Serializing parsed HTML trees

[jberkel]

Hello,

I'm using jsoup as part of a data processing pipeline I've built for MediaWiki HTML dumps, using Spark. This involves parsing gigabytes of HTML and works very well. But it occurred to me a that a lot of CPU time is spent reparsing the same HTML over and over, so I was wondering if there's anything to "pre-parse" the HTML (once) into a serialized format which can be read more efficiently. Java's serialization or something like kryo might be an option, or maybe it's worth writing a custom serializer. Any recommendations?