The standards say that sed uses Basic Regular Expresions. In BREs [0-9]\+ matches a single digit followed by a literal plus sign. Some versions of sed ignore the standard on this and treat [0-9]\+ as an RE that will match one or more digits. The way to match one or more digits portably using standard BRE syntax is [0-9][0-9]* (which matches a single digit followed by zero or more digits). And, you're right in noting that we both missed the -i option the OP used. Looking at what was requested by the OP again, it might have been better to suggest:

Code:

sed -i 's/^a[1-9][0-9]*/aw/'

since it doesn't look like a leading zero should be allowed in the string of digits following the a or A at the start of the line.

Note also that ^a[123]\([0-9]\{1,3\}\)\?\+ (on systems that accept that form of RE) wouldn't do what was wanted even if it was a valid BRE): it precludes a4 through a9, a40 through a99, and a400 through a999 which, presumably, would have been wanted as part of the vertical elipsis between the lines starting with a3 and a3000.

If the OP wanted a leading a or A followed by any string of alphanumerics ending with a digit, that would be:

Code:

sed -i 's/^a[[:alnum:]]*[0-9]/aw/'

or:

Code:

sed -i 's/^a[[:alnum:]]*[0-9] /aw /'

And, of course, both of us are still assuming that the g at the end of the sed substitute command should be ignored and that the anchor at the start of the RE was intended to take precedence over the g flag at the end.