[TIKA-2511] Slowness parsing SQLite database file

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.16
• Fix Version/s: 1.17
• Assignee: Unassigned
• Reporter: Eamonn Saunders
• Created: 27/Nov/17 17:10
• Updated: 27/Nov/17 20:58
• Resolved: 27/Nov/17 19:55

Description

Parsing of the attached urlclassifier3.sqlite database is approximately 5 times slower in Tika 1.16 than it was in 1.14.

I've performed some profiling and it appears as though the problem lies in the number of times a new TikaConfig instance gets created. See attached screenshot

I notice that EmbeddedDocumentUtil has 2 methods to get a TikaConfig, one of which caches the config instance. Are both methods needed? Why does EmbeddedDocumentUtil.getExtension() use the version that doesn't cache the config?