extract/delete line from file to other.
 

hi, guys

I need your help on a piece of code.

i have 2 files.

file1.txt with content:
song1
song2
song3
song4

and an empty file2.txt

and i want to extract and delete randomly a line from file1.txt and put it on file2.txt

the result would be:

file1.txt with content:
song1
song2
song4

file2.txt with content:
song3

Hi and Welcome to LQ.

Ok, where exactly are you stuck?

I'm stuck with the command sed can't make it working :(

So what sed have you tried? Are you able to do the 2 individual tasks? ie print line to other file and delete line from first file.

So what's your code so far?

i don't have it yet, i know that sed can't remove line with an expression sed '/Hello/d' file.txt but i can't implement on the way to get a random line and put it on the other file!

i'm stuck in this, with this sed line i can get a random line and print on file2.txt but i can't insert this as variable on sed to delete the line.

GETLINE=$(sed -n $((1+$RANDOM%`wc -l file1.txt | cut -f 1 -d ' '`))p file1.txt > file2.txt)

sed -e "/$GETLINE/d" file1.txt > file1_.txt

mv file1_.txt file1.txt

this is to run every 30 minutes and when file1.txt gets empty do:

find /home/mp3/ -type f -name "*.mp3" > file1.txt

It might be better to calculate a random line number (within range), and then use:

sed -n "${num}p" text1 >> text2
sed -i "${num}d" text1

You don't use slashes around the line number for one thing.


Break up the problem so that your calculations is outside the sed command.

declare -i lines random_line
lines=$(($(wc -l text1)))
random_line=$((${RANDOM}%${lines}+1))

There will be an error here. Declaring lines, and random_lines helped pick up the mistake.
..
declare -i lines random_line
lines=$(($(wc -l text1.txt | cut -d' ' -f1)))
random_line=$(( ${RANDOM%${lines}+1 ))

That's better.

sed -n "${random_line}p" file1.txt >> file2.txt
sed -i "${random_line}d" file1.txt

A common catch is to not use $(( ... )) when dealing with integers.

jschiwal impossible to do that because the total of number of lines it's not always equal. but thanks anyway, your code gave me a better ideia.

i found the solution:

N=$((1+$RANDOM%`wc -l file1.txt | cut -f 1 -d ' '`))

sed -n "${N}p" file1.txt >> file2.txt

sed -i "${N}d" file1.txt


now i only got to implement an if statement when file1.txt gets empty.

I thought you might like this tidy up:

grail: don't you have to put the cut -f 1 -d ' ' part to remove the empty line on the file????

No you are using the cut because if you do:
The output is:
However when using the cat command, wc see the input as coming from stdin and output is:

ok got it!

SOLVED!!!!!!!!!!!!!!!!

Cool ... look under thread tools to mark as SOLVED.

Yes you need to count the number of lines each time.
My main point is that separating the tasks into separate lines can make debugging easier. Also use "declare -i" for integer variables to expose hidden errors where a string would be assigned to the variable.