I've printed out the sed man page, and having skimming over it I don't see any "individual" edit commands for portions of a sentence. I'm starting to get the impression that you have to modify the entire sentence and make sed replace it just to modify a small portion of the original sentence.


Isn't there a command, for example, where I can say "Insert this phrase at this line after X characters from the beginning of the line"?

With sed you can edit just a portion of the sentence. Given a file called something.txt, containing:
-----------------------
one two three
four fix six seven
-----------------------

Example.

This will replace only 'two' with 'surprise'. And the replacement will be in the position of the original.

Can you give a specific example with real data?

Well, allow me to present this following example:

-------------------------
cat cat cat bird dog dog cat cat cat bird owl dog cat
-------------------------


It's quite repetitive isn't it?

Suppose, for example, I want to change the 2nd word "cat" into "dog"

-------------------------
cat dog cat bird dog dog cat cat cat bird owl dog cat
-------------------------


Or... suppose I wanted to add "xxxx" after he 4th "cat":

-------------------------
cat cat cat bird dog dog catxxxx cat cat bird owl dog cat
-------------------------

or remove the 3 "cats" in the middle:

-------------------------
cat cat cat bird dog dog bird owl dog cat
-------------------------







I'm given the impression that Sed CANNOT do these modifications directly- rather, I think Sed has to replace the entire line to make one change (even if the length is the same, such as if I replace ONE "cat" with "dog"). I mean of course you can still do the same replacements, but I was just wondering if Sed could do these changes to individual portions of a sentence.

Not it makes a difference or anything, I just find it very inefficent to replace the whole line when you only need to replace certian portions of it (I can understand if this makes some programming complications, but like I said I was just curious- after all, the guys at GNU are rather clever in their design)

This: This: This: (The last one's not pretty. Maybe someone has a cleaner solution.)

Anyway, all of the cases you mentioned can be handled with sed. I hope this is helpful.

seems good enough but I just realized that you had a word in each example to reference to- so allow me to present one problem that will answer my question:


You have the following data:
--------------
qwertyuiopasdfghjklzxcvbnm

poiuyt finis dsed fddd
3dgc6

5rfd
---------------


No words! Now... add "xxxx" after at point 15 (assuming the first character is point 0, as in an array in programming)

and add "5555" at line 3 position 5


showing me the command should answer my question- since I'll be able to see what I am missing.


Thanks

Ok. I see what you're asking. The sed (stream editor) program is probably not the appropriate tool for this.

lol great, this isEXACTLY what I was asking. You would need some sort of a front end in which you modify everything in a buffer, and use sed to flush it- rewriting the entire line even if it has one small change.


Not that it ruins sed for me lol- I just had a hard time finding a command to do this...cause it doesn't exist lol


Thank you very very much

It is a bit out of topic, but can you not use awk / gawk instead or in addition to sed
and solve your problem?

sed -e '1s/\(.\{15\}\(.*\)$/\1xxxx\2/' \
-e '3s/\(.{5}\)\(.*\)$/\1555\2/' testfile >testfile2

This will do what you asked on post #6.

If this data is broken into records, then indeed awk would be a much better choice. It merely records the first 15 characters in line 1, then records the rest of the line, and spits them out with your characters inserted.

(from post #1
Here is how I print out the print out man pages. Just in case you haven't noticed the -t option to man.
man -t sed | lp

jschiwal I am very impressed
Btw there is a missing \)
should read
'1s/\(.\{15\}\)\(.*\)$/\1xxxx\2/'
instead of
'1s/\(.\{15\}\(.*\)$/\1xxxx\2/'

and in the second -e expression {5} it should read \{5\}
but maybe it is because I have tested this with a "dos version" of sed.
I am sure you have planted those typos on purpose...

I was doing some reading on sed; anybody new to sed
and looking at advanced commands should really look at

http://sed.sourceforge.net/sedfaq.html

Thank you for catching my typo's. I had tested it on another computer before typing it in.

Regular expressions can be hard to read, there is no doubt about that. I find myself using egrep instead of grep simply to be able to use something like '[248]{1,3}\*' instead of '[248]\{1,3\}\*'.


The corrected sed command:
sed -e '1s/\(.\{15\}\)\(.*\)$/\1xxxx\2/' -e '3s/\(.\{5\}\)\(.*\)$/\1555\2/' temp >testfile2
jschiwal@hpand64:~> cat testfile2

This was actually an easy one. Sed can get difficult when you have patterns, that can span more than one line. Then you need to use the commands that add more lines to the pattern space and use labels and such.

If you use vim, grep or even search with regular expressions in less, simple sed commands become easier with time.