Shell Programming and Scripting

Hi All,

Please help me with the below problem

Code:

if [ "$a" == "RAHUL" ]
then
$a=" "
fi

echo "ABC1abc" | sed 's/a/'$a'/'

The required output is :

Code:

ABC abc

But I am getting the below error:

Code:

sed: -e expression #1, char 4: unterminated `s' command

Please help.

I don't understand what you're trying to do. Why would a script that attempts to use a sed substitute command to replace the 1st occurrence of the character a with the contents of an undefined variable be expected to change the digit 1 in your sample input to a <space> character???

If you want to change the 1st occurrence of the digit 1 on each input line to the value you have assigned to the shell variable a (assuming that $a does not expand to a string containing any /, \, &, or NUL characters; you could use:

Code:

a="some string"
echo "ABC1abc" | sed "s/1/$a/"

Note that unless you quote the entire substitute command, you can't include blanks in the substitution string because the shell will see sequences of blanks in the substitute command as field separators; not as data included in the replacement string.

I have this requirement to put space as a substitution for some cases.

Isn't there any option to give space to a variable and use it in sed?
Because, I am using sed for replacing other strings in my script, including space is one of the requirement.

In the above code, I am trying to replace '1' with string a (some string / space ) depending on the input read.

Yes, there is a way to do it. Do it exactly as I showed you in post #2 in this thread.

I'd like to respectfully disagree: you can do that, but it takes a little effort with the quoting to make sure the string doesn't undergo the shells field splitting:

Code:

a=" "
print - "yxyxyxy" | sed 's/x/'"$a"'/'
y y y y

Of course all the usual problems with shell strings used as sed program(part)s persist: metacharacters, which might lead to the regexp meaning something very different from the expected, etc..

I hope this helps.

bakunin

Hi bakunin,
You are absolutely correct in saying that the entire substitute command doesn't have to be quoted (only characters subject to field splitting, pathname expansions, and other expansions that have already been discussed have to be quoted), but note that in your example:

Code:

sed 's/x/'"$a"'/'

you have the entire substitute command quoted (some in single quotes and some in double quotes, and backslashes can also be used to quote characters in strings). But, we could have written this specific command as:

Code:

sed s/x/"$a"/

(without the single quotes) and still get the same results. But, if the substitute pattern had contained bracket expressions, asterisks, question marks, or parentheses, they would have to be quoted to protect against possible pathname expansions. I don't know of any case where someone writing a script using sed needs to write an unquoted substitute command to make it work correctly. Therefore, I generally tell noobies to always quote the entire substitute command. They just have to decide which quotes are appropriate to use to get dollar signs in the pattern and in the replacement text to work as desired.

D'accord!

I understood your first quote to mean you have to enclose everything into one quotation. Personally. i prefer to enclose to biggest-possible part of my sed-scripts always into single-quotes (to not quote them at all didn't even come to my mind, even if this might be possible for a few select scripts, as you pointed out) and only the parts where i need parameter expansion i enclose in double-quotes. This results in the way i wrote it, with the alternating single- and double-quotes.

Anyway, i take it, we used different words to mean the same.

bakunin