Delete lines from file with SED or AWK

Question

Ive seen many variations, very confused on how to solve these 3 problems.

1. deleting all rows except the first from a file
2. deleting a row from file with a line number
3. deleting rows from a file with a range of line numbers

Solution 1

Using sed:

Delete 1st line:

sed '1d' file-name

Delete 10th line:

sed '10d' file-name

Delete line # 5 to 10

sed '5,10d' file-name

All above sed commands will write output on stdout that you can redirect to another file if you want or use -i flag of sed to inline edit the file.

Solution 2

With awk:

# delete line 1
awk 'NR == 1 {next} {print}' file

# delete line number stored in shell variable $n
awk -v n=$n 'NR == n {next} {print}' file

# delete between lines $a and $b inclusive
awk -v m=$a -v n=$b 'm <= NR && NR <= n {next} {print}' file

To save a few chars, {print} can be replaced just with 1

To overwrite the original file, you have to do something like this

awk '...' file > tmpfile && mv tmpfile file

Comments

[bluetickk]

ok this is working but the lines are just printing out and not saving to the file?

[Beta]

To delete all lines but the first, sed '2,$d' filename, or sed '1!d' filename, or sed -n '1p' filename.

[jilles]

The way you use it, read -r will still strip leading and trailing whitespace. You need to do IFS= read -r line. Furthermore note that from a shell script this method is faster for small files because it avoids a fork, but slower for large files because read is inherently inefficient and usually reads one byte at a time or does one read and lseek call per invocation and string processing in bash tends to be inefficient (less so in other shells).

[bash-o-logist]

@jilles, thanks I forgot about IFS= on these cases. And yes, i do know that read is inefficient on large files with bash. If OP's files are large sizes and performance is an issue, then use a better tool.

[dubiousjim]

@Beta, first two are exactly right, but the third will print the first line only.

[anubhava]

@dubiousjim: Print first line only is same as delete all lines but the first that's why sed -n '1p' is also correct.

[dubiousjim]

@anubhava, sorry, mental glitch. I thought you were trying to do something else; don't know on what basis I thought so.

[Andrew Dalke]

To delete line 1, use awk 'NR!=1'. The default action is to print the line. All of your '{next} {print}' terms can be removed.

[SKJ]

Lifesaver for loading a sql dump without destroying / creating the table. Many thanks :)