Hfrhyu

I have a csv file that looks like:

data/train/4/36280.png,four
data/train/2/10317.png,two
data/train/2/57890.png,two
data/train/1/53448.png,one
data/train/8/58233.png,eight
data/train/4/23599.png,four
data/train/2/35051.png,two
data/train/1/12323.png,one
data/train/9/18562.png,nine
data/train/8/46629.png,eight
data/train/7/1746.png,seven

Where first column is path and second is class. I would like to change the second column conditionally. In pseudo code I want something like:

If second column "four" change it to the next row's class

So far I've tried just to see if I can change ne class to another, but without success:

awk ' if ($2 == "zero") $2="one"; print $0 ' train.csv > new_file.csv

gives me the same csv.

Basically I would like a script that gives me following output csv:

data/train/4/36280.png,seven
data/train/2/10317.png,four
data/train/2/57890.png,two
data/train/1/53448.png,two
data/train/8/58233.png,one
data/train/4/23599.png,eight
data/train/2/35051.png,four
data/train/1/12323.png,two
data/train/9/18562.png,one
data/train/8/46629.png,nine
data/train/7/1746.png,eight

bash solution:

readarray -t filename < <(cut -d ',' -f1 train.csv)
readarray -t class < <(cut -d ',' -f2 train.csv)

for (( i=0; i<$#filename[@]; i++ )); do
 printf '%s,%sn' "$filename[$i]" "$class[$((i-1))]";
done > new_file.csv

Equivalent awk solution:

awk -F, '
 
 filename[NR]=$1
 class[NR]=$2
 
 END 
 OFS=","
 print filename[1],class[NR]
 for (i=2;i<=NR;i++) 
 print filename[i],class[i-1]
 
 
' train.csv > new_file.csv

Both solutions first read the lines into arrays filename and class. Only difference is that the bash array starts with 0, the awk array with 1.

Then, we loop over the arrays and print the desired output. In the awk solution, we need to treat the first line separately as, unlike in bash, array[-1] does not return the last element.

The awk solution is a bit faster.