Only once, and then it’ll work for 10000 documents. Hopefully. And “the source HTML” is all you have, anyway. If you think about converting the PDF to HTML – don’t. That’ll probably not yield a better structured HTML than the original one.
In any case, a sample PDF and original HTML would help to figure out if/how to get at the data. As it stands, it’s all a bit foggy.
But there is no HTML/XML source available for this record. You’re kind of reading it backwards. source is meant to access the HTML/XML of an HTML/XML record. Not to add stuff to arbitrary records. And in the case of an HTML record, one can actually modify source, creating all kinds of havoc.
Since it’s possible to convert a PDF to HTML (at least in the one case I tried it with), the documentation seems in fact to imply that this conversion might happen on the spot, creating an HTML document in the source property. Which doesn’t happen, though. @cgrunenberg would have to comment on that