3

I have a file foo.txt

test
qwe
asd
xca
asdfarrf
sxcad
asdfa
sdca
dac
dacqa
ea
sdcv
asgfa
sdcv
ewq
qwe
a
df
fa
vas
fg
fasdf
eqw
qwe
aefawasd
adfae
asdfwe
asdf
era
fbn
tsgnjd
nuydid
hyhnydf
gby
asfga
dsg
eqw
qwe
rtargt
raga
adfgasgaa
asgarhsdtj
shyjuysy
sdgh
jstht
ewq
sdtjstsa
sdghysdmks
aadfbgns,
asfhytewat
bafg
q4t
qwe
asfdg5ab
fgshtsadtyh
wafbvg
nasfga
ghafg
ewq
qwe
afghta
asg56ang
adfg643
5aasdfgr5
asdfg
fdagh5t
ewq

I want to print all the lines between qwe and ewq in a separate file. This is what I have so far :

#!/bin/bash

filename="foo.txt"

#While loop to read line by line
while read -r line
do
    readLine=$line
    #If the line starts with ST then echo the line
    if [[ $readLine = qwe* ]] ; then
        echo "$readLine"
        read line
        readLine=$line
        if [[ $readLine = ewq* ]] ; then
            echo "$readLine"
        fi
    fi
done < "$filename"

closed as unclear what you're asking by Kusalananda, G-Man, GAD3R, don_crissti, Jeff Schaller Oct 12 '17 at 0:21

Please clarify your specific problem or add additional details to highlight exactly what you need. As it's currently written, it’s hard to tell exactly what you're asking. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 3
    sed '/qwe/,/ewq/ w other.file' foo.txt – Costas Jan 18 '16 at 19:19
  • 1
    @Costas I cant use that because I need to do some logic between these lines . . – gkmohit Jan 18 '16 at 19:21
  • 3
    Sure you can... If you're using while..read you're almost always doing it wrong. – don_crissti Jan 18 '16 at 19:35
  • 2
    read this – don_crissti Jan 18 '16 at 19:42
  • 3
    @BinaryZebra - what problem ? If you know what the real problem is here please explain it so that we can understand it because the question is (obviously) some sort of XY question but the OP doesn't even bother explaining what X is (the real problem). – don_crissti Jan 18 '16 at 20:11
7

You need to make some changes to your script (in no particular order):

  • Use IFS= before read to avoid removing leading and trailing spaces.
  • As $line is not changed anywhere, there is no need for variable readLine.
  • Do not use read in the middle of the loop!!.
  • Use a Boolean variable to control printing.
  • Make clear the start and end of printing.

With those changes, the script becomes:

#!/bin/bash

filename="foo.txt"

#While loop to read line by line
while IFS= read -r line; do
    #If the line starts with ST then set var to yes.
    if [[ $line == qwe* ]] ; then
        printline="yes"
        # Just t make each line start very clear, remove in use.
        echo "----------------------->>"
    fi
    # If variable is yes, print the line.
    if [[ $printline == "yes" ]] ; then
        echo "$line"
    fi
    #If the line starts with ST then set var to no.
    if [[ $line == ewq* ]] ; then
        printline="no"
        # Just to make each line end very clear, remove in use.
        echo "----------------------------<<"
    fi
done < "$filename"

Which could be condensed in this way:

#!/bin/bash
filename="foo.txt"
while IFS= read -r line; do
    [[ $line == qwe* ]]       && printline="yes"
    [[ $printline == "yes" ]] && echo "$line"
    [[ $line == ewq* ]]       && printline="no"
done < "$filename"

That will print the start and end lines (inclusive).
If there is no need to print them, swap the start and end tests:

#!/bin/bash
filename="foo.txt"
while IFS= read -r line; do
    [[ $line == ewq* ]]       && printline="no"
    [[ $printline == "yes" ]] && echo "$line"
    [[ $line == qwe* ]]       && printline="yes"
done < "$filename"

However, it would be quite better (if you have bash version 4.0 or better) to use readarray and loop with the array elements:

#!/bin/dash
filename="infile"

readarray -t lines < "$filename"


for line in "${lines[@]}"; do
    [[ $line == ewq* ]]       && printline="no"
    [[ $printline == "yes" ]] && echo "$line"
    [[ $line == qwe* ]]       && printline="yes"
done

That will avoid most of the issues of using read.


Of course, you could use the recommended (in comments; Thanks, @costas) sed line to get only the lines to be processed:

    #!/bin/bash
filename="foo.txt"

readarray -t lines <<< "$(sed -n '/^qwe.*/,/^ewq.*/p' "$filename")"

for line in "${lines[@]}"; do

     : # Do all your additional processing here, with a clean input.

done 
  • 2
    This answer very nicely answers the question of "how do I cope with most of the issues of using read". Sometimes though, the OP needs to be unasked: using read for this task requires all your explanation and is manifestly the wrong tool for the job when half-line sed comment just works. (±0 if it matters). – msw Jan 18 '16 at 20:25
  • @mikeserv BinaryZebra, please stop having this sort of discussion in the comment threads. Comments are not for extended discussion; this conversation has been moved to chat. – terdon Jan 19 '16 at 11:54
4

As @Costas pointed out, the correct tool to use for this job is sed:

sed '/qwe/,/ewq/ w other.file' foo.txt

There may be other processing needed on the lines to be printed. That's fine; just do it like so:

sed -e '/qwe/,/ewq/{w other.file' -e 'other processing;}' foo.txt

(Of course, "other processing" isn't a real sed command.) The above is the pattern to use if you need to do your processing after you print the line. If you want to do some other processing and then print a changed version of the line (which seems more likely), you would use:

sed -e '/qwe/,/ewq/{processing;w other.file' -e '}' foo.txt

(Note that it is necessary to put the } into its own argument, otherwise it will be interpreted as part of the other.file name.)


You (the OP) haven't stated what "other processing" you have to do on the lines, or I could be more specific. But whatever that processing is, you can definitely do it in sed—or if that becomes too unwieldy, you could do it in awk with very little change to the above code:

awk '/qwe/,/ewq/ { print > "other.file" }' foo.txt

Then, you have all the power of the awk programming language at your disposal to do processing on the lines before you execute that print statement. And of course awk (and sed) are designed for text processing, unlike bash.

  • The solutions do not provide a way for the OP to "I need to do some logic between these lines" in shell. Yes, sed or awk may probably do the same (unknown) processing. – user79743 Jan 18 '16 at 22:58
  • 1
    @BinaryZebra, see the last sentence of my answer. He didn't say he needs to do the logic in shell, just that he needs to do some logic. But it's good that your answer provided a way he can do the logic in shell if he needs to. – Wildcard Jan 19 '16 at 6:41
1
qwe(){ printf %s\\n "$1"; }
ewq(){ :; }
IFS=   ### prep  the  loop, only IFS= once
while  read -r  in
do     case $in in
       (qwe|ewq)
           set "$in"
       ;;
       ("$processing"?)
           "$process"
       esac
       "$1" "$in"
done

That's one really slow way to do it. With a GNU grep and a regular infile:

IFS=
while grep  -xm1 qwe
do    while read  -r  in  &&
            [ ewq != "$in" ]
      do    printf %s\\n "$in"
            : some processing
      done
done <infile

...would at least optimize out half of the inefficient reads...

sed  -ne '/^qwe$/,/^ewq$/H;$!{/^qwe$/!d;}' \
      -e "x;s/'"'/&\\&&/g;s/\n/'"' '/g"    \
      -e "s/\(.*\) .e.*/p '\1/p" <input    |
sh    -c 'p(){  printf %s\\n "$@"
                for l do : process "$l"
                done
          }; . /dev/fd/0'

And that would avoid the inefficiencies of read altogether for most sh's out there, though it does have to print the output twice - once quoted to sh and once unquoted to stdout. It works differently because the . command tends to read input by block rather than by byte for most implementations. Still, it elides ewq - qwe altogether, and would work for streamed input - such as a FIFO.

qwe
asd
xca
asdfarrf
sxcad
asdfa
sdca
dac
dacqa
ea
sdcv
asgfa
sdcv
qwe
a
df
fa
vas
fg
fasdf
qwe
aefawasd
adfae
asdfwe
asdf
era
fbn
tsgnjd
nuydid
hyhnydf
gby
asfga
dsg
qwe
rtargt
raga
adfgasgaa
asgarhsdtj
shyjuysy
sdgh
jstht
qwe
asfdg5ab
fgshtsadtyh
wafbvg
nasfga
ghafg
qwe
afghta
asg56ang
adfg643
5aasdfgr5
asdfg
fdagh5t
-1
sed '/./=' input2.file | sed -n '/./N;s/\n/  /; /qwe/,/ewq/p'

Not the answer you're looking for? Browse other questions tagged or ask your own question.