Steven W. Orr wrote:
On Sunday, Jun 18th 2006 at 14:14 -0700, quoth Don Russell:
=>I need some basic CLI help. :-) I've googled, and read, and I can't find how
=>to erase a bunch of files in one go.
=>
=>Specifically, I need a command that will erase *.zip files, regardless of the
=>text case of the .zip part....
=>
=>So far, I have
=> ls | grep -iE \\.zip$
=>
=>That gives me the correct list of files.... but I don't know how to get rm to
=>process that list.. It doesn't look like rm has an option to read the file
=>name from stdin, it's expecting the file name as a CLI argument.
=>
=>I know this is basic stuff.... I'm obviously missing some fundamental concept
=>of command line processing. :-)
You have gotten three different suggestions so far.
1. Use xargs
ls | grep -i '\.zip$' | xargs rm
2. Use find with -exec
find . -iname \*.zip -exec 'rm {]' \;
3. Use process substitution.
for i in $(find -iname "*.zip"); do rm $i; done;
Always, wait, should I say that again in reverse, blinking underlined
text?, *ALWAYS* use form 1.
If you don't understand it, read the man page. Use it. Love it.
Use of 1 will run one skinny process.
2. will execute a seperate process for *EACH* file to be deleted.
3. will execute a seperate process for *EACH* file to be deleted and would
potentially overflow before the for loop even starts.
I see how
ls | grep -i '\.zip$' | xargs rm
is the preferred method of the three above....
But... in reading the man pages for xargs, it says this may be
problematic if the file name contain blanks.
What's your opinion of
rm *.[zZ][iI][pP]
Although, if I understand properly what's happening here, the pipe
solution allows files to be erased as they are found in the filelist,
while the gobbing method first requires that all files that match be
found, then each command is executed one after another. If there are too
many files, a buffer could overflow trying to hold the entire list of
globbed file names.
So, the most robust solution seems to be using pipes and xarg, but WITH
the --null option.... the downside being that it further complicates the
pipe line.... How to add the terminating null to the strings coming out
of grep?
Don
