Shell Programming and Scripting

Hi, I'm a newbie to sed and learning about sed 1 liners and captures. I was wondering if someone could help me with the code below:



So that's taking a URL and splitting it into URL,TLD & Filename. Any help on how I can do this is much appreciated.

Apologies for posting an image. I don't have enough post count to insert a URL.

Code:

echo "www.google.com/sample.log" |awk -F \/ '{print $0,$1,$2}'

nice awk solution but OP asked for sed:

Code:

echo "www.google.com/sample.log" |sed -r 's/(.*)\/(.*)/\1\/\2 \1 \2/g'

Explaination:

Breaking this sed line down we have 4 main parts.

1. The sed command and it's flags
- The -r tells sed to use regex-extended matching.

2. The search pattern
- (.*)\/(.*)
- \/ is the escaped form of /
- We escape the '/' since sed sees '/' as a search pattern delimiter

3. The replace pattern
- \1\/\2 \1 \2
- \/ is the escaped form of /
- We escape the '/' since sed sees '/' as a search pattern delimiter

4. The sed commands
- 's' at the start
- This tells sed that we are searching and replacing
- The character following s used as the pattern delimiter
- You can make the delimiter what ever you like (|, /, :, ^, etc...)
- 'g' at the end
- This tells sed to match more than one occurrence per line if found
- If not used and the line has multiple occurrences of pattern only 1 is matched and replaced

The (.*)\/(.*) says match everything but keep the parts before and after the / (which is escaped as \/) for later use.

In the replace pattern of the search we use \1 and \2 several times along with the escaped / (\/) again.

\1 is the first item we kept from the search (everything to the left of '/' or www.google.com)
\2 is the second part we kept (everything to the right of '/' or sample.log) .

So '\1\/\2 \1 \2' becomes 'www.google.com/sample.log www.google.com sample.log'

Beautiful. Thank you both & special thanks to ddreggors for the wonderful explanation.

Like ddreggors have mentioned we can use any other character as a delimiter instead of "/" in sed. This will help in avoiding confusion if we have "/" in the string we are using. Below is a slightly modified code using "!" as delimiter.

Code:

#using !
echo "www.google.com/sample.log" | sed 's!\(.*\)/\(.*\)!\1 \2!'

#using =
echo "www.google.com/sample.log" | sed 's=\(.*\)/\(.*\)=\1 \2='

regards,
Ahamed

Also note that Ahamed did not use a 'g' at the end of the regex search/replace. In this situation (mine included) the g is not needed.

Here is why:

With a match like we have '(.*)\/(.*)' we would match the whole line. Everything before the '/' and everything after.

Now take the following line:

ABC123ABC456ABC

if sed matches ABC and replaces it with XYZ but does not use the g command (sed 's/ABC/XYZ/') only the first ABC would be replaced.

Examples:

Code:

$ echo "ABC123ABC456ABC" |sed 's/ABC/XYZ/'
XYZ123ABC456ABC
$ echo "ABC123ABC456ABC" |sed 's/ABC/XYZ/g'
XYZ123XYZ456XYZ

BEGIN PEDANTRY

Except for backslash and newline.

END PEDANTRY



Regards,
Alister