Hello - Hoping somone can help me here :

I have a few thousand files in a few hundred sub-directories.

IE
/categories/cars
/categories/houses
/categories/boats
/categories/business

some of the filenames within these directories contain incorrect characters that I want to just delete

/categories/business/someö_filenme.htm

I want to scan thru the entire directory, and batch delete the special character...

Any idea how to do this from command line?
thanks
-Arron

What characters do you allow? a-z, A-Z, numbers, underscores, spaces(?), etc?
Try something like:
find top_dir -type f > list_of_files
you can then:
-either edit the list_of_files, deleting all files you want to keep
-use "grep" to get the files you don't want to keep (use regular expression)
Something like
grep -e [^a-zA-Z _] list_of_files
should do the trick. [] indicate a set of characters you want to match. Ranges are indicated by "-" (ie a-z
== abcdefghijklmnopqrstuvwxyz), the ^ after the [ tells grep that it should match any lines that have a
character that is NOT in the set (ie an unwanted character).
So, you may want to add additional characters to the set (just add them before the ] but after the ^) and
you will want to check grep's results afterwards.
Finally, you may also need to use grep's -E (extended regular expressions) option. See "man grep" for details
on how to use regular expressions with your version of grep.

Once you have the list of files, the rest is easy (in Bash):
for i in `cat list_of_files`; do
echo "Removing ${i}";
rm ${i};
done;
You may want to try the above loop, commenting the "rm" command first, just to see which files would get deleted. If it mentions only files you want to throw away, run the loop again, this time not commenting the "rm" command.

Hint: "find" will actually scan the directory and all subdirectories, if there are any.
If you don't want to look in the subdirectories, replace the "find" command I gave you with a simple "ls", but you'll need to omit the subdirectories then (for instance with a "if [[ ! -d ${i} ]];" test in the loop).

There is also the tool mmv, once you have understood the syntax, its pretty efficient.
The method with find also works and doesn't need this additionnal tool

sed (streaming editor) will read text (or input in text format from the cli), look for a specific pattern and chenage it to anything elso accoding to your specifications. The basic syntaxis for the command you'd need is:
sed -e 's/<what to look for>/<what to substitute it with/g'
if you combine it with grep as explained in post #2 and pipe the output from grep into sed you should have a good working sollution.

There are also the tools convmv and detox that could help, specially if this is caused by different char sets beetween a Samba server and a windows client.

regards,