scripts/file_utils/README.md

dirpacker.py
============

This scripts allows you to group (pack) a bunch of files or directories into
fixed-size volumes, optimizing the occupied size of the volumes. The original
use case was to burn MP3 albums to CD-ROMs to play them in the car, while
minimizing the amount of wasted space on each disc. Of course, it can be used
to backup any kind of files to any kind of medium.

The particularity of this program, compared for example to datapacker (from
which it is loosely inspired) is that it works with directories instead of
regular files only. The files inside a directory won't be separated on several
archives, they will be on the same volume.

Note: I kept the datapacker's terminology in which a volume is also called a
bin.

The default bin size is 703 MiB, i.e. the capacity of a 80-minute CD-ROM; note
that this can lead to slightly exceed the capacity when actually burning the
disc, so if you don't want to overburn a few hundreds KiB, choose a lower bin
size (702 MiB should be fine).

By default, dirpacker displays a list of bins and the files they would contain,
along with the size of each file and some statistics. With the option
`--machine-readable`, this list will be printed in a machine-readable format:
each line contains a bin's name, then a tabulation, then a file that belongs to
this bin.

With the option `--move`, a directory will be created for each volume and the
files will be moved to the corresponding volume.

To see the full usage, call the program with `-h`.


mvparent.sh
===========

`mvparent.sh` was originally written to be integrated in the ROX-Filer
directory “Send to” menu. It moves the contents of a directory into its parent
directory, then deletes the empty directory.

To install this script, just copy it in a directory which is in your `PATH`, or
use the `bin` directory provided with the repository, as explained in the main
`README.md`.

If you want to integrate it in ROX-Filer:

    mkdir -p ~/.config/rox.sourceforge.net/SendTo/.inode_directory
    cp mvparent.sh ~/.config/rox.sourceforge.net/SendTo/.inode_directory/"Move here"


prefixsubdirs.sh
================

The use case for this script is to flatten a file hierarchy, for instance
switching from:

    dir1
    |-- dirA
    |-- dirB
    `-- dirC

to:

    dir1 ; dirA
    dir1 ; dirB
    dir1 ; dirC

However, `prefixsubdirs.sh` will only rename the subdirectories (not the
regular files) within the given directories, it will not actually flatten the
hierarchy. You can use `mvparent.sh` for this purpose (see above).


unln.py
=======

I wrote this script after accidentally merging all my “duplicate” files with
the cleaning tool `fslint` (thanks guys for a totally confusing UI!). It does
the opposite of the `ln` command, that is it separates file names given as
arguments from their respective inodes by doing the equivalent of `cp -p file
tmp && mv tmp file`. Checks are done in order to work only on regular files
that have more than one hard link, so the minimal amount of copies are done.

A summary of the file names on which errors were raised is displayed at the end
of the execution. Note that if a file is under a directory which is not
executable (i.e. the program can't `cd` in), it will be considered as
non-existent and ignored, and its name won't be displayed.

/!\ **Warning #1**: the Python copy function used is `shutil.copy2()`, which
tries to preserve all the metadata it can, but as of Python 3.3 it is not
guaranteed that extended attributes will be preserved. See the documentation
for details: <http://docs.python.org/3/library/shutil.html>.

/!\ **Warning #2**: file owner and group cannot be preserved, as a file is
always created (upon copy, in our case) with the current user's UID and GID.

**Tip**: if you have a file listing all the file names you want to work on,
with one name per line, you can use `xargs` with `--delimiter='\n'`:

    xargs --delimiter='\n' <list.txt unln.py

(Without this option, you would have troubles with file names containing
spaces.)