Hi all!

I'm struggling with a problem with sed (or awk).
I have several different files which need a change on a certain line.
The files contain lines like:
B-1234/ 0001ABCD-0023ASDF-W-5--
The next file has a line like:
B-1234/ 0001ABCD-0023EFVB-W-5--

I need to find a way to change ASDF and EFVB into ABCD.
What I don't want is to search for the strings that has to be replaced, because they change over time.
What I want is to replace the 4 characters starting of position 22 with ABCD.

I'm searching for two days now but it seems I can't get hold of the right construction.

Please help me find the right construction.

TIA!
Onno.

Hi,

If I understand correctly, you only want to change positions 23 thru 26 with ABCD, no matter what the first 22 or the last 6 are. If that is the case:

sed 's/\(.\{21\}\)\(.*\)\(.\{6\}\)/\1ABCD\3/' infile

Example run:
It's up to you to figure out the sed command used If you cannot figure it out, just ask.

Hope this helps.

Thereby proving that one can write totally inscrutable commands using SED.......

3 groups in "\()\" for backreference....loosely translated as:

group1: 21 of any character
group2: any # of characters
group3: 6 of any character

Replace with: group1 + ABCD + group3

Note that this will be VERY dependent on the format of the original file.

Another method would be to replace the 2nd group of 4 capital letters:

Even this is dependent on the exact structure of the file entries

Hi,

@pixellany: Shouldn't that be: Replace with: group1 + ABCD + group3

I do like the shorter example you give (keeping it 'as simple' as possible).

Of course!!! (fixed in the original)

Many thanks guys!!

It indeed solved the 'problem'.
Now I'm changing the scripts so it will only update a certain line, because each file only contains one line like C-7906/ 0001AASU-0031DNGS-M-9-- and the rest of the file don't need any changes.
With the current command to many lines get changed.

So I'm going to use grep -i -n "<needle>" <haystack> to get the right linenumber and feed that one to the sed command!!

Thanks again!!
From a windy netherlands,
Onno.

Hi,

You probably can do that with one sed command (no need to use grep and sed). Something like this:

sed '/^C-4321/s/\(.\{21\}\)\(.*\)\(.\{6\}\)/\1ABCD\3/' infile will only change a line if it starts with C-4321

Wildcards are also possible:

$ sed '/^[ACE]/s/\(.\{21\}\)\(.*\)\(.\{6\}\)/\1ABCD\3/' infile This will change only those lines that start with an A, C or E.

Hope this helps.

Thanks for the tip druuna.
But I still have to use grep since the lines start everytime with a different 'key'. I already changed my script and it's working without a hitch.

But thanks again!
Onno.