I have the following JSON input
... "somefield":"somevalue", "time":"timevalue", "anotherfield":"value" ...
inside my KornShell (ksh) script, and I wish to replace timevalue with my value. So I created this regular expression using groups with works just fine
data=`cat somefile.json`
echo $data | perl -pe "s|(.*time\"\s*\:\s*\").*?(\".*)|\1%TIME%\2|g" | another-script.sh
... "somefield":"somevalue", "time":"%TIME%", "anotherfield":"value" ...
However ... I cannot use number as substitution because Perl uses numbers to define groups .. so this one obviously doesn’t work:
perl -pe "s|(.*time\"\s*\:\s*\").*?(\".*)|\120:00:00\2|g"
I can overcome this by doing a two-step substitution,
perl -pe "s|(.*time\"\s*\:\s*\").*?(\".*)|\1%TIME%\2|g" | perl -pe "s|%TIME%|20:00:00|"
... "somefield":"somevalue", "time":"20:00:00", "anotherfield":"value" ...
but I am sure there is a better and more elegant way to do it.
Perl doesn't use \1 for substitution. If you had enabled warnings (e.g., with perl -w), Perl would have told you it's $1. Which can be disambiguated from surrounding digits by adding { }:
perl -pe 's|(.*time"\s*:\s*").*?(".*)|${1}20:00:00$2|g'
(I also removed all the redundant backslashes from the regex.)
On another note, what's the point of matching .* if you're just going to replace it by itself? Couldn't it just be
perl -pe 's|(time"\s*:\s*").*?(")|${1}20:00:00$2|g'
?
I'm not a big fan of .* or .*?. If you're trying to match the inside of a quoted string, it would be better to be specific:
perl -pe 's|(time"\s*:\s*")[^"]*(")|${1}20:00:00$2|g'
We're not trying to validate the input string, so now there's really no reason to match that final " (and replace it by itself) either:
perl -pe 's|(time"\s*:\s*")[^"]*|${1}20:00:00|g'
If your Perl is not ancient (5.10+), you can use \K to "keep" leading parts of the string, i.e. not include it in the match:
perl -pe 's|time"\s*:\s*"\K[^"]*|20:00:00|g'
Now only the [^"]* part will be substituted, saving us from having to do any capturing.