Gfilui

I have the following bash script to make a set of directories, move to that directory, execute some actions (not shown), and move back up to the parent directory:

#!/bin/bash
if [ -e 'dump' ]; then
 rm -r dump
 mkdir dump
 D="dump/var"
else
 mkdir dump
 D="dump/var"
fi

for d in 1..24

do
echo $D$d
mkdir $D$d/
cd $D$d
cd ..

done

When this script is executed, the result is that odd-numbered directories are made while even-numbered directories throw the error 'No such file or directory'. If I put in the option to make a parent directory, as in mkdir -p $D$d/, a nested directory results dump/var1, dump/dump/var2, dump/dump/dump/var3, et cetera.

If I remove the directory changes, cd $D$d and cd .., then the script executes without error. If I use absolute paths for the directory, as in D="/path/to/directory/dump/var", then there is 'No such file or directory' for all but the first directory created.

What is happening in the shell?

Core of the issue: your script cd back to dump, not to dump/'s parent.

On first iteration:

• You start out in dump's parent directory, where dump/ exists


• 
  dump/var1 is created


• you cd into dump/var1.


• when cd .. occurs you go back to dump/. Current working directory is dump and there's only var1 there, nothing else.

On the second iteration:

• When you try mkdir dump/var2/ you are in dump/ and there's only var1 there. The dump/ of dump/var2 is the non-existent path. Of course it fails. Hence the error message and hence the duplicate dump/dump/var2 when you use mkdir -p flag.


• Your script still does cd .., so before 3rd iteration current working directory changes from dump/ to dump/'s parent.

For 3rd iteration:

• You're inside dump/'s parent directory, the path dump/ exists, hence mkdir dump/var3 won't fail.


• 
  cd dump/var3 occurs, then cd .., and what is the current working directory now ? dump/ , the one level above dump/var3, where you try to do mkdir dump/dump/var4, but there's only var1 and var3 there, no dump. The mkdir fails, you cd .. which goes one level above dump/, and the whole thing repeats again.

You can see this clearly when you execute our script with set -x for debugging output appended after #!/bin/bash line or by doing pwd as first command on each iteration of the script.

The way I'd recommend fixing your script is via cd into dump/ first, and then run the for loop. In other words:

cd "$D"
for i in 1..24
do
 mkdir "$i"
 cd "$i"
 # do whatever you want inside "$i"
 cd ..
done

Among other things, I'd suggest using [ -d "dump" ] instead of -e, because that'll help you make sure the file that exists is actually the directory, although this might take extra considerations.