Recursively moving contents of directory

Question

I wanted to recursively move files from a folder (Pictures) to another (Picturesnew).

The "Pictures" Folder had many subfolders and and hence I used this command after following up the posts here.

Both Pictures and Picturesnew were in the same directory. I just wanted to get rid of all the subfolders and combine the data.

I ran the following command from the directory these folders were situated in.

find ./Pictures -type f -name "*.jpg" -print0 | 
xargs -0 -Imysongs mv -i mysongs ./Picturesnew

Now seemingly the Picturesnew folder which should have appeared didn't appear at all and hence I am confused as to where 20000 JPG files of mine went.

Answer 1

I don't know how come there was no confirmation while overwriting, which should be present due to the -i parameter to mv. But assuming you didn't use it and given there was a jpg file Picturesnew after the operation, as you state in the comment under your question, well - you overwrote the same file 20000 times.

xargs -0 -Imysongs mv -i mysongs ./Picturesnew

This means for each argument this command will be executed:

mv -i argument ./Picutesnew

Since there was no directory ./Picturesnew the command mv thought you mean it should move the file here and rename to ./Picturesnew. It did so 20000 times.

After the pipeline has finished executing, there should be just one file ./Picturesnew left. The last one that was moved. The previous ones are gone.

This form is therefore much safer:

mv -i argument ./Picturesnew/

That's because the / at the end eliminates confusion as to what Picturesnew is supposed to be.

As for -i, you wouldn't put it in this pipeline, as it wouldn't work. It demands confirmation.

Answer 2

You really wanted

mkdir Picturesnew
find Pictures -type f -iname "*.jpg" -exec mv -i {} + Picturesnew/

mv does not make directories.

The -i option to mv causes it to not work when run in a pipeline like that. (Though, on the bright side, it hopefully told mv to not stomp all over the first file renamed to be Picturesnew instead of whatever it was originally.)

In my solution, I'm telling find to move multiple files per instance of mv executed if useful, to minimize the chance that a mistake such as leaving the trailing / off of the target directory name and not having made the target directory would actually rename a file, rather than generating the error desired. It also has the advantage of allowing mv to actually prompt for confirmation for overwriting a file rather than simply taking the next filename as "no".

Answer 3

find Pictures -type f -iname "*.jpg" -print0 | xargs -0 -I {} mv -n "{}" Picturesnew/

Would be a more syntactically correct solution, so restore your backup and try that.

Answer 4

I think you can achieve your goal with following command:

find Pictures -type f -print0 | xargs -0 mv -t Picturesnew

where Pictures is source directory and Picturesnew is destination directory.

Both mv (and cp) have format mv -t directory source... which fits good in use with xargs.

But this method will leave files with duplicate names in their original positions (maybe that's not bad since you can review them and copy to the desitnation later), since mv -t won't work interactively when input redirection is used. That is because xargs already reads values from redirected standard input (which is output of find), and mv, which is ran by xargs, tries to read answers from the same standard input too. So you cannot use mv with pipes in interactive fashion.

Maybe job can be solved by some clever input-handle reassigning, but let's try more simple way: get rid of output redirection. Also we cannot use -print0 in command substitution because it doesn't allow 0-bytes. We need to iterate over filenames with for, using newline as input-field separator:

TMP=$IFS; 
IFS=$'\n'; 
for i in $(find Pictures -type f); do 
    mv -it ../Picturesnew $i; 
done; 
IFS=$TMP

(you may copy it as a one-liner).

In this case, mv will not only ask you for each duplicate, but wait for answer as well.