Circumvent the sed backreference limit \1 through \9

Question

The sed manual clearly states that the available backreferences available for the replacement string in a substitute are numbered \1 through \9. I'm trying to parse a log file that has 10 fields.

I have the regex formed for it but the tenth match (and anything after) isn't accessible.

Does anyone have an elegant way to circumvent this limitation in KSH (or any language that perhaps I can port to shell scripting)?

Answer 1

Can you user perl -pe 's/(match)(str)/$2$1/g;' in place of sed? The way to circumvent the backreference limit is to use something other than sed.

Also, I suppose you could do your substitution in two steps, but I don't know your pattern so I can't help you out with how.

Answer 2

Split the stream with -e, as long as the replaced elements are with in the group that you split them with. When I did a date split so I could re-org the date-time into a string of 14 digits, I had to split the stream up 3 times.

echo "created: 02/05/2013 16:14:49" |  sed -e 's/^\([[:alpha:]]*: \)//' -e 's/\([0-9]\{2\}\)\(\/\)\([0-9]\{2\}\)\(\/\)\([0-9]\{4\}\)\( \)/\5\1\3/' -e 's/\([0-9]\{2\}\)\(\:\)\([0-9]\{2\}\)\(\:\)\([0-9]\{2\}\)/\1\3\5/'

20130205161449

Answer 3

You're asking for a shell script solution - that means you're not limited to using just sed, correct? Most shells support arrays, so perhaps you can parse the line into a shell array variable? If need be, you could even parse the same line multiple times, extracting different bits of information on each pass.

Would that do?