如何使用 sed 或其他命令将现有字符串替换为每行不同的字符串

Question

for i in 1 2 3 4 5 6 7; do
 sed "s/\number\b/& =$i/" print.txt 
done

print.txt contains:

number elephant
number giraffe
number dogs
number cats
number mouse
number pigs
number snake

How do I replace a word/string in ONE line of a text file in a for loop without changing every line? Whenever I run my script it makes all the 'number' into one specific number and by the time the last iteration runs, all the 'number' is replaced by 7. I understand why it does that, just not sure how to fix it.

In this simple example, I want it to be outputted like this:

1 elephant
2 giraffe
3 dogs
4 cats
5 mouse
6 pigs
7 snake

But let's say, in my for loop, it defines a variable to be goat and the second iteration to be chicken, I want pigs to be changed to goat and snake to be changed to chicken, how do I do that inside my for loop?

Answer 1

If I understand you want to replace "number" in each line with increasing values 1-7 , then you don't need (or want) a loop at all. Simply pipe the result of sed removing the "number" string from the beginning of each line to nl -w1 -s' ' to number the lines. Example:

$ sed 's/^number\s*//' print.txt | nl -w1 -s' '
1 elephant
2 giraffe
3 dogs
4 cats
5 mouse
6 pigs
7 snake

( note: you can use grep -o '\\w*$' print.txt | nl -w1 -s' ' as well)

You can adjust the number format and separator with options. See man 1 nl .

Why don't you want a loop? How many times are you invoking sed and spawning an additional subshell within your loop? (one per-iteration). This is horribly inefficient. Any time you are looping in shell, you want to minimize the number of subshells spawned within your loop.

Can you do it with a loop as you were attempting? Of course, but it is one of the least efficient solutions you could come up with. To make it work in a loop, you have to limit your operation per-iteration to the wanted line in the file. You can do that by suppressing normal output and only outputting the line you changed, eg

for i in 1 2 3 4 5 6 7; do
    sed -n "${i}s/number/$i/p" print.txt
done

( but DON'T do it this way... )

If you want an even more efficient solution that eliminates the pipe and additional subshell, just use awk and a counter ( n below), eg

$ awk -v n=1 '{print n++ " " $2}' print.txt
1 elephant
2 giraffe
3 dogs
4 cats
5 mouse
6 pigs
7 snake

You can likewise further control the number and separator format using printf instead of print above.

Let me know if you have questions or if you are wanting something different.

---

Answer to Comment Question

If you have 3 lines with an identical field you want to change to something else, you first have to know what you want to change the identical fields to. That means you must have a replacement table or lookup table of replacements somewhere.

Say your text file is actually:

$ cat print2.txt
number elephant
number giraffe
number dogs
number cats
number pigs
number pigs
number pigs

And you want to change lines 5, 6, 7 from "pigs" to "goat", "cow", "chicken" in that order, then with awk you can read the replacement values into an array and make the replacements using a counter to determine which replacement to use. (below we just fill an array with the three replacements)

You can do something similar to the following:

awk 'BEGIN {n=1; rep[1]="goat"; rep[2]="cow"; rep[3]="chicken"}
    $2=="pigs"{$2=rep[n++]}
    {print}
' print2.txt

Copy/pasting at the command line would yield:

$ awk 'BEGIN {n=1; rep[1]="goat"; rep[2]="cow"; rep[3]="chicken"}
>     $2=="pigs"{$2=rep[n++]}
>     {print}
> ' print2.txt
number elephant
number giraffe
number dogs
number cats
number goat
number cow
number chicken

So you can do what you ask in the comment -- you just need to know what you will replace repeats by in advance.

Answer 2

A pure bash shell solution without any external tools/utility.

!/bin/bash

n=1
while read -r line; do
  printf '%s\n' "${line//number/"$n"}"
  ((n++))
done < print.txt

Using POSIX sh.

#!/bin/sh

n=1
while read -r line; do
  printf '%d. %s\n' "$n" "${line#* }"
  n=$((n+1))
done < print.txt

For a small size/set of data/input it should be fine.