Ttdfjt

In a ksh shell script I am using a grep command to get a specific word as shown below.

$ cat file.txt
abc xyzdef.123 def.jkl mnopqrst

$ grep -o "wdefw" file.txt
xyzdef
def

I want output to be xyzdef.123 and def.jkl

It is not fetching the value after . Is there any other way to grep this word also I don't know the exact word to grep only I know a pattern. I am working on ksh shell.

Looks like you just want the string def and all non-whitespace characters around it. If so, you can use:

$ grep -Eo 'S*defS*' file.txt 
xyzdef.123
def.jkl

The S means non-whitespace and is supported by GNU grep with either the -E or -P flags.

Using grep -o and POSIX character classes:

grep -o '[^[:blank:]]*def[^[:blank:]]*' file.txt

This is essentially what terdon suggests, albeit using slightly different syntax (and no -E). This would match the string def and any non-blank characters on either side of that string (a non-blank character is a character that is not a space or a tab).

Alternatively,

tr '[:blank:]' 'n' <file | grep -F 'def'

This just breaks the line down into multiple lines, one line per blank-separated word (where a "blank" is a tab or a space character). Then it applies a plain string match with grep -F on the generated lines to find the ones that you're interested in.

Your pattern, wdefw, which with GNU grep, if using POSIX character class names, is the same as [[:alnum:]_]def[[:alnum:]_], requires that the string def is flanked by a alphanumeric character or underscore on either side. A dot is neither an alphanumeric character nor an underscore.

The pattern also would not match def if it occurs at the very start or end of a line.

With the attempt you have, you can't match the whole word to be returned. The -o flag of grep only returns the matched regex portion defined. Also w is not a POSIX defined extension for grep and might be available only in the GNU versions which support the PCRE syntax. On which you could do

grep -oP '(w*)def[.](w*)'

The -P flag turns on the PCRE regex mode in GNU grep and the -o flag returns the whole word matched the regex defined. The regex is translated as match zero or more number of alphanumeric characters followed by def and a literal . ( enclosed in a bracket expression ) and followed by zero or more number of alphanumeric characters.

Using POSIX character classes for alphanumeric characters would be doing below. But remember the flag -o is still a GNU extension

grep -o '([[:alnum:]]*)def[.]([[:alnum:]]*)'