Description

For simplicity, we assume that a word is a sequence of lowercase alphabets. You can extend the regular expressions used in the scripts to match POSIX-defined words.

The job to do is reversing every word of a line.

Raw Input		Desired Output
112358 is a fibonacci sequence... a test line 124816 1392781 final line...		112358 si a iccanobif ecneuqes... a tset enil 124816 1392781 lanif enil...

Script and Comments

Script₁

[ 1] s/[a-z][a-z]*/\n&/g
[ 2] :loop
[ 3] s/([a-z]*\n)([a-z])([a-z]*)/\2\1\3/g
[ 4] t loop
[ 5] s/\n//g

Comments

The `-r' option of GNU sed must be used or you have to escape every parenthesis.
Step [1] is used to insert a newline character before every word of a line.
Steps [2] thru [4] constitute a loop where each iteration will move the character after the newline character of every word to the beginning of it.
For example, the process of reversing `abcd' is as follows: abcd \nabcd a\nbcd ba\ncd cba\nd dcba\n