Metadata 0.5

Ruby
1

Konrad M. [email protected] wrote:

Any chance you could wrap this up as a gem?

I already have a gemspec file, but gem screws up bin/chardet by
plastering it with #!/usr/bin/ruby boilerplate (it’s a python file).

And I don’t know how to turn it off.

Another bug (Sorry :D):
$ mdh -p ~/music/Limp\ Bizkit\ -\ Rollin'\ (edited).ogg
sh: -c: line 0: syntax error near unexpected token (' sh: -c: line 0: ogginfo ‘/home/konrad/music/Limp Bizkit - Rollin'
(edited).ogg’’

(Last line was broken up to email length.) You’re already escaping single
quotes for the shell, need to escape start-parens and end-parens as well.

Argh, amateurish mistake on my part, thanks for catching that. Fixed.
If in a bit over-engineered way (creating a safely named link to the
file.)
Probably impossible to safely pass a filename like “-f -i -l -e -z”
to a shell command that doesn’t support “–” in any other way, though.

Also:
For mp3 id3v2 tags, the binary string “\xCB\x99\xC5\xA3” is being inserted
at the front of all the string fields.

[snip]

I think this is an id3v2 thing. Also, it happens in more than one file and
amaroK sees the tags “correctly”, so I’m thinking it’s on the metadata’s
end. Thanks!

Right you are. Fixed. No idea what was causing it. Moved to
using id3lib for the tags (it extracts embedded album art as well!) and
mplayer for the rest of the metadata.

Here we go, 0.5:

tarball: http://dark.fhtr.org/repos/metadata/metadata-0.5.tar.gz
git: http://dark.fhtr.org/repos/metadata

Description

This package Metadata' comes with a library called metadata’ and
a small program called `mdh’.

The library probes files for their metadata (e.g. jpeg dimensions
and camera make, mp3 artist, pdf word count) and returns the metadata
as a Hash.

Mdh can print out file metadata as YAML and package the metadata
with the file.

This package has many dependencies since there is no single universal
metadata header format that all files use. Blame resource forks,
filename
extensions, bags of bytes and mimetypes.

Usage

print out metadata header

mdh -p myfile.jpg

create myfile.jpg.mdh, which consists of metadata header +

myfile.jpg
mdh myfile.jpg

print out metadata header from mdh file

mdh -e -p myfile.jpg.mdh

strip out metadata header from mdh file and save it to myfile.jpg

mdh -e myfile.jpg.mdh

print out list of flags

mdh -h

irb> Metadata.extract(‘myfile.jpg’)
irb> Metadata.extract_text(‘myfile.jpg’)
irb> Pathname.new(“myfile.jpg”).metadata

List of supported formats

Audio:
Whatever you manage to make mplayer play.
Plus FLAC, m4a and wma handled specially.
Successfully tested with:
mp3, flac, ogg, wav
Should also work:
wma, m4a

Video:
Whatever you manage to make mplayer play.
Successfully tested with:
wmv, mov, divx, xvid, flv, ogm, mpg

Images:
Should handle pretty much anything (apart from XCF and ORF.)
Successfully tested with:
jpeg, png, gif, nef, dng, crw, pef, psd

Documents:
Successfully tested with:
pdf, ppt, odp, sxi, ps, ps.gz, html, txt
Should work:
- OpenOffice docs work to some degree (personally, I’m using unoconv
to
convert OO docs to temp PDFs for the text & dimensions extraction,
so
those bits of data are missing.)
- MS Office docs to some degree (ppt at least, doc and xls should
work too,
dimensions missing due to the above temp PDF -thing.)

Others:
Whatever extract spits out on the five or six bits of metadata I’m
using
from it. Archive contents at least.

Requirements

  • Ruby 1.8

  • Tons of metadata extraction programs and libs,
    list of gems:
    flacinfo-rb
    wmainfo-rb
    MP4info
    id3lib-ruby
    list of debian packages:
    dcraw
    libimlib2-ruby
    extract
    libimage-exiftool-perl
    poppler-utils
    mplayer
    html2text
    imagemagick
    unhtml
    pstotext
    antiword
    catdoc
    shared-mime-info

  • You do want to install the latest versions of dcraw and
    shared-mime-info to be able to handle camera raw images.
    http://cybercom.net/~dcoffin/dcraw/
    shared-mime-info

  • Python + chardet library
    http://chardet.feedparser.org/

Install

De-compress archive and enter its top directory.
Then type:

($ su)
# ruby setup.rb

These simple step installs this program under the default
location of Ruby libraries. You can also install files into
your favorite directory by supplying setup.rb some options.
Try “ruby setup.rb --help”.

License

Ruby’s

2

Quoth Ilmari H.:

$ mdh -p ~/music/Limp\ Bizkit\ -\ Rollin'\ (edited).ogg
to a shell command that doesn’t support “–” in any other way, though.

Also:
For mp3 id3v2 tags, the binary string “\xCB\x99\xC5\xA3” is being inserted
at the front of all the string fields.

[snip]

I think this is an id3v2 thing. Also, it happens in more than one file
and
tarball: http://dark.fhtr.org/repos/metadata/metadata-0.5.tar.gz
and camera make, mp3 artist, pdf word count) and returns the metadata
Usage

Documents:
Successfully tested with:
pdf, ppt, odp, sxi, ps, ps.gz, html, txt
Should work:
- OpenOffice docs work to some degree (personally, I’m using unoconv to
convert OO docs to temp PDFs for the text & dimensions extraction, so
those bits of data are missing.)
- MS Office docs to some degree (ppt at least, doc and xls should work
too,

  libimage-exiftool-perl
  • You do want to install the latest versions of dcraw and
    De-compress archive and enter its top directory.

License

Ruby’s

–
Ilmari H. <ilmari.heikkinen gmail com>
http://fhtr.blogspot.com

Another bug, here we go:

undefined method audio_x_vorbis_ogg' for Metadata:Module undefined method audio_x_vorbis_ogg’ for Metadata:Module
/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:365:in
video_x_theora_ogg' /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:142:in send’
/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:142:in `extract’

That code that seems to be failing is:

def video_x_theora_ogg(filename, charset)
h = video(filename, charset)
wma = audio_x_vorbis_ogg(filename, charset)
%w(
Artist Title Album Genre ReleaseDate TrackNo VariableBitrate
).each{|t|
h[‘Video.’+t] = wma[‘Audio.’+t]
}
h
end

This makes sense, as audio_x_vorbis_ogg() doesn’t exist anywhere else.
:smiley:

3

Quoth Ilmari H.:

–
Ilmari H.
http://fhtr.blogspot.com

As a matter of fact (I only had to kill mplayer 5-7 times):
Genre
Name
Artist
Creation Date
Album
Track
Title
Year
Comment
name
author
Comments

HTH,

4

On 9/18/07, Konrad M. [email protected] wrote:

Just FYI – I’ve been running your script since about 6.5 hours ago, I’ll
reply again when it’s actually done. So far no problems.

Big library, eh :slight_smile:
If it’s still running, I doubt it’s going to do much more…

What has it printed out?

Thanks!

5

Quoth Ilmari H.:

–
Ilmari H.
http://fhtr.blogspot.com

Ok, now it successfully runs over my entire collection. Yay! Now it’s
time
to fix all those tag-less songs.

Thanks much,

6

Quoth Ilmari H.:

$ mdh -p ~/music/Limp\ Bizkit\ -\ Rollin'\ (edited).ogg
to a shell command that doesn’t support “–” in any other way, though.

Also:
For mp3 id3v2 tags, the binary string “\xCB\x99\xC5\xA3” is being inserted
at the front of all the string fields.

[snip]

I think this is an id3v2 thing. Also, it happens in more than one file
and
tarball: http://dark.fhtr.org/repos/metadata/metadata-0.5.tar.gz
git: http://dark.fhtr.org/repos/metadata

…

–
Ilmari H. <ilmari.heikkinen gmail com>
http://fhtr.blogspot.com

Thanks, trying it out now. (I’m basically running it on every file
in my collection and running back to you when I get errors. :D)

7

On 9/16/07, Konrad M. [email protected] wrote:

http://dark.fhtr.org/repos/metadata/metadata-0.6.tar.gz
from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:153:in `extract’

Not sure what that is, and frankly atm my brain is a bit too weak to think
about it. But you should be fresh and able to solve that.

Apparently temporary hardlinks weren’t such a hot idea after all. Nuts.

Ok, now escaping filename by default, only trying to “ln rescue cp” for
filenames starting with a dash. Running it against my downloads-dir
presently, been working ok thus far. YMMV of course :slight_smile:

http://dark.fhtr.org/repos/metadata/metadata-0.7.tar.gz

8

Quoth Ilmari H.:

On 9/16/07, Konrad M. [email protected] wrote:

Another bug, here we go:

undefined method `audio_x_vorbis_ogg’ for Metadata:Module

This makes sense, as audio_x_vorbis_ogg() doesn’t exist anywhere else. :smiley:

Fixed. And 0.6 :slight_smile:
http://dark.fhtr.org/repos/metadata/metadata-0.6.tar.gz

Ooh, here’s another: :slight_smile:

/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:735:in unlink': No such file or directory - _tmp_metadata_temp_22720__604265598_1189946590.27022 (Errno::ENOENT) from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:735:in secure_filename’
from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:590:in
extract_extract_info' from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:153:in extract’

Not sure what that is, and frankly atm my brain is a bit too weak to
think
about it. But you should be fresh and able to solve that.

9

On 9/18/07, Konrad M. [email protected] wrote:

Thanks!
Track
Title
Year
Comment
name
author
Comments

Alright, thanks a lot! I was missing ‘name’ and ‘author’, the rest I
had already.

10

On 9/17/07, Konrad M. [email protected] wrote:

else. :smiley:

    from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:735:in

Video.Artist: Korn
Album: Greatest Hits Vol. 1
Track: 16

Ah, it uses Name instead of Title. Thanks!
Added it and made 0.8.

http://dark.fhtr.org/repos/metadata/metadata-0.8.tar.gz

Now I wonder what other synonyms mplayer uses…
I’d really appreciate it if you could run the following over
your media library and tell what field names it spews out:

find $MEDIA_LIBRARY_DIR -type f |
mplayer -identify -ao null -vo null -frames 0 -playlist - |
grep ID_CLIP_INFO_NAME | sed ‘s/^.*=//’ | sort | uniq

(replace $MEDIA_LIBRARY_DIR with the directory name)

Thanks again,

11

Quoth Ilmari H.:

On 9/16/07, Konrad M. [email protected] wrote:

Quoth Ilmari H.:

On 9/16/07, Konrad M. [email protected] wrote:

Another bug, here we go:

undefined method `audio_x_vorbis_ogg’ for Metadata:Module

This makes sense, as audio_x_vorbis_ogg() doesn’t exist anywhere
else. :smiley:

Fixed. And 0.6 :slight_smile:
http://dark.fhtr.org/repos/metadata/metadata-0.6.tar.gz

Ooh, here’s another: :slight_smile:

/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:735:in unlink': No such file or directory - _tmp_metadata_temp_22720__604265598_1189946590.27022 (Errno::ENOENT) from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:735:in secure_filename’
from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:590:in
extract_extract_info' from /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:153:in extract’
http://dark.fhtr.org/repos/metadata/metadata-0.7.tar.gz
The title tag isn’t being parsed out of oggs:

$ mdh -p music/korn_-_clown.ogg
Video.TrackNo: 16
Video.Artist: Korn
Video.Genre: Hard Rock
Video.Album: Greatest Hits Vol. 1

vs mplayer:

Ogg file format detected.
Clip info:
Genre: Hard Rock
Name: Clown
Artist: Korn
Album: Greatest Hits Vol. 1
Track: 16

Cheers,

12

On 9/16/07, Konrad M. [email protected] wrote:

Another bug, here we go:

undefined method `audio_x_vorbis_ogg’ for Metadata:Module

This makes sense, as audio_x_vorbis_ogg() doesn’t exist anywhere else. :smiley:

Fixed. And 0.6 :slight_smile:
http://dark.fhtr.org/repos/metadata/metadata-0.6.tar.gz

13

Quoth Ilmari H.:

On 9/18/07, Konrad M. [email protected] wrote:

Quoth Ilmari H.:

On 9/18/07, Konrad M. [email protected] wrote:

Just FYI – I’ve been running your script since about 6.5 hours ago,
I’ll
http://fhtr.blogspot.com
Comment
name
author
Comments

Alright, thanks a lot! I was missing ‘name’ and ‘author’, the rest I
had already.

Alright, glad to help.

14

On 9/17/07, Konrad M. [email protected] wrote:

Quoth Ilmari H.:

find $MEDIA_LIBRARY_DIR -type f |
mplayer -identify -ao null -vo null -frames 0 -playlist - |
grep ID_CLIP_INFO_NAME | sed ‘s/^.*=//’ | sort | uniq

(replace $MEDIA_LIBRARY_DIR with the directory name)

I’d love to run that but mplayer dies rather early on on some of my files.

Hmm, here’s a ruby version that should work through those:

media_dir = “music”
seen_names = {}
mpc = “mplayer -identify -ao null -vo null -frames 0 -playlist -
2>/dev/null”
Dir[“#{media_dir}/**/*”].each{|fn|
if File.file?(fn)
IO.popen(mpc, “r+”){|mp|
begin
mp.puts fn
mp.close_write
tags = mp.read.strip.split(“\n”).grep(/^ID_CLIP_INFO_NAME/)
names = tags.map{|t| t.split(“=”, 2)[1] }
names.each{|n|
seen_names[n] ||= (puts n; true)
}
rescue
end
}
end
}

I’d guess one of the libraries you’re using for parsing is giving back 40
as a genre or track number (a bit high, but might be tagged wrong) and it
needs to be converted to a string before you can use it.

Good catch, thanks. Fixed.
http://dark.fhtr.org/repos/metadata/metadata-0.9.tar.gz

15

Quoth Ilmari H.:

This makes sense, as audio_x_vorbis_ogg() doesn’t exist anywhere
else. :smiley:

Fixed. And 0.6 :slight_smile:
http://dark.fhtr.org/repos/metadata/metadata-0.6.tar.gz

Ooh, here’s another: :slight_smile:

/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:735:in `unlink’: No
such
Not sure what that is, and frankly atm my brain is a bit too weak to
think
The title tag isn’t being parsed out of oggs:
Clip info:
http://dark.fhtr.org/repos/metadata/metadata-0.8.tar.gz

Thanks again,

Ilmari H.
http://fhtr.blogspot.com

I’d love to run that but mplayer dies rather early on on some of my
files.
Also, seems like we have another bug (not sure what kind of file it’s
on,
sorry):

undefined method empty?' for 40:Fixnum undefined method empty?’ for 40:Fixnum
/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:765:in enc_utf8' /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:352:in video’
/usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:142:in __send__' /usr/lib/ruby/site_ruby/1.8/metadata/extract.rb:142:in extract’

I’d guess one of the libraries you’re using for parsing is giving back
40
as a genre or track number (a bit high, but might be tagged wrong) and
it
needs to be converted to a string before you can use it.

Thanks!

16

Quoth Ilmari H.:

–
Ilmari H.
http://fhtr.blogspot.com

Well, I’m having it run over all of them, then print the output. But
sometimes mplayer just starts eating 100% cpu and doesn’t exit, so the
script gets stuck on the one song.

17

Quoth Ilmari H.:

Hmm, here’s a ruby version that should work through those:

media_dir = “music”
seen_names = {}
mpc = “mplayer -identify -ao null -vo null -frames 0 -playlist -
2>/dev/null”
}
Good catch, thanks. Fixed.
http://dark.fhtr.org/repos/metadata/metadata-0.9.tar.gz

Thanks!

Konrad M. [email protected] http://konrad.sobertillnoon.com/

Just FYI – I’ve been running your script since about 6.5 hours ago,
I’ll
reply again when it’s actually done. So far no problems.

HTH,

18

Ilmari H. wrote:

tarball: http://dark.fhtr.org/repos/metadata/metadata-0.5.tar.gz
git: http://dark.fhtr.org/repos/metadata

These links don’t work. Is there somewhere else I can find this project?
Is it available as a gem?

Description

This package Metadata' comes with a library called metadata’ and
a small program called `mdh’.

The library probes files for their metadata (e.g. jpeg dimensions
and camera make, mp3 artist, pdf word count) and returns the metadata
as a Hash.

Mdh can print out file metadata as YAML and package the metadata
with the file.

This package has many dependencies since there is no single universal
metadata header format that all files use. Blame resource forks,
filename
extensions, bags of bytes and mimetypes.

Thanks.