need help with script to remove all metachars from filenames

BrianK · 08-08-2005, 02:19 PM

I have a bunch of files on a mac that allows metachars in the file names.. I need to move those files to a linux box that doesn't allow metachars in the file names, so I need to change the file names before they get to the Linux machine.

I'm at a bit of a loss on a couple of things - how to search/replace more than one thing at a time (i.e. search for '[' and ']' and '|' and '?' and replace with '.' all in one operation), and how to recurse directories while doing so.

My intial guess would be change the file names on their way into a tar archive that's piped over ssh from the mac to a linux box, that way the files aren't modified on the mac, but I'm not sure how to do that (or if it's even possible). I can change the file names on the mac if necessary, but always prefer to leave files untouched wherever possible.

Language doesn't matter, I had planned on using a shell script with simple tools like sed, but could easily use a perl script or even a C program if necessary.

Matir · 08-08-2005, 02:31 PM

Something LIKE this: (if you have enough disk space to copy them)

Code:

#!/bin/bash
mkdir COPY
find . -print0 | xargs -0 -i  'cp {} COPY/`echo {} | sed 's/[\[\]\|\?]/./g'`'

I'm not 100% sure that will work, and it's a kludge, but I think it will.

BrianK · 08-19-2005, 10:18 PM

Well, as it turns out, it's not so trivial. The problem is, when you do any sort of find the directory list gets stored at the start of the find, so if you modify a directory name, you no longer have access to that directory and therfore can't recurse down into it. I ended up writing a perl script to do it. I suppose it could have been done any number of other ways, but they all need to manually recurse, rather than relying on some sort of a glob of files from some given start point.

eddiebaby1023 · 08-20-2005, 02:56 PM

Quote:

Originally posted by Matir
Something LIKE this: (if you have enough disk space to copy them)

Code:

#!/bin/bash mkdir COPY find . -print0 | xargs -0 -i 'cp {} COPY/`echo {} | sed 's/[\[\]\|\?]/./g'`'

I'm not 100% sure that will work, and it's a kludge, but I think it will.

It won't. You don't have to escape characters inside [ ]s and if you want to match ] it must the first character after the [. This will do it (I've added * to the list for completeness):

Code:

sed 's/[][|?*\]/./g'

Matir · 08-20-2005, 08:30 PM

oops, thanks for the catch.

Matir · 08-20-2005, 11:10 PM

Quote:

Originally posted by BrianK
Well, as it turns out, it's not so trivial. The problem is, when you do any sort of find the directory list gets stored at the start of the find, so if you modify a directory name, you no longer have access to that directory and therfore can't recurse down into it. I ended up writing a perl script to do it. I suppose it could have been done any number of other ways, but they all need to manually recurse, rather than relying on some sort of a glob of files from some given start point.

I think the important part would be to modify the directory name AFTER recursing it. (Post-order recursion). Pseudocode:

Code:

function handledir{
    cd $1
    for file in *
        do [ -d $file ] && handledir $file
        update_file $file
}