Insane redirection behavior

Shell Programming and Scripting
1

Hi guys,

I know computers don't misbehave.
But I'm puzzled by what's happening right know in a script :

I simplified the example to point out what seems weird to me.
Don't try to find any sense to this stupid script.

There are 10 rows in /tmp/tmp.txt

i=0
tmpfile=/tmp/tmp.txt
while read -r; do
    echo "$REPLY" > /dev/null
    sed "1,$((++i))d" $tmpfile | wc -l
done < $tmpfile

i=0
tmpfile=/tmp/tmp.txt
while read -r; do
    echo "$REPLY" > /dev/null
    sed "1,$((++i))d" $tmpfile > /tmp/output; wc -l /tmp/output
done < $tmpfile

In the first script, I count line from the output of sed. In the second one, I first output sed to a file and then count the lines of that file.
I see no reason why the output would be different... YET:

root@debian:~# ./script1
9
9
9
9
9
9
9
9
9
9
root@debian:~# ./script2
9 /tmp/output
8 /tmp/output
7 /tmp/output
6 /tmp/output
5 /tmp/output
4 /tmp/output
3 /tmp/output
2 /tmp/output
1 /tmp/output
0 /tmp/output

This is driving me crazy.
Do you have any sane explanation?!?

Regards
Santiago

2

It should be easier if you also post the input file and the desired output.

3

The first script is running sed in a subshell environment; the second is not. The subshell is a byproduct of the shell's pipeline implementation. The difference is that when $i is incremented in a subshell, the new value is not available in the parent shell.

One workaround would be to increment the variable prior to calling sed.

Regards,
Alister

2 Likes
5

The input file is random. It just contains 10 rows but we don't care what they are.
The output should be counting down from 9 to 0.

---------- Post updated at 16:40 ---------- Previous update was at 16:23 ----------

Hi alister,

You were right.
Here are the two new scripts...

i=0
tmpfile=/tmp/tmp.txt
while read -r; do
    echo "$REPLY" > /dev/null
    ((++i))
   sed "1,${i}d" $tmpfile | wc -l
done < $tmpfile

i=0
tmpfile=/tmp/tmp.txt
while read -r; do
    echo "$REPLY" > /dev/null
    ((++i))
    sed "1,${i}d" $tmpfile > /tmp/output; wc -l /tmp/output
done < $tmpfile

... and their output...

root@debian:~# ./script1
9
8
7
6
5
4
3
2
1
0
root@debian:~# ./script2
9 /tmp/output
8 /tmp/output
7 /tmp/output
6 /tmp/output
5 /tmp/output
4 /tmp/output
3 /tmp/output
2 /tmp/output
1 /tmp/output
0 /tmp/output

Regards
Santiago