Confused by the sed command

Shell Programming and Scripting
1

Hi Gurus,

I am a little bit confused by the sed command.

my file is below

 
cat sample
aaa
bbb
ccc
ddd
eee
fff
ggg
hhh
iii

below is command and output.

 
sed  '/ddd/{x;p;}' sample.txt 
aaa
bbb
ccc
 
       --blank
       --blank
eee
fff
ggg
hhh
iii

as per my understanding, {x;p;} when match found, x -- exchange hold place with pattern place which is null with ddd, then p; print, I thought it only print once, but here it prints twice.

Please help me to understand this, thanks in advance.

Ps, I think I figured out this. it works as below. x -- exchange hold place with pattern place which is empty. with ddd then p; print. after finish these two. the default is "print pattern space", so it prints the pattern space content again which is empty.

Please correct me if anything incorrect.

2

sed -n may be your friend. Thus don't print unless you explicitly say so.

3

Actually, in your sample output there are three blank lines that in real life aren't there; it's just two.
And yes, you are right with your analysis - for ddd lines, there is one explicit p rint after the x change, and one implicit. If you want just the explicit one, append (man sed)

4

This SED command and the MAN pages drive me nuts, its never as simple to follow and even if you use google its almost impossible to get a simple example and visual representation of what you actually need to do. Anybody can read a story , but just show the actual practicality of it, what?

I have a file which has broken lines, unintended line feeds.
I want to repair them, so my thoughts are , find "something" unique about those particular broken lines and then do something to make them append to the rest of the line accordingly.
I have found out how to do this in AWK and GAWK.

I wish to know how to do this using SED which i understand is the most correct method.
I understand i need to use SED '/find something/ {N something something, i dont undersgand the syntax and the material i have referenced doesn't explain this in a one liner so its just ridiculous.
I have tried for a month already on and off but i am not able to win here.

The character which is unique to the broken records is a "$ at the end of the line....which is CORRECT and just a $ character is the incorrect broken line.
So i need to say find lines that end in ONLY $ and then append the NEXT line to that line with the $ character only.
So again, the lines ending with "$ are OK and the lines ending with $ NOT OK.

I use cat -vet filename.txt to see the $ characters.

I know i need to use N as an option to SED.

Appreciate the simple example and answer. Thanks

5

@lmre: the n command seems more appropriate here.
To delete an empty line following a line ending with a "

sed '/"$/{n;/^$/d;}' file
1 Like
6

I need to search for lines ending with $ and not "$
The i want to append the next line to that line,
I dont want to delete anything, i dont know why i need to delete in your example.
There is no empty line here.

23445w34,"djhahfhake,shdjgekagdgejs"$
23445w34,"djhahfhake","shdjgekagdgejs"$
23445w34,"kjshdfd","shdjg$
"."ekagdgejs"$
23445w34,"djhamdhzkjfsdtheree,jgekagdgejs"$
23445w34,"doubletroubkeinhungary","shdjgekagdgejs"$

Line 3 is the broken line and ends with $ and i want to APPEND line 4 to line 3.
Thats it.

I have used GAWK and AWK like this :
rskjhbvsfs1:icsoka:~/code> awk '{if ( /.$/ != /\"$/ ) printf $0 ; else print $0 } southafrica.txt

Now i want to win using SED.

7

Ah, then it's

sed '/"$/!{N;s/\n//;}' file

If not ending with a " append the next line, delete the embedded \n

---------- Post updated at 05:20 PM ---------- Previous update was at 03:18 PM ----------

The awk command is still easier here, an improved version is

awk '{if ( /"$/ ) print $0 ; else printf "%s",$0 }' file

sed lacks the capability to print a line without a newline at the end.
So more effort is needed.
Correct implementations with sed are, using a loop and a look-ahead append:

sed '                                                       
:loop
/"$/b
$!N;s/\n//
t loop
' file

or, using the hold space:

sed '
H;/"$/!{$!d;}
s/.*//;x;s/\n//g
' file
1 Like