How to compute a thumbnail filename from the shell?

Question

I started noticing some time ago in Xfce4 that when I sent some file to the Trash, tumbler (the Xfce4 thumbnailer) would cause very high I/O load for quite some time. Upon investigating the issue, I found that it was scanning the ~/.thumbnails directory, which was very large in size.

So I decided to write a cron script that will periodically clean the ~/.thumbnails directory, but there is a certain directory of large video files that tumbler takes a bit of time, and sometimes even fails, to create thumbnails for.

The idea is removing all thumbnails, except the ones for these videos. But in order to keep these thumbnails, I have to find what their names are. The problem is that thumbnails are stored named with a md5sum of the URI, plus the PNG extension.

Upon looking at the tumbler source, I found the name for the thumbnail is generated in the following line:

md5_hash = g_compute_checksum_for_string (G_CHECKSUM_MD5, uri, -1);

The documentation for g-compute-checksum-for-string says:

g_compute_checksum_for_string(GChecksumType checksum_type,
                              const gchar *str,
                              gssize length);

checksum_type: a GChecksumType 
str:           the string to compute the checksum of
length:        the length of the string, or -1 if the string is null-terminated.

To put it short, the thumbnail for a file named /home/teresaejunior/File 01.png will be stored in the .thumbnails/ directory as a8502be3273541e618b840204479a7f9.png

According to the ThumbnailerSpec, URI is file://filename. I did some research on the "null character", and thought \0 would do the trick. In order to achieve the result a8502be3273541e618b840204479a7f9, I believed the following should work:

printf "file:///home/teresaejunior/File 01.png\0" | md5sum

but that returns f507285c45d293fa66bc0b813d17b6e6 instead.

Can someone give me some advice? I believe my printf line is flawed. What is my command doing different from g_compute_checksum_for_string?

Answer 1

The NUL character is not included when the MD5 is calculated. Rather, it's the space character that's causing your problem. The filename is URL-encoded:

$ printf '%s' 'file:///home/teresaejunior/File%2001.png' | md5sum
a8502be3273541e618b840204479a7f9  -

Here's one way to do the conversion with Perl:

$ perl -MURI::file -MDigest::MD5=md5_hex \
  -e 'printf "%s.png\n", md5_hex(URI::file->new(shift))' \
  '/home/teresaejunior/File 01.png'
a8502be3273541e618b840204479a7f9.png