Had a bunch of python scripts in a directory. They all needed some munging. It was too complicated for my sed skills, so I wrote a python script to munge the python scripts.
Here's how I ran it: for f in UC*.py; do cp $f $f.orig; ./munge $f.orig > $f; echo $f; done Only problem was, the script was called munge.py, not munge, so I got a whole bunch of: bash: ./munge: No such file or directory Oops. up-arrow twice to previous command and change to ./munge.py: for f in UC*.py; do cp $f $f.orig; ./munge.py $f.orig > $f; echo $f; done Just about to hit return, when I decided to quickly check the directory listing. Yep. Every file matching UC*.py was of zero length. If I had run the correct version, It would have copied the zero-length version over the $f.orig version, and all would have been lost! Except, of course, the files are in a version control system, so all would not have been lost. Moral of the story: The shell interprets shell redirects, not the program. Even if the program fails, the redirect (such as > foo) will create a zero-length foo, clobbering any existing foo. I think you can turn off this useful feature with a noclobber option in the shell. Second Moral: Backups or version control systems are a Good Idea. Cheers, Carl.
