directory tree size browsing with column view

[Michael Brand <michael.brand@alumni.ethz.ch>]

Hi all,

I would like to use the column view for browsing a hierarchical tree with a summed up property like e. g. the directory sizes (inclusive cluster waste and subdirectories from `du -sk`) of a directory tree. Just similar (only unsorted) to the upper left pane of this GUI screenshot of WinDirStat here
http://windirstat.info/images/windirstat.jpg

Let me make an example how I think to do the same with column view when given a directory with the two shell outputs

----------------------------------------
$ ls -lR dir_node
total 16
drwxr-xr-x  2 usr  grp      136 Oct 24 12:00 a
drwxr-xr-x  2 usr  grp      102 Oct 24 12:00 b
-rw-r--r--  1 usr  grp     9070 Oct 24 13:00 t.txt

dir_node/a:
total 2400
-rw-r--r--  1 usr  grp     4535 Oct 24 12:00 x.txt
-rw-r--r--  1 usr  grp  1217312 Oct 24 12:00 y.txt

dir_node/b:
total 16
-rw-r--r--  1 usr  grp     4535 Oct 24 12:00 z.txt
----------------------------------------

----------------------------------------
$ find dir_node -type d -exec du -sk {} \;
1220	dir_node
1200	dir_node/a
8	dir_node/b
----------------------------------------

For this I would like to have `some software' which uses the path to this directory `dir_node' as input and produces the following output file

----------------------------------------
-*- mode: org; eval: (org-columns) -*-
#+STARTUP: odd hidestars
#+PROPERTY: size
#+COLUMNS: %20ITEM %size{+}

* dir_node
*** a
    :PROPERTIES:
    :size: 1200
    :END:
    -rw-r--r--  1 usr  grp     4535 Oct 24 12:00 x.txt
    -rw-r--r--  1 usr  grp  1217312 Oct 24 12:00 y.txt
*** b
    :PROPERTIES:
    :size: 8
    :END:
    -rw-r--r--  1 usr  grp     4535 Oct 24 12:00 z.txt
*** .
    :PROPERTIES:
    :size: 12
    :END:
    -rw-r--r--  1 usr  grp     9070 Oct 24 13:00 t.txt
----------------------------------------

Opening this file in org-mode using column view now is comfortable to browse and can look e. g. like here

----------------------------------------
ITEM                 | size |
* dir_node           | 1220 |
*** a                | 1200 |
    :PROPERTIES:...
    -rw-r--r--  1 usr  grp     4535 Oct 24 12:00 x.txt
    -rw-r--r--  1 usr  grp  1217312 Oct 24 12:00 y.txt
*** b                | 8    |...
*** .                | 12   |...
----------------------------------------

where all not of interest can be left hidden and one can see easily that
  - the directory `a' is by far the largest
  - most of the disk usage is caused only by the file dir_node/a/y.txt.
  - etc

I would be able to implement this `some software' in a shell script (if I would take the necessary time some day) but not within Emacs itself which would be easier to use and IMHO would not be concerned that much with portability issues.

What do you think?