[SOLVED] Why does `sed -e 's/[0-9]*/& &/'` have very different results with minor input changes?

Issue

I am trying to understand the below result and differences, why the result is not abc 123 123 when [0-9]* would match 123

• Why is this first answer not abc 123 123?

  $ sed -e 's/[0-9]*/& &/' <<<'abc 123'
   abc 123
  

• Why is the second result 123 123 abc?

  $ sed -e 's/[0-9]*/& &/' <<<'123 abc'
  123 123 abc
  

• Why does the third one have extra spaces?

  $ sed -e 's/[0-9]*/& &/g' <<<'abc 123'
   a b c  123 123
  

Solution

sed will use the first match. Since [0-9]* matches not just 3 digits 123 but also 0 digits (the empty string), it will match the empty string marked with parentheses here: ()abd 123 and replace it with () ()abd 123

If you instead use one of:

sed -r 's/[0-9]+/& &/'
sed 's/[0-9][0-9]*/& &/' 

then you require that it matches at least 1 digit, and then it'll behave the way you expect.

Answered By - that other guy

Answer Checked By - Katrina (WPSolving Volunteer)