How to edit previous line from current in text file? - bash

Question

So what I need exactly.
I have a file that I looping line by line and when I'll found the word "search" I need to return on previous line and change the word "false" to "true" inside that line, but only on that line not for all file. I'm newbie in bash and that all that I have.
file="/u01/MyFile.txt"
count=0
while read line
do
((count++))
if [[ $line == *"[search]"* ]]
then
?????????????
fi
done < $file

Answer 1

You could do the whole thing in pure bash like this:
# Declare a function process_file doing the stuff
process_file() {
# Always have the previous line ready, hold off printing
# until we know if it needs to be changed.
read prev
while read line; do
if [[ $line == *"[search]"* ]]; then
# substitute false with true in $prev. Use ${prev//false/true} if
# several occurrences may need to be replaced.
echo "${prev/false/true}"
else
echo "$prev"
fi
# remember current line as previous for next turn
prev="$line"
done
# in the end, print the last line (it was saved as $prev) in the last
# loop iteration.
echo "$prev"
}
# call function, feed file to it.
process_file < file
However, there are tools that are better suited to this sort of file processing than pure bash and that are commonly used in shell scripts: awk and sed. These tools process a file by reading line after line1 from it and running a piece of code for each line individually, preserving some state between lines (not unlike the code above) and come with more powerful text processing facilities.
For this, I'd use awk:
awk 'index($0, "[search]") { sub(/false/, "true", prev) } NR != 1 { print prev } { prev = $0 } END { print prev }' filename
That is:
index($0, "[search]") { # if the currently processed line contains
sub(/false/, "true", prev) # "[search]", replace false with true in the
# saved previous line. (use gsub if more than
# one occurrence may have to be replaced)
}
NR != 1 { # then, unless we're processing the first line
# and don't have a previous line,
print prev # print the previous line
}
{ # then, for all lines:
prev = $0 # remember it as previous line for the next turn
}
END { # and after the last line was processed,
print prev # print the last line (that we just saved
# as prev)
}
You could also use sed:
sed '/\[search\]/ { x; s/false/true/; x; }; x; ${ p; x; }; 1d' filename
...but as you can see, sed is somewhat more cryptic. It has its strengths, but this problem doesn't play to them.
Addendum, as requested: The main thing to know is that sed reads line into something called the pattern space (on which most commands operate) and has a hold buffer on the side where you can save things between lines. We'll use the hold buffer to hold the current previous line. The code works as follows:
/\[search\]/ { # if the currently processed line contains [search]
x # eXchange pattern space (PS) and hold buffer (HB)
s/false/true/ # replace false with true in the pattern space
x # swap back. This changed false to true in the PS.
# Use s/false/true/g for multiple occurrences.
}
x # swap pattern space, hold buffer (the previous line
# is now in the PS, the current in the HB)
${ # if we're processing the last line,
p # print the PS
x # swap again (current line is now in PS)
}
1d # If we're processing the first line, the PS now holds
# the empty line that was originally in the HB. Don't
# print that.
# We're dropping off the end here, and since we didn't
# disable auto-print, the PS will be printed now.
# That is the previous line except if we're processing
# the last line (then it's the last line)
Well, I did warn you that sed is somewhat more cryptic than awk. A caveat of this code is that it expects the input file to have more than one line.
1 In awk's case, it's records that don't have to be lines but are lines by default.

Answer 2

A very simple approach would be to read 2 lines at a time and then check for the condition in the second line and replace the previous line.
while read prev_line # reads every 1st line
do
read curr_line # reads every 2nd line
if [[ $curr_line == *"[search]"* ]]; then
echo "${prev_line/false/true}"
echo "$curr_line
else
echo "$prev_line"
echo "$curr_line"
fi
done < "file.txt"

Answer 3

The correct version of your way of doing this would be:
file="/u01/MyFile.txt"
count=0
while read line
do
((count++))
if [[ $line == *"[search]"* ]]
then
sed -i.bak "$((count-1))s/true/false/" $file
fi
done < $file