I would like to pipe standard output of a program while keeping it on screen.
With a simple example (echo
use here is just for illustration purpose) :
$ echo 'ee' | foo
ee
<- the output I would like to see
I know tee could copy stdout to file but that's not what I want.
$ echo 'ee' | tee output.txt | foo
I tried
$ echo 'ee' | tee /dev/stdout | foo
but it does not work since tee output to /dev/stdout
is piped to foo
echo 'ee' | tee /dev/stderr
works though, so if your "on screen" requirement is satisfied by stderr too, that'll do.
Here is a solution that works at on any Unix / Linux implementation, assuming it cares to follow the POSIX
standard. It works on some non Unix environments like cygwin
too.
echo 'ee' | tee /dev/tty | foo
Reference: The Open Group Base Specifications Issue 7 IEEE Std 1003.1, 2013 Edition, §10.1:
/dev/tty Associated with the process group of that process, if any. It is useful for programs or shell procedures that wish to be sure of writing messages to or reading data from the terminal no matter how output has been redirected. It can also be used for applications that demand the name of a file for output, when typed output is desired and it is tiresome to find out what terminal is currently in use. In each process, a synonym for the controlling terminal
Some environments like Google Colab have been reported not to implement /dev/tty
while still having their tty
command returning a usable device. Here is a workaround:
tty=$(tty)
echo 'ee' | tee $tty | foo
or with an ancient Bourne shell:
tty=`tty`
echo 'ee' | tee $tty | foo
Another thing to try is:
echo 'ee' | tee >(foo)
The >(foo)
is a process substitution.
Edit: To make a bit clearer, (.) here start a new child process to the current terminal, where the output is being redirected to.
echo ee | tee >(wc | grep 1)
# ^^^^^^^^^^^^^^ => child process
Except that any variable declarations/changes in child process do not reflect in the parent, there is very few of concern with regard to running commands in a child process.
foo
will have to be part of the process substitution. Here's an example: echo 'ee' | tee file.txt >(wc -c | tr -d ' ')
echo 'ee' | tee >(cat) | foo | bar
.
foo
sees 2x the ee
and none is printed on terminal. Try echo 'ee' | tee >(cat) | grep .
vs. echo 'ee' | tee >(cat) | grep x
.
Try:
$ echo 'ee' | tee /dev/stderr | foo
If using stderr is an option, of course.
Access to "/dev/stdout" is denied on some systems, but access to the user terminal is given by "/dev/tty". Using "wc" for "foo", the above examples work OK (on linux, OSX, etc.) as:
% echo 'Hi' | tee /dev/tty | wc Hi 1 1 3
To add a count at the bottom of a list of matching files, I use something like:
% ls [A-J]* | tee /dev/tty | wc -l
To avoid having to remember all this, I define aliases:
% alias t tee /dev/tty
% alias wcl wc -l
so that I can simply say:
% ls [A-J]* | t | wcl
POSTSCRIPT: For the younger set, who might titter at its pronunciation as "titty", I might add that "tty" was once the common abbreviation for a "teletype" terminal, which used a roll of yellow paper and had round keys that often stuck.
first you need to figure out the terminal associated with your screen (or whichever screen you want the output to display on):
tty
then you can tee the output to that terminal and pipe the other copy through your foo program:
echo ee | tee /dev/pty/2 | foo
t
is useless. You can use echo ee | tee $(tty) | foo
but it still has a useless command (tty
), given the fact /dev/tty
just works.
Success story sharing
/dev/tty
is a mandatory Unix device. Are you running in a BSD jail?tee
like this:cat some.log | tee /dev/tty | tee -a other.log | grep -i 'foo' >> foo.log
to 1) get it all to the console, 2) get it all appended to another file, 3) get thefoo
lines to a different file.