How can I show the results of each step in a pipeline of continuous transformations without looping? - bash

Question

Consider the following example Bash one-liner in which the letters "h", "e" and "o" are removed from the word "hello" one at a time, in that order. Only the two "l" letters remain;
$ echo "hello" | tr -d h | tr -d e | tr -d o
ll
I am trying to find a method for displaying the output of each command to the screen within the one liner, so others running it can see what is going on. Continuing with the above example I would like output as follows;
$ echo "hello" | tr -d h | tr -d e | tr -d o
hello
ello
llo
ll
Is this possible? As per the operation of the one-liner above, we are carrying the output from command to command with the vertical pipe. So I assume I would have to break from the pipe to print to stdout, which would then interrupt the "command chain" I have written. Or perhaps tee can be used here, but I can't seem to achieve the desire affect. UPDATE: tee won't work because it's output is still within the boundaries of the pipe, duh!
Many thanks.

Answer 1

This only works on a terminal:
echo hello | tee /dev/tty |
tr -d h | tee /dev/tty |
tr -d e | tee /dev/tty |
tr -d o
The /dev/tty device redirects output to the current terminal, no matter where the normal output goes.

Answer 2

You can tee the output directly to the terminal:
echo hello | tee /proc/$$/fd/1 |
tr -d h | tee /proc/$$/fd/1 |
tr -d e | tee /proc/$$/fd/1 |
tr -d o
$$ is the shell's PID.

Answer 3

IMO, if this is real world problem and not how is it possible to do in bash, then instead of the bash madness starting X tr and Y tee processes, you can try a simple perl oneliner:
echo hello | perl -nale 'foreach $i (qw(h e o)) {#F=map{s/$i//g;print $_;$_}#F}'
will print:
ello
llo
ll
or if your really want bash (as #chepner suggested)
echo "hello" | while read w; do for c in h e o; do w=${w//$c};echo $w; done;done
or
while read w; do for c in h e o; do w=${w//$c};echo $w; done;done < wordlist

Answer 4

With a small little loop:
w="hello" ; for c in h e o .; do echo $w; w=$(echo $w | tr -d $c); done
The . is only used for a brief solution. After reading the question more carefully, I tested and found, that it works in a pipe chain too:
w="hello" ; for c in h e o .; do echo $w; w=$(echo $w | tr -d $c); done | less
# pure bash buildins, see chepner's comment
w="hello" ; for c in h e o .; do echo $w; w=${w/$c}; done

Answer 5

But this WILL work:
tmp=`echo "hello"`; echo $tmp; tmp=`echo $tmp | tr -d h`; echo $tmp; tmp=`echo $tmp | tr -d e`; echo $tmp; echo $C | tr -d o
It's ugly, but it works. The solution creates variables to store each pass and then show them all:
tmp=`echo "hello"`
echo $tmp
tmp=`echo $tmp | tr -d h`
echo $tmp
tmp=`echo $tmp | tr -d e`
echo $tmp
echo $C | tr -d o
I cannot find a better solution.

Answer 6

You could do it using a fifo (or a regular file) on which you listen:
mkfifo tmp.fifo
tail -f tmp.fifo
Then just run your command on a separate shell:
echo hello | tee -a tmp.fifo | tr -d e | tee -a tmp.fifo | tr -d o

Answer 7

You can use tee - as - represents the console:
echo "hello" | tee - | tr -d h | tee - | tr -d e | tee - | tr -d o
Hope that helped!
EDIT:
This solution WILL NOT WORK because stdout gets blocked and reused by all the commands that produce output by means of stdout.
I will not delete the answer as it clearly states that it doesn't solve the problem but ALSO SHOWS that this kind of solution is NOT VALID and why. If someone finds himself in the same situation will discover this information, and that will be useful.

Answer 8

Room for one more?
echo "hello" | tee >(tr -d h | tee >(tr -d e | tee >(tr -d o) ) )
or
echo "hello" |
tee >(tr -d h |
tee >(tr -d e |
tee >(tr -d o ) ) )
output is:
hello
ubuntu#ubuntu:~$ ello
llo
ll
a little persuasive coercion is required to eliminate corruption by ubuntu#ubuntu:~$
cat <(echo "hello" | tee >(tr -d h | tee >(tr -d e | tee >(tr -d o) ) ) )
output is:
hello
ello
llo
ll