This is a nice approach, and works fine with gawk and other awk implementations that use FS="" to split each input character into a field. A similar approach that doesn't depend on this behavior (which the standards say produces unspecified results) would be:

Code:

awk '
        {n = split($2, T, "[+-]")
         split($2, S, "[0-9.]+")
	 $2 = 0
         for(i = 1; i < n; i++) $2 = $2 + sprintf ("%f", S[i+1]"1" * T[i])
        }
1
' FS="|" OFS="|" file

which also has a little feature creep allowing a decimal point to be included in the input numbers (sparked by RudiC's code using %f instead of %d in the sprintf() format string).

If you change:

Code:

         gsub (/[0-9]*/, _, $2)

in RudiC's script to:

Code:

         gsub (/[0-9.]*/, _, $2)

his script will also handle floating point values in awk versions where:

Code:

         split ($2, S, "")

doesn't produce a syntax error.