Pimp My XSLT
Deeply nerdy stuff here. I attempt to impress you all with my l337 XSLT skillz.
Slicing and dicing iTunes playlists. Lets say you want to extract a list of files for the songs in a given playlist. Maybe you want to see which files are stored where. I dunno. Anyway, how would you do it?
If you’re sane, you probably reach for Perl, Python or, my (and everyone else’s) new infatuation, Ruby, to digest the iTunes Music Library.xml
file. These tools can slice’n’dice with the best of them. And for a quick and dirty hack, I’m right there with you.
But here’s some food for thought. Check out the following XSLT script.
<?xml version="1.0"?> <xsl:stylesheet version="1.0" id="stylesheet" xmlns:xsl="http://www.w3.org/1999/XSL/Transform"> <xsl:output method="text"/> <xsl:param name="playlist" select="'Library'"/> <xsl:key name="playlists-by-name" match="/plist/dict/array[preceding-sibling::key[1]='Playlists']/dict" use="string[preceding-sibling::key[1]='Name']"/> <xsl:key name="tracks-by-id" match="/plist/dict/dict[preceding-sibling::key[1]='Tracks']/dict" use="integer[preceding-sibling::key[1]='Track ID']"/> <xsl:template match="/"> <xsl:apply-templates select="key('playlists-by-name',$playlist)" mode="playlist"/> </xsl:template> <xsl:template match="dict" mode="playlist"> <xsl:for-each select="array[preceding-sibling::key[1]='Playlist Items']/dict"> <xsl:variable name="trackid" select="integer"/> <xsl:apply-templates select="key('tracks-by-id',$trackid)" mode="track"/> </xsl:for-each> </xsl:template> <xsl:template match="dict" mode="track"> <xsl:value-of select="string[preceding-sibling::key[1]='Location']"/><xsl:text> </xsl:text> </xsl:template> </xsl:stylesheet>
Go ahead and try it! Save it to your iTunes directory as playlist.xsl
(or whatever) and apply your favourite XSLT processor. MacOS users have xsltproc
available by default:
xsltproc --novalid playlist.xsl iTunes\ Music\ Library.xml
(To print a different playlist, use --stringparam playlist "Other Playlist"
)
Now maybe I’m just deluding myself, but I think this is quite readable. For XSLT anyway. Here’s what it’s doing.
- The
xsl:key
constructs are creating hash tables on the input document. As the names imply they are indexing playlists and tracks. - The
/
template is the “start” of the stylesheet, and it selects the playlist to process using thekey()
XPath function. - The “playlist” template iterates over the tracks which are referenced by ID. We look up the playlist items and iterate over them with the
xsl:for-each
construct. We select the track to process by again using thekey()
function. - Finally, the “track” template outputs the location.
The readability of all this is impacted by the preceding-sibling::key[1]
stuff. And this is mainly because Apple chose to use a generic Property List vocabulary rather than one which is specific to iTunes. Incidentally, I came across this blog which explained how to convert from a property list into a more comprehensible format. I happen to think that the property list is quite usable as-is, provided that you employ the use of xsl:key
as above.
In fact, I think this use of xsl:key
is the deal-maker. Some of the other attempts to process iTunes Libraries with XSLT don’t seem to take advantage of this, and I think it really helps.
Notwithstanding the workarounds for the weird XML vocabulary, I would (again with my jaundiced eye) judge this to be quite a maintainable piece of code. If you had to change it to operate on a given album or a set of tracks from a genre, or even output a different field from each track, I’m guessing that this would be pretty easy, even for XSLT n00bs. Also if Apple decide to change the output format by (for instance) rearranging the fields inside a <dict>
element or changing the whitespace, this script should still cope OK.
And it’s fast. 3 seconds to write out all the files in my Library playlist.
OK so the point of all this is not to say that I am an XSLT god (although feel free to draw that conclusion), rather to show the expressiveness and power of the language. No, it’s not suitable for every coding task, but I think it’s a valuable tool to have in the arsenal. Particularly on MacOS X, where property lists show up everywhere. Including iTunes libraries.
2 Comments