[PDFBOX-3503] 2.0 much slower than 1.8 for text extraction with certain PDF files

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 2.0.2
• Environment: Windows 2012 R2, Oracle Java 1.8.0_91 64-bit
• Assignee: Unassigned
• Reporter: Andy McMullan
• Created: 19/Sep/16 10:20
• Updated: 19/Sep/16 18:54
• Resolved: 19/Sep/16 12:43

Description

While testing Tika 1.13 we've hit a performance problem with PDF conversion to text for certain files from the GovDocs corpus.

I have attached an example PDF that shows the problem. The problem can be reproduced using pdxbox-app.jar. Running the extraction with 1.8.12 takes around 1 second:

java -jar pdfbox-app-1.8.12.jar ExtractText 074031.pdf 074031.pdf.txt

Doing the same with 2.0.2 takes around 30 seconds:

java -jar pdfbox-app-2.0.2.jar ExtractText 074031.pdf 074031.pdf.txt

This is a small PDF so taking 30 seconds seems excessive.

Issue Links

• is duplicated by: PDFBOX-3442 OOM for single page pdf file