How does a Unix program know to change output if it's being piped?

Question

Calling something like ag (a grep alternative) results in an output that's human readable. Results are printed with headers. Useful for humans, horrible for programs.

$ ag 'filter\('
tests/string_lookup/tests.py
75:        self.assertSequenceEqual(Article.objects.filter(submitted_from__contains='192.0.2'), [a])
77:        self.assertEqual(Article.objects.filter(submitted_from__contains='32').count(), 0)

tests/test_runner/test_debug_sql.py
16:            Person.objects.filter(first_name='pass').count()
20:            Person.objects.filter(first_name='fail').count()

Same program, piped into something else, outputs:

$ ag 'filter\(' | cat
tests/string_lookup/tests.py:75:        self.assertSequenceEqual(Article.objects.filter(submitted_from__contains='192.0.2'), [a])
tests/string_lookup/tests.py:77:        self.assertEqual(Article.objects.filter(submitted_from__contains='32').count(), 0)
tests/test_runner/test_debug_sql.py:16:            Person.objects.filter(first_name='pass').count()
tests/test_runner/test_debug_sql.py:20:            Person.objects.filter(first_name='fail').count()

This output is much better suited for use by other programs, so I assume that ag is aware that it's being piped.

How is ag aware of where it's output is being sent?

Answer 1

The program calls isatty() on the stdout file descriptor (i.e. 1). If the function returns true, the output is connected to a tty device (a terminal).