Randomly selecting (units) from a file where a unit is 2 lines.

Question

I want to select from a file random lines/units but where the units are consisted of 2 lines.

For example a file looks like this

Adam
Apple
Mindy
Candy
Steve
Chips
David
Meat
Carol
Carrots

And I want to randomly subselect lets say 2 units group

For example

Adam
Apple
David
Meat

or

Steve
Chips
Carol
Carrots

I've tried using shuf and sort -R but they only shuffle 1 lines. Could someone help me please? Thank you.

Have you considered preprocessing the data so that you only need to select single lines at random, and then reformat the selected lines as two lines each. — Jonathan Leffler, Jun 13 '17 at 22:46

ilkkachu · Accepted Answer · 2017-06-14T07:24:33.363

2

You could do it with shuf by joining the lines before shuffling (that might not be a bad idea for a file format in general, if the lines describe a single item):

$ < file sed -e 'N;s/\n/:/' | shuf | head -1 | tr ':' '\n'
Carol
Carrots

The sed loads two lines at a time, and joins them with a colon.

edited Jun 14 '17 at 07:24

answered Jun 13 '17 at 21:49

ilkkachu

6,221
16
30

You don't need the p, just remove the -n flag. Also if your sed supports `\n` then you don't need the -e either and it accepts filenames as arg so you don't need `<`. It could be written as `sed 'N;s/\n/:/' file` – 123 Jun 13 '17 at 22:08

score 0 · Answer 2 · answered Jun 13 '17 at 22:12

0

Pick a random number in the correct range, ensure that it is odd (if desired), then use sed to print the 2 lines:

$ a=$(expr $RANDOM % \( $(wc -l < input) / 2 \) \* 2 + 1)
$ sed -n -e ${a}p -e $((a+1))p input

answered Jun 13 '17 at 22:12

William Pursell

204,365
48
270
300

Any reason you chose `expr` over `$(())` or `(())`? – 123 Jun 13 '17 at 22:19
1

No. It was just the first thing I typed. Any reasonable person would use `a=$(( $RANDOM % ( $( wc -l < input ) / 2 ) * 2 + 1))` – William Pursell Jun 13 '17 at 22:21
1

Well, any reasonable person would probably do something else entirely :) But this works. – William Pursell Jun 13 '17 at 22:22

William Pursell · Answer 3 · 2017-06-13T22:30:59.663

0

Rather than selecting lines to print, you could walk the file and print each "unit" with a particular probability. For example, to print (roughly) 10% of the "units" in the file, you could do:

awk 'BEGIN{srand()} NR%2 && (rand() < .1) {print; getline; print}' input

edited Jun 13 '17 at 22:30

answered Jun 13 '17 at 22:26

William Pursell

204,365
48
270
300

This prints nothing for me 90% of the time. – 123 Jun 13 '17 at 22:27

Randomly selecting (units) from a file where a unit is 2 lines.

3 Answers3