Shell Programming and Scripting

You are right. The second "%" was a typo on my part, thank you for pointing it out. I corrected it in the posting above.

bakunin

Btw, Yoda, the awk did not work, it returend the same file, however the parameter expansion one is great.
(OS=hp-ux ksh )

Code:

echo "rp8-file.1.cr_cr.cr" | awk -F'.' '{NF-=1}1' OFS='.'
rp8-file.1.cr_cr.cr

Thanks, for , parameter expansion way it worked great:

Code:

Filename=rp8-file.1.cr_cr.cr
echo "${Filename%.cr}"

I would like to know how this works, I never come across this before the parameter expansion way . Thanks all for your time.

I don't know why it didn't work for awk in HP-UX.

But I know for sure that it is buggy, for example it prints FILENAME in BEGIN rule which is not correct as per the standard!!

Code:

awk 'BEGIN { print FILENAME } { print $0 }' file
file
test

You can read up on "Parameter Expansion" in the man page of your shell. This, btw., is the reason why it is so fast: it is done completely inside the shell, whereas any external program (awk, sed, tr, ...) would have to be loaded and started first.

Here are some basics, for a head start:

You can address the full content of a variable like this:

Code:

${variable}

but you can also address arbitrary parts of a variables content by using "modifiers":

Code:

${variable#pattern}
${variable##pattern}
${variable%pattern}
${variable%%pattern}

What this does is: the content of "$variable" is reduced by the pattern if it matches. "%" will remove "pattern" from the right side, "#" will remove from the left side:

Code:

variable="start.middle.end"
echo ${variable#*.}
middle.end
echo ${variable%.*}
start.middle

The pattern is "*." in the first example, and it is applied at the left side (#) of the content, whereas the pattern is ".*" in the second example and is applied at the right side. "Pattern" means a file glob in this case, like you search for files "foo.*" to find all files starting with "foo." and any extension.

If you double the "%" or "#" symbols you switch from shortest-possible match to longest-possible match. In case of a single possible match it will make no difference, but in the above case, where "*." could match two possible strings:

Code:

variable="start.middle.end"
echo ${variable##*.}
end
echo ${variable%%.*}
start

You can even nest these expressions and use the outcome of one as the pattern for another:

Code:

variable="get first 2 characters of a string regardless of length"
echo ${variable%${variable#*??}}
ge

Here, "{variable#*??}" produces the third to last character and this is used as a pattern to "${variable%<pattern>}".

Notice, that without an explicit assignment the content of "$variable" is NOT CHANGED. The expansion is just addressing different portions of it:

Code:

variable="start.middle.end"
echo ${variable#*.}
middle.end
echo ${variable}
start.middle.end

I hope this helps.

bakunin

Didn't work because:The following should work:

Code:

echo "rp8-file.1.cr_cr.cr" | awk 'BEGIN{FS=OFS="."}{NF--;$1=$1;print $0}'

but when parameter substitution does the job, why use awk?