sed regular expression

[Pat Denton]

I have a value aaa.c11234.001.xml and I am needing to use the sed command to pull out the x11234 and 001 as individule variables. So far, I am getting the 001 with the following command in my script

var1=`echo aaa.c11234.001.xml | sed -e 's/^.*\.\(.*\)\..*$/\1/'`

But I am having issues with getting the c11234 part. I can't seem to nail down the regular expression to do so. Any help would be greatly appreciated.

[Ernest Friedman-Hill]

Well, one option (deliberately similar to yours)

var2=`echo aaa.c11234.001.xml | sed -e 's/[^.]*\\.\([^.]*\)\\..*$/\1/'`

The regexp is (with group 1 parenthesized) "Any number of not-a-dot characters, followed by a dot, (followed by any number of not-a-dot characters), followed by a dot, followed by anything."

[Pat Denton]

Thanks Ernest. That did the trick. And it helped me with some others I had to create as well. The best part was the worded explination of the regex. That made difference in me understanding it.
[ November 09, 2005: Message edited by: Pat Denton ]

[Harald Kirsch]

Why use sed?

Assuming you use (ba)sh, try this:

% a=aaa.c11234.001.xml 
% a="${a%.*}" ; echo $a
aaa.c11234.001
% b="${a##*.}" ; echo $b
001
% a="${a%.*}" ; echo $a
aaa.c11234
% b="${a##*.}" ; echo $b
c11234

For the details of this percent and hash business see:
http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#tag_02_06_02
[ November 09, 2005: Message edited by: Harald Kirsch ]

[Stefan Wagner]

Regexpressions are often unreadable one-way-programming.
You may write them, but try to read them!

A little help is the possibility to combine multiple statements with semicolons: "sed 's1;s2'".
So you could snip away the text 'aaa.c' and '.xml' with those two commands:

 
snipped=$(echo aaa.c11234.001.xml | sed 's/[^0-9]*//;s/.xml//')
# and then split it 
part1=$(echo $snipped | sed 's/.*\.//g')
part2=$(echo $snipped | sed 's/\..*//g')

By replacing `foo` with $(foo), nesting is more easy. Using backticks is discouraged for the bash for that reason.

[Ernest Friedman-Hill]

Feh. Regular expressions can be ugly, so comments certainly help. One thing in their favor is if the shell script gets turned into a Perl, Ruby, or Python program, or recoded in Java, the regexp can go along for a ride; the other techniques would require a total rewrite.