mjhjmtu

I am trying to figure out how to replace the '/' in a date (20/01/1990) with '.' (20.01.1990). I do understand to do the whole file is 's///./g', but there are urls in the file so it replaces those '/' with '.' which is not what I need. Is there any way to make it apply to only certain columns? I say there is a way with perl but I am working in bash

Just use a different delimiter that's not part of your line.

echo 'http://www.google.com (20/01/1990)' | sed -r 's@([0-9]+)/([0-9]+)/([0-9]+)@1.2.3@g'
http://www.google.com (20.01.1990)

bash is a shell, a command line interpreter. It's role is to run commands. sed and perl are two commands that bash or any other shell can run.

Both happen to be interpreters of some programming language, but that's not relevant here. Both are great at text processing as they have been designed for that. sed is a standard command but there are many different implementations which are incompatible if you stray outside of what POSIX specifies, perl is not but there is only one implementations (though many different versions with different feature level).

Here, using standard sed syntax, you can do:

sed 's|([[:digit:]]2)/([[:digit:]]2)/([[:digit:]]4)|1.2.3|g'

Which should work regardless of the sed implementation provided it's POSIX compliant within the limits specified by POSIX (in this case, as long as the input is valid text, which in practice excludes very long lines, sequences of bytes not forming valid characters or non-delimited lines, though some implementations like GNU sed may accept those as an extension).

With any version of perl and any input, you can also write it:

perl -pe 's|(dd)/(dd)/(d4)|$1.$2.$3|g'

Or with recent versions or perl:

perl -pe 'sdd/dd/d4$&=~yge'

If you want to avoid replacing 1000/10/99999 with 1000.10.99999, you can use word boundary operators which are available in perl, but not in standard sed (though some implementations support </> or [[:<:]]/[[:>:]] for that as an extension):

perl -pe 's|b(dd)/(dd)/(d4)b|$1.$2.$3|g'

Or look-around operators (again, a feature of perl regexps, but not standard sed regexps).

perl -pe 's|(?<!d)(dd)/(dd)/(d4)(?!d)|$1.$2.$3|g'

You can do something equivalent with standard sed with a bit of programming (sed's conditional looping construct):

sed -e :1 -e 's|^(.*[^[:digit:]])0,1([[:digit:]]2)/([[:digit:]]2)/([[:digit:]]4)([^[:digit:]].*)0,1$|12.3.45|g; t1'