I currently migrate to Tika 1.5, and fall into this behaviour, which leads to double entries in my database for one pdf file as I work directly with the handler.
Here are the two calls:
First call is in PDF2HTML, line 197: handler.endDocument();
this is part of the PDF2XHTML.process(pdfDocument, handler, context, metadata, localConfig);
invocation from PDFParser, line 143.
The second call is then directly in PDFParser, line 151: handler.endDocument();
Must stay at Tika 1.4 for now - still thanks for good work!