Replacing numbers in a long list by other numbers - bash

Question

I am trying to change numbers in a long list by other numbers. For example,
cat inputfile.txt
A 254 B 456 C 546
D 548 E 548 F 458
A 244 B 416 C 566
D 148 E 558 F 428
And I want to change B's value by adding a percentage on it. For example I want to increase the B in the first array by 3% and b in the next one by 2 % as following:
cat inputfile.txt
A 254 B 469.68 C 546
D 548 E 548 F 458
A 244 B 424.32 C 566
D 148 E 558 F 428
I tried the following but it didn't work.
a=(456 416)
b= (469.68 424.32)
for i in ${a[#]};
for j in ${b[#]}; do
sed -i -- "s/${i}/${j}" inputfile.txt
done

Answer 1

There are multiple problems.
cat inputfile.txt
A 254 B 456 C 546
D 548 E 548 F 458
A 244 B 416 C 566
D 148 E 558 F 428
a=(456 416)
b= (469.68 424.32)
for i in ${a[#]};
for j in ${b[#]}; do
sed -i -- "s/${i}/${j}" inputfile.txt
done
a)
Your assignment to b fails. There is no space allowed around assignment.
b) The substitution pattern isn't closed:
sed -i -- "s/${i}/${j}/" inputfile.txt
c) Now it should run with values not empty, but this will try to replace 456 with 469.68 and then with 424.32. This can't work, since the value is already changed to 469.68. Then it will try to change every 416 with both values.
a=(456 416)
b=(469.68 424.32)
for i in ${a[#]};
for j in ${b[#]}; do
sed "s/${i}/${j}/" inputfile.txt
done
You have two corresponding values which need to be in sync, because you want to replace the first by the second. So you have to iterate once, and by the index:
max=${#a[#]}
for i in $(seq 0 $((max - 1))); do
sed "s/${a[$i]}/${b[$i]}/" inputfile.txt
done
I removed the -i for testing from sed.
The last problem is, that there might be number collisions, for example replacing 456 in 1456 with 469.68 or 416 in 416.02 with 424.32.
To prevent this from happening we can put a blank before the number to match and a boundary for matching blank or line end in the end:
sed "s/ ${a[$i]}\b/ ${b[$i]}/" inputfile.txt
The \b-notation has the advantage (over [ $]) that we don't need to catch it, to push it back in the values, it is non-consuming.
a=(456 416)
b=(469.68 424.32)
max=${#a[#]}
for i in $(seq 0 $((max - 1))); do
sed -i "s/ ${a[$i]}\b/ ${b[$i]}/" inputfile.txt
done
I don't know the source of your a and b values - maybe my following reasoning doesn't apply, but the storage of as and bs in arrays seems not optimal. They have to be of same length but aren't guaranteed to be. You may test it, but if one value gets lost, it's hard to find out where it was, to remove the corresponding b-value.
Not combining all originals and all replacements, binding pairs seems the better idea:
a=(456 469.68)
b=(416 424.32)
But that's not far from the final sed-expression, which would be:
a="s/ 456\b/ 469.68/"
b="s/ 416\b/ 424.32/"
Now that's a bit more verbose, but we save the loop complete:
sed -i "${a};${b}" inputfile.txt
and the input file has only to be read once, and now can be testet without -i in complete.
If you happen to have mass data, you can just generate a file like that:
s/ 456\b/ 469.68/
s/ 416\b/ 424.32/
and name it numcorrect.sed, and call it by:
sed -f numcorrect.sed inputfile.txt