UNIX for Dummies Questions & Answers

The behaviour of GNU sed is the more "strictly" correct :

first match : abc 123

second match : \([a-z]*\) \([0-9]*\) means :
one string made of 0 or more characher within range a to z
one space
one string made of 0 or more digit within range 0 to 9

Code:

"<empty string><space><empty string>" = " "

So it match "the string made of 0 caracter followed by space followed by another 0 length digit string. " Which in fact means "space"
So the first space that comes just after the first pattern matching... match the second pattern matching ... that is why you get this result.

So It looks like the implementation of Sun sed and GNU sed differs ( this case demonstrate this)

So to get the same result, you should (just for the example) to force one of the string to match at least one caracter so it can not match empty string:

Code:

# echo  abc 123 def 456  |  sed 's|\([a-z]*\) \([0-9][0-9]*\)|\1|2'
abc 123 def
# echo  abc 123 def 456  |  sed 's|\([a-z][a-z]*\) \([0-9]*\)|\1|2'
abc 123 def