How to move some lines to the top of the file?

Question

I have a file like this:

aaa.bbb.1.ccc
xxx.bbb.21
mmm.ppp
xxx.eee
mmm.qqqq
xxx.hhh.12.ddd

I want to move all the lines starting with xxx. at the top of the file with a simple command (using sed, awk, grep...).

So my new file will look like this:

xxx.bbb.21
xxx.eee
xxx.hhh.12.ddd
aaa.bbb.1.ccc
mmm.ppp
mmm.qqqq

How to do this?

score 6 · Answer 1 · answered Mar 04 '15 at 10:15

6

One sed solution:

sed -n '/^xxx/! { H }; // p; $ { x; s/^\n//; p }' infile

The H appends the content that does not begin with xxx to the hold space, and the opposite match is printed inmediatly // p. At last line $ recover the hold space contents, remove leading newline character and print.

It yields:

xxx.bbb.21
xxx.eee
xxx.hhh.12.ddd
aaa.bbb.1.ccc
mmm.ppp
mmm.qqqq

answered Mar 04 '15 at 10:15

Birei

35,723
2
77
82

An alternative solution but in the same vein: `sed '/^xxx/p;//!H;$!d;x;s/\n//' file` – potong Mar 04 '15 at 11:51

score 4 · Answer 2 · answered Mar 04 '15 at 10:05

You can process the file twice with awk:

$ awk 'FNR==NR {if (/^xxx\./) {print; a[FNR]} next} !(FNR in a)' file file
xxx.bbb.21
xxx.eee
xxx.hhh.12.ddd
aaa.bbb.1.ccc
mmm.ppp
mmm.qqqq

In the first one, we match the lines starting with xxx. and: print them, store its number. In the second reading of the file, we just skip those stored lines.

Jotne · Answer 3 · 2015-03-04T11:41:19.413

3

Another awk

awk '!/^xxx/ {a[$0];next} 1; END {for (i in a) print i}' file
xxx.bbb.21
xxx.eee
xxx.hhh.12.ddd
aaa.bbb.1.ccc
mmm.ppp
mmm.qqqq

!/^xxx/ If line does not start with xxx
a[$0];next store it in array a
1 print line with xxx
END {for (i in a) print i} At end print lines in array a. (the one without xxx)

As fedorqui (thank you) points out, if order of lines are important, use this:

awk '!/^xxx/ {a[++c]=$0;next} 1; END {for (i=1;i<=c;i++) print a[i]}' file
xxx.bbb.21
xxx.eee
xxx.hhh.12.ddd
aaa.bbb.1.ccc
mmm.ppp
mmm.qqqq

edited Mar 04 '15 at 11:41

answered Mar 04 '15 at 10:17

Jotne

40,548
12
51
55

note this maybe does not preserve the order ([your own question!](http://stackoverflow.com/q/20683474/1983854)) – fedorqui Mar 04 '15 at 11:04

score 3 · Answer 4 · answered Mar 04 '15 at 10:26

3

Another awk for those not in the mood to use arrays:

$ awk 'NR == FNR { if (/^xxx/) print; next } !/^xxx/' top.txt top.txt
xxx.bbb.21
xxx.eee
xxx.hhh.12.ddd
aaa.bbb.1.ccc
mmm.ppp
mmm.qqqq

NR == FNR { if (/^xxx/) print; next } First pass, print lines starting with xxx
!/^xxx/ Second pass, print lines not starting with xxx

answered Mar 04 '15 at 10:26

jas

10,715
2
30
41

You can write this `top.txt top.txt` like this `top.txt{,}` – Jotne Mar 04 '15 at 11:04
@jas, Why do you need to equate NR==FNR? Can you please explain the purpose of it – Sadhun Mar 04 '15 at 13:01
1

Sure, @Sadhun, it's actually quite well explained here: http://backreference.org/2010/02/10/idiomatic-awk/ (search for the section headed "Two-file processing"). In this case, of course, instead of process two files we're processing the same file twice (putting the file on the command line twice). As `NR` is the total number of recs read by awk until this point, and `FNR` is the number of recs within the current file, when `NR == FNR` we know we are reading the file for the first time. The `next` keps awk from going to the next statement, which therefore can only be reached on second readng – jas Mar 04 '15 at 13:34
@jas - I got the answer. NR==FNR will run the first operation in first file and once fails it goes for next operation in second file – Sadhun Mar 04 '15 at 13:51

score 2 · Answer 5 · edited Mar 04 '15 at 11:05

2

With grep:

grep "^xxx" File > NewFile && grep -v "^xxx" File >> NewFile

Redirect all lines starting with xxx to file NewFile. Then grep for lines not starting with xxx and append to NewFile.

It uses && between commands, so that one is not executed if the former exited with the bad status.

edited Mar 04 '15 at 11:05

fedorqui

275,237
103
548
598

answered Mar 04 '15 at 10:00

Arjun Mathew Dan

5,240
1
16
27

you probably want to say `&&` instead of `;`. This way, the final `mv tmp File` won't be done if something failed before. Otherwise, you may end up truncating `File`. – fedorqui Mar 04 '15 at 10:10
1

or `{ grep ; grep ; } > NewFile`. Bad new is 2 subshell at least but very fast on huge file – NeronLeVelu Mar 04 '15 at 12:03

score 1 · Answer 6 · answered Mar 04 '15 at 12:05

1

sed '/^xxx/!{H;$!d;}
     ${x;s/.//;}' YourFile

using the d behaviour (cycle without going further in script)

answered Mar 04 '15 at 12:05

NeronLeVelu

9,908
1
23
43

How to move some lines to the top of the file?

6 Answers6