Selecting lines according to position in a file
 

Have: a file called numbs consisting of positive integers, one per line ...
The numbers are not sorted and not necessarily unique.

Have: a file called words consisting of text ...
The text could just as well be the names of colors or states or lines from a poem.

The desired output is a file consisting of lines selected from words according to the numbers in numbs ...
As a matter of personal style I strive to write code which does not have explicit loops. I've tried several approaches which don't work. If you can do this without loops, please show how. Solutions which use commands such as sed, grep, and awk are preferred. I haven't learned perl or ruby.

These are failed attempts:
Daniel B. Martin

Is this what you are looking for (bash only solution):

Thank you for working code... but... I am seeking a loop-less solution. Your bash solution uses two while loops.

Daniel B. Martin

From a technical point of view those are 2 loops, but....

The way they are implemented in my solution they work like sed or awk (read each file once, one line after another). If that is considered a loop then both sed and awk cannot be used either ;)

Why? Seems like a pretty silly requirement.

Let's start with obvious loop:
We can use xargs to make the loop implicit:
Both cases use O(nw) time, O(1) space (n is size of numbs, w is size of words)

awk solution (equivalent to druuna's bash solution):
This trades space O(w) for time O(n + w).

I am retired and do programming as recreation, hoping to stave off the inevitable Old Age Brain Rot. I dream up silly problems with silly requirements for amusement. This is one which I couldn't solve on my own so I called on the LQ experts.

Daniel B. Martin

Good idea. I would suggest donating your time to open source projects as that not only will help stave off the Old Age Brain Rot, but it will also produce something that others can use and work off of. Plus you'll learn a whole lot. That will, of course, force you to learn a programming language and not rely on Linux commands. :)

Superb!

For the sake of learning: Please explain why my attempt with sed ...
... generates this OutFile ...
All the desired words are there, but not in the desired order.

Can my sed be changed to make it work, and still be loop-less?

Daniel B. Martin

Hi, Daniel.

Your sed solution does not work, because on the second line you execute sed script $Work on each line of file $Words. For example, you apply script "3p;6p;2p;8p;8p;2p;8p;" to an input line number 2. The script contains '2p' two times, so sed prints second line two times.

EDIT: Wow, it's my 600-th post here :)

The basic idea of using sed to write a sed program can be used like so:

Unfortunately, unless sed does some heroic optimization, I think this solution is O(nw) in time and O(w) in space.

Hi again!

Maybe try another tool?:
(45 goes to standard error)

This works, because ed treats numbers on stdin as line addresses to print.

Read info ed for more info.

Congratulations on achieving this milestone and garnering 336 reputation points. That's impressive!

This wins a prize for being concise.

Daniel B. Martin

Nice solution!

Hi.

Hmm.. After your post my reputation suddenly jumped 12 points up. That's impressive! :) Thanks guys!

Daniel, please keep asking your questions -- they are very good for limbering up my brains, I enjoy solving them. And your english reminds me of Shakespeare :)