Shell Programming and Scripting

Instead of searching for the delimiters do it the other way round: Search for what you are interested in, everything else would be a delimiter.

With this in mind, suppose, for example, the text you are looking for are always letters (or letters and digits, or only uppercase letters, etc.). In this case you could use the following line:

# sed 's/^\([A-Za-z][A-Za-z]*\).*$/\1/' <source> > <target>

This collects every letter (uppercase or lowercase) starting at the beginning of the line to the sed variable "\1" until a non-letter is reached. The remainder of the line is snipped off then and the variables content put out.

If the text you are looking for can be further limited change the char classes in the regular expression above. For example, if you only look for uppercase letters, you could write "[A-Z][A-Z]*" above.

bakunin

Thanks bakunin........... It worked

Anyway, can anyone suggest me the awk version on SUN OS

Very Interesting solution...I never thought one can collect the fields into a variable using sed. Can someone explain it further.Will the above solution work if spaces is the delimiter.
Also which solution is the better one to follow awk or sed??
Thanks,
srikanth

I know you got a working solution.

Just got access to an SunOS and continuing on the first suggested solution, this will work

Code:

/usr/xpg4/bin/awk -F"[*~]" '{ printf $1 "\n" }' kum.txt

Vino

1. The preferable solution is sed of course, since it uses less system resources.

2. Of course it will work if the delimiters are spaces, since the sed script is not relying on the delimiters but the content. Delimiter is anything not part of the content, so to say.

3. In case of a specified single delimiter char (including blanks) the most efficient solution would be to use "cut", like in "cat file | cut -d'<delimiterchar>' -f1"

4. The given awk solution is correct AFAICS, I don't know why it doesn't work on kumariaks system

5. Last point: How do sed variables work:

sed variables are named \1...\9 and are defined by enclosing parts of the search string by \(...\). They can even be nested. To fully understand the effect the best thing is to try it with some short file and some different regexps. For instance, a script exchanging the first and the second word in every line ("word" meaning something delimited by a space) the syntax would be:

s/\([^ ][^ ]*\) \([^ ][^ ]*\]/\2\1/

everything matched by the expression between the first \(...\) goes to \1 in the result, everything matched by the expression enclosed in the second \(...\) pair goes to \2. In the replacement expression the are simply exchanged.

bakunin

If you are dealing with EDI data, the sed solution is not complete and will not work. If I may ask what EDI software are you using?

Try this:

Code:

awk -F"$(head -1 test.txt | cut -c4)" '{print $1}' test.txt

where test.txt is the name of the input file. (Better to supply it in a variable, but I just did this quick for the purpose of showing it.) The problem isn't knowing ALL of the field delimiters you're dealing with, it's knowing which one you're dealing with...

Input file:
ISA*123
GS*456
ST*789

Output:
ISA
GS
ST

I tried it with "~", "*", and "@".

djp