[PDFBOX-1653] Fix pdfbox eating up big chunks of memory for identical CID mappings

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 1.8.1, 1.8.2, 2.0.0
• Fix Version/s: 1.8.3, 2.0.0
• Component/s: FontBox
• Labels: PatchAvailable
• Assignee: Andreas Lehmkühler
• Reporter: Christian Kohlschütter
• Created: 27/Jun/13 22:49
• Updated: 30/Nov/13 17:02
• Resolved: 31/Aug/13 12:35

Description

pdfbox currently handles the PDF beginbfrange command (which creates a character mapping for a range of CIDs to Unicode characters) in a very inefficient way.

If a PDF document contains a range of CID 0 to CID 65535 with a mapping offset of 0 (which translates to "CID values map 1:1 to Unicode characters", pdfbox would nevertheless map each and every CID.

There apparently are PDFs with a lot of these 0-65535 mappings, and such a single PDF may cause an OutOfMemoryError.

This patch detects zero-offset ranges and basically just skips them from an explicit mapping.
There is some special handling for the space character included in the patch, which might or might not be relevant.

Issue Links

• is depended upon by: PDFBOX-1692 java.lang.OutOfMemoryError: Java heap space