Shell Programming and Scripting

Hello Alister,
Neither of the "sed's" working. The above one is giving "infinite number of xyz's in output" and I have to close the terminal to come out.

The sed with -f option has thrown the below error:

Code:

/home/ravi>sed -f sed_one rem
sed: /xyz/! d is not a recognized function.

Here , sed_one does have below code:

Code:

 
/home/ravi/>cat sed_one
#n
/xyz/! d
:top
H
/xyz/ {
    s/.*//
    x
    s/\n//
    p
}
n
b top

Regards
Ravi

Removed : the code was containing an error of logic...

That one should be better :

Code:

awk '{A[NR]=$0}/one/{m=x?m:NR;x=NR}END{while(m<=x)print A[m++]}' input

Strange, this:

Code:

awk '/one/{f=1;x=NR}f{A[++c]=$0}END{while(++d<x)print A[d]}' input

works fine on a HP-UX machine but not on my Debian box with gawk

@Franklin

Ah ... forget my code ... there is an error of logic in it (confusing "line number of 'one' pattern" and "total number of line that match the requirement"
(i have just removed that erroneous code from my previous post)

In fact it is because the first occurrence of "one" in my tst file was appearing at the second line so the dift with final NR of the "one" occurrence was -1 ...
NR or NR-1 (x or x-1) doesn't necessary match the number of wanted lines !!!

That code should be fine :

Code:

awk '{A[NR]=$0}/one/{m=x?m:NR;x=NR}END{while(m<=x)print A[m++]}' input

Hmmm. I know it sounds like the easy way out, but I suspect that your system's sed is broken (at least with respect to POSIX). This is not a complicated sed script, it uses standard syntax, and it works fine on the three different implementations that I tested. I don't have access to a HP-UX system, so I cannot take a deeper look.

Regards,
Alister