Shell Programming and Scripting

So, I need to find a bunch of files and delete them (this example, but sometimes I need it for something else) and my trusty go-to command has always been:

Code:

find . -type f -name '*file*' | xargs -I## rm '##'

Works wonders... But:

Code:

touch file\ file\'.txt
touch file.txt
touch file\ file.txt
touch file\".txt
touch file\'\ \".txt

and run the command above it will produce the following error:

Code:

xargs: unmatched single quote; by default quotes are special to xargs unless you use the -0 option

Now... I cannot get the -0 option to work properly, can someone help?

Why use xargs? Why not just do it in find:

Code:

find . -type f -name '*file*' -exec rm {} +

Completely valid mr. Cragun. I agree, too often I forget that the find command has a built in exec possibility and in this example it is valid, however it's not always valid if you for instance need something like:

Code:

find . -type f -name "*.jpg" | grep -i summer | xargs -I## mv '##' ./SummerPictures/

in this case I need the xargs (or in a more complicated case than this, it was the first fast idea I got). Anyways, I would love to find how I take care of ' and " and spaces in xargs and still get it to work.

Thank you for your solution in this example though

you can eliminate the pipe to grep if

Code:

find . -iname \*summer\*.jpg -type f

Concerning your first post

Code:

find . -type f -name '*file*' -print0 | xargs  -0 -I file rm file

The -0 option to xargs and both -iname and -print0 primaries in find are extensions to the standards that are not available in several implementations. This is a little harder to write, but is both portable and more efficient:

Code:

find . -type f -name '*[sS][uU][mM][mM][eE][rR]*[.][jJ][pP][gG]' -exec mv '{}' ./SummerPictures/ \;

This works even when filenames contain spaces, tabs, quotes (of any kind), and even newlines (which should never be included in a filename, except in examples explaining why newlines should never be used in filenames).

@Don Cragun

Out of the nth times you have written the same about find and the standards, you have never said why it is more efficient. Next time, maybe?

In the shell, and in UNIX and Linux in general, fork and exec are relatively expensive operations. So:

Code:

find . -type f -name name

is more efficient than:

Code:

find . -type f | grep name

unless there is some reason to believe that grep is so much faster at processing a regular expression to match a filename than find that the overhead needed to write many more filenames into a pipe, fork and exec grep, and have grep read those filenames from the pipe, and perform the same regexp processing that find would have performed; doing it all in find will use fewer processes, pass less data between processes, and use less system memory than using find and grep to do the same work.

On systems that don't have find -print0 and xargs -0, piping a list of filenames that might contain whitespace and/or quote characters is not only less efficient, it is somewhere between harder and impossible to reconstruct the filenames that need to be processed while there is absolutely no problem figuring what filename should be used in find -exec command '{}' + and in find -exec command '{}' other operands \;.

That is why it is usually much more efficient to do all of the processing in find that you can do in find instead of firing up multiple processes in a pipeline to do the same work.