find command with wildcard directory

Shell Programming and Scripting
1

I want to look if there is any file inside a specific directory which was modified before 2 days.

I wrote the find command, but the problem is there is one directory and that is a random directory generated by unix, so not sure on how to code for that on the find command.

find /Production/ST/st*/Outbound/Prod/PROD-*/bcfa0dbbcf17f28c768d1d34da6b48a4/PGP -name '*.*' -mtime +2
find /Production/ST/st*/Outbound/Prod/PROD-*/63acf2caf91bc136cb9bcce8a85c7fa8/PGP -name '*.*' -mtime +2

the folders bcfa0dbbcf17f28c768d1d34da6b48a4 and 63acf2caf91bc136cb9bcce8a85c7fa8 are random and generated by unix.

there are several hundreds of folders like this, so I want a generic soultion to code for the random directory

any help would be appreciated as this is urgent.

2

I doubt UNIX generates them, I think some application's probably doing so.

"*.*" is a DOS thing, what it amounts to for find isn't "find all files" but "find all files or folders with . in their name". If you want to find all files, leave off -name and try -type f (to limit it to files, by default it prints files and folders alike)

What is your system?

How about:

# Find all directories in .../PROD-* which are exactly 32 chars long
find /Production/ST/st*/Outbound/Prod/PROD-* -type d -name '????????????????????????????????' -print -name '*' -prune
while read DIR
do
        find "${DIR}/PGP" -type f -mtime +2
done

The -name '*' prune tells it to not recurse deeper into those folders, which is the job of the second find.

3

yes yours should work and I also found an other way and changed the code as suggested by you.

find /Production/ST/st*/Outbound/Prod/PROD-*/[abcdefghijklmnopqrstuvwxyz]*/PGP -type f -mtime +2 -exec rm -f '{}' \;

find /Production/ST/st*/Outbound/Prod/PROD-*/[0123456789]*/PGP -type f -mtime +2 -exec rm -f '{}' \;

so the above code will delete the files under folder PGP.

after deleting the files inside PGP I also want is to delete the PGP folder then remove the folder ??????????????????????????????? also.

I can delete the PGP folder using rmdir PGP

I want to know how to handle the 32char folder?

4

If you'd only said "dozens" I'd have have suggested this in the first place, but this is dangerous when you have lots of files and folders. There's a limit to how many things one glob can find -- in some shells, no more than a page or two worth.

The version with two finds has no limit at all.

You're only deleting some of the files. If they're not empty, I doubt you really want them deleted.

You'll need to use the two-find version to do this anyway, since it'd be torturous to get the right directory in one find and use it only once. I'll use 0-9, a-f if that worked for you.

# Find all directories in .../PROD-* beginning with [0-9a-f]
find /Production/ST/st*/Outbound/Prod/PROD-* -type d -name '[0-9a-f]*' -print -name '*' -prune
while read DIR
do
        # Find files in "${DIR}/PGP", delete them if old enough
        find "${DIR}/PGP" -type f -mtime +2 | xargs -d '\n' echo rm

        # Remove these directories only if they're empty
        rmdir "${DIR}/PGP" && rmdir "${DIR}"
# the rmdir's will cause some error messages when they fail, redirect that to /dev/null
done 2> /dev/null

find | xargs rm will run 'rm file1 file2 file3 ...' where find -exec rm would run 'rm file1; rm file2 ; rm file3 ...' so xargs makes it much faster. The -d '\n' is to tell xargs to consider anything but newlines as part of the filename.

The 'echo' is just a test, to print filenames instead of deleting as a test. Remove it once you're sure it's doing what you wanted.

5

can you tell me a bit more about two finds?

so in my case do I have to give it as 0-9,a-z as it could be anything between 0-9 and a-z as the first letter of the folder. is there any restriction on the # of directories I can search using this wildcard?

also won't the rmdir "${DIR}" delete the full structure? I just need only till the 32 char folder deleted and the rest should be intact.

P.S:
I just confirmed that it will be 32 chars and it will be consistent? so can i use the ??? approach? would that be better?

6

I think I forgot a pipe. No wonder it was puzzling you.

# Find and print all directories in .../PROD-* beginning with [0-9a-f]
find /Production/ST/st*/Outbound/Prod/PROD-* -type d -name '[0-9a-f]*' -print -name '*' -prune |
# Read lines like /Production/ST/.../bcfa0dbbcf17f28c768d1d34da6b48a4 into DIR, one-by-one
while read DIR
do
        # We feed that whole directory into another find, telling it to look
        # for files.  It prints them one by one, piped into xargs,
        # which uses them as arguments for rm.
        find "${DIR}/PGP" -type f -mtime +2 | xargs -d '\n' echo rm
 
        # Remove these directories only if they're empty
        rmdir "${DIR}/PGP" && rmdir "${DIR}"
# the rmdir's will cause some error messages when they fail, redirect that to /dev/null
done 2> /dev/null

If that didn't answer your question you'll need to ask about something specific.

I don't think so, those look like hex: binary numbers encoded as ASCII digits 0-9 and a-f.

There's no limit on find -name '*', no, because it can process them one-by-one. When you expand * inside the shell itself, it has to find all of them at once, before find is even run, potentially running out of room.

"rmdir /path/to/folder" only removes 'folder'. And it won't even do that unless it's empty. So I try to remove /Production/ST/st*/Outbound/Prod/PROD-/bcfa0dbbcf17f28c768d1d34da6b48a4/PGP, and if that succeeds, I try to remove /Production/ST/st*/Outbound/Prod/PROD-/bcfa0dbbcf17f28c768d1d34da6b48a4 itself.

And as I said, don't remove the echo until you're sure it does what you want.