UNIX for Dummies Questions & Answers

Hi,

Which way is faster

Code:

rm -rf /path/

or

Code:

find / -name  -exec rm {} \;

and why?

Code:

rm -rf /path/

will be orders of magnitude faster than

Code:

find / -name  -exec rm {} \;

Which operation do you think is more costly, invoking the rm utility once to remove a million files or invoking the rm utility one million times having each invocation remove one file?

plane 'rm' is faster than with 'find' as find will consume time and 'rm' with '-exec' is slower as it will remove one file at a time...
If you have no other option but to use 'find', I would suggest you use 'xargs' so that all the files will be removed at a time.

Code:

find / -name "*.log" | xargs rm

I suppose - once rm is faster than million times

---------- Post updated at 02:29 AM ---------- Previous update was at 02:28 AM ----------

What about

Code:

find /path/ -type f -delete

is this a fastest method to remove a million files more than 200gb ?

With a million files to be removed, xargs rm will not remove all of the files in one invocation of rm. Furthermore, if any of the files to be removed have names containing whitespace characters, single quotes, double quotes, or backslashes, using xargs that way will not work.

If you need to use find to select a subset of files to be removed by name, size, date, etc., use:

Code:

find /path selection_criteria_operands -exec rm {} +

Using the find utility's -exec command {} \; executes command once for each selected file. Using the find utility's -exec command {} + executes command with groups of one or more file operands chosen such that system ARG_MAX limits will not be exceeded. (This last form matches what xargs tries to do, but doesn't suffer from the filename limitations that you run into by passing filenames through a pipe and having xargs try to find filename boundaries.)

You're comparing apples and oranges:

Code:

rm -rf directory

attempts to remove directory and any links to files (of any type) in the file hierarchy rooted in directory.

On the other hand:

Code:

find directory -type f -delete

attempts to remove only links to regular files in the file hierarchy rooted in directory (leaving the directory structure and all other file types as is). Furthermore, the -delete primary in find is an extension to the standards that is available on some systems, but is not available on other systems.

If actual elapse time is critical because of ongoing processing, could I suggest:-

Code:

mv /path /path-to-del
mkdir -m 777 /path
chown user:group /path

Doing this will be a near instantaneous operation and then with the directory renamed and a new empty one created, your processing can continue whilst you dispose of what is now called /path-to-del at your leisure.

The -m 777 on the mkdir sets the permissions for the directory. The user:group on the chown sets both the owner and group in one operation, rather than having to do a chown and a chgrp


I might be way off course, but I was confused as to why the timing matters so much. If you still want a choice for your original question, I would go with rm -f /path


Another alternate might be available if this is a filesystem you can unmount for a few moments:-

Code:

umount /path
newfs your-options /dev/logical-volume
mount /path

You don't give us an OS or any clue about what devices the disks are on, so I've tried to be generic here. You make have to use mkfs, crfs, fscreate or other things depending what you are running. Of course, this may not be an option.



I hope that these suggestions might help.

Robin
Liverpool/Blackburn
UK