line options).

Answer 2

Yes, that's correct. tee is like a T-junction in a water pipe, or a dual-output headphone jack splitter. It reads the input one and duplicates it to two output files, stdout and files specified on the command line. (Actually any number of output files; you can specify more than one on the command line. The rest of this answer assumes the simple case of one arg.)

---

Perhaps it would help to understand how you might write a simple implementation of tee.
In C, a loop body like this:

ssize_t bytes_read = read(0 /*stdin*/, buf, 4096);
write(1 /*stdout*/,          buf, bytes_read);  // error checking not shown; also
write(cmdline_arg_output_fd, buf, bytes_read);  // loops to finish short writes

The loop condition would be something like while(bytes_read > 0) to catch EOF (read returns 0) or error (negative), perhaps written like while( (bytes = read()) > 0 ) with assignment and compare inside the controlling expression. Since it needs to check the return value before doing the write calls.

Before that loop, fd = open(argv[1], O_WRONLY|O_CREAT, 0666); or something.

One fun usage is foo | bar | tee /dev/tty | baz as kind of a "debug print" to see data going through that stage of a pipeline. (Opening /dev/tty actually gives you whatever device is the controlling TTY for that process, i.e. the terminal. Like /dev/pts/19 in a Konsole terminal tab.)

Other useful args include regular files like you're using, as well as named pipes, or even process-substitution like >( grep ... > filtered.txt).

Answer 3

'|' the pipe character is a instruction to the shell that the three programs (cat,tee and wc) should be run in a very specific manner.

THEY are basically oblivious to the fact that stdout (in this case for cat and tee) is redirected to something which isnt the terminal itself.

The same for stdin for the programs tee and wc.