Removing columns from a file using `sed`

indiancosmonaut · 09-24-2007, 02:39 AM

Hi,

I have a file

aa1:123:sometext1:name1:zz1
bb1:457:sometext2:name2:yy2
cc1:78:sometext3:name3:xx3

Could you help me with a `sed` command to remove entire column-3 i.e. the sometext column.

PS: Share with me some good literature for sed.

Kind regards and thanks in advance.

indiancosmonaut

druuna · 09-24-2007, 05:56 AM

Hi,

There's an easier way to cut out this column: The cut command.

cut -d: -f1,2,4,5 infile

If you do want/need a sed solution:

sed 's/$.*:.*$:.*:$.*:.*$/\1:\2/' infile

A good reference for Sed (and AWK): O'Reilly's 'sed & awk'
Or on-line: the SED $HOME

Hope this helps.

ghostdog74 · 09-24-2007, 06:14 AM

awk is the tool to use for working with columns

Code:

awk 'BEGIN{OFS=FS=":"}
     {$3="";print }
' "file"

osvaldomarques · 09-24-2007, 06:24 AM

Hi Indiancosmonaut,

I can't share any literature with you as I just read "man" and "info" for the trivial tasks and, for the rest, I google.

The following code uses regexp to get 3 groups of non-colon characters separated with a colon, from the start of the line; the first 2 can be referenced after by enclosing them between parenthesis; they will be referenced as variable 1 on the second part of the sed expression:

Code:

sed 's/^\([^:]*:[^:]*\):[^:]*/\1/' sometext

indiancosmonaut · 09-27-2007, 07:14 AM

Hi Druuna and ghostdog,

Thanks for the solutions and the Link. Its got lots of information!!!
I'll try to find the good ones...

Hi osvaldomarques,

Thanks for the solution you gave but I'm unable to fully understand it.
It goes over my head! I'll try to read more about this. Meanwhile, could you kindly explain it in a little detail so that I am able to understand it.

Kind regards,

indiancosmonaut

pixellany · 09-27-2007, 08:19 AM

Another good reference:
http://www.grymoire.com/Unix/Sed.html

indian...

osvaldo gave you a pretty complete explanation--what makes it hard is that the "backreference" is embedded in other code.

First, look at the simple syntax for a backreference. This code finds the first occurrence of "yp" and replaces it with "y":
sed 's/$y$p/\1/' filename
In a backreference, everything in the escaped parentheses is inserted where the "\1" occurs. Obviously, this could be done with:
sed 's/yp/y/' filename

The best way to learn sed is to write very simple commands first--once you know how they work, then add more layers of complexity.

indiancosmonaut · 10-04-2007, 01:42 AM

Thanks pixellany...

I read some literature and understood some of it.
Will try to learn it in detail... :-)

Kind regards,

indiancosmonaut