Move a range of words to the end of the string

Question

I have the following string:

"5  S comment=whatever Unspecific text fruit=apple animal=mouse sky=blue"

I need to move to end of the string, all that begin with comment to the word text.

The start is always be comment= and the delimiter is always fruit. The value between comment=... and fruit delimiter is variable.

any suggestion?

Can there be more than one comment= and/or fruit in that string? Do you need that process for every line of a file, or just in a string stored in a shell variable? Can that string contain newline characters? — Stéphane Chazelas
– Stéphane Chazelas, Commented May 27, 2015 at 14:57
Yes, there may be many lines in the file. And only one field with comment= for each line. Also, the end of the string can be variable, this end could be sky=blue or raze=human, or whatever word, is unspecific. and i using bash . Thanks for reply. — Teb
– Teb, Commented May 27, 2015 at 15:28

Stéphane Chazelas · Accepted Answer · 2015-05-27 15:22:43Z

If that string is in a shell variable, and that shell is zsh, you can do:

string='5  S comment=whatever Unspecific text fruit=apple animal=mouse sky=blue'
new_string=${string/(#b)(comment=*)(fruit*)/$match[2] $match[1]}

If that shell is ksh93:

new_strong=${string//@(comment=*)@(fruit*)/\2 \1}

If that shell is a POSIX shell (like those two, bash or the sh of modern systems):

a=${string%%comment=*}
b=${string#"$a"}
b=${b%fruit*}
c=${string#"$a$b"}
new_string=$a$c$b

For each line of a file:

sed 's/\(comment=.*\)\(fruit.*\)/\2 \1/' < file

All three would match the part from the left-most comment= to the rightmost fruit if there were more than one per line/string.

1 Answer 1