Delete last two lines

[parmy_4]

Hi,
I have a file in which I don't need the last two lines. How can run a command to achieve this?

I have a work around for my collegue but he is not happy with it.

sed '$d' my_file | sed '$d'

I am expecting some thing like

sed -e '$d' -e '$d' my_file.

Please help us.
Parmy

[Pete Randall]

Parmy,

Use "sed -e '$d' my_file |sed -e '$d'"

that should do it.


Pete

[harry d brown jr]

vi my_file<
live free or die
harry

[Jean-Luc Oudart]

with vi cmd
vi +$ my_file<
Regards
Jean-Luc

[harry d brown jr]

Another way:

expr `wc -l my_file | cut -d" " -f1` - 2 | xargs echo | sed -e "s/^/head -/" -e "s/$/ my_file/" | sh > new_my_file

live free or die
harry

[john korterman · Accepted solution]

Hi,

I think this works (at least it seems so on my screen):

# tail -2 | comm -23 -

regards,
John K.

[Jean-Luc Oudart]

awk solution :
awk -v nb=$(wc -l my_file | cut -f1 -d" ") '{if(NR>(nb - 2)) exit; print}' my_file

Regards,
Jean-Luc

[Elmar P. Kolkman]

A possible solution:

head $(wc -l my_file | awk '{print "-" $1 - 2}') my_file

In vi you can use the command:
$-1,$d
to delete the last two lines. Should work in ed too.

Or a complete awk solution:
awk '{ if (NR > 2) { print prev2 } prev2=prev1;prev1=$0}' < my_file

[Bharat Katkar]

Hi
This works ..but don't know you like it or not.

# let A=`cat file1 | wc -l`-2
# head -$A file1 >file2

That works.
Regards,

[Jan Sladky]

hi Parmy,

I have no HP-UX where to try it but on Linux
exists the tac command (opposit to cat)

so maybe the following row will help you

tac file | sed '1,2d' | tac | tee file1


br Jan

[Rodney Hills]

Here is a perl one-liner that only requires one pass through the file.

perl -ne 'push(@last2,$_); if (scalar @last2 > 2) { print shift @last2}'

By pushing each line onto a stack this script will not print the last 2 lines.

HTH

-- Rod Hills

[Rodney Hills]

Parmy,

Assigning points to indicate which solution you think works the best is what helps the forum help others...

-- Rod Hills

[Rick Garland]

cat $file | sed '1,2d'

[Hein van den Heuvel]

Here is a one-pass in awk which does no create a temp file.
It uses a 'circular buffer' of the target 'missing line' count 'L'.
This is done with an array 'line' indexed by the current record number (NR) modulus that L.
You provide a desired value for L as argument.
We start printing from the array only after L lines have come by, 'falling behind' by L lines.

Sounds tricky? Here is the 'one line' code:

awk -vL=2 '{if(NR>L){print line[(NR-L)%L];line[NR%L]=$0}}' your-file


Hein.

[Elmar P. Kolkman]

Hein, your solution is a variant of my awk solution... but my mathematic mind says you need to remove the -L in your solution. It does nothing in the end... since (NR-L)%L equals NR%L.
And you have a closing curly bracket wrong, resulting in missing the first lines...

So the right solution would be:

awk -vL=2 '{if(NR>L){print line[NR%L];}line[NR%L]=$0}' my_file

[Elmar P. Kolkman]

Strange point assignment... I've tested the highest rated solution, but it doesn't seem to work. At least nog on the /etc/passwd files I have on Linux and HP-UX... But it could be me...

[Bharat Katkar]

It looks like Parmy is taking it very lightly and on the contrary all the responders are struggling to be as perfect as possible..

[Hein van den Heuvel]

Elmar> Hein, your solution is a variant of my awk solution...

Ayup, just a touch more general purpose. I had missed the awk part of your solution.

Elmar> but my mathematic mind says you need to remove the -L in your solution.

Ayup, a simplication I overlooked this time.

Elmar> It does nothing in the end... since (NR-L)%L equals NR%L.

Ayup, other then to make the code more clear.

Elmar> And you have a closing curly bracket wrong, resulting in missing the first lines...

Ooops! Thanks for catching that.

Groetjes,
Hein.