Dominik Schmidt created TIKA-4857:
-------------------------------------
Summary: embedded-limits.maxDepth counts parser layers, not
embedding levels
Key: TIKA-4857
URL: https://issues.apache.org/jira/browse/TIKA-4857
Project: Tika
Issue Type: Bug
Reporter: Dominik Schmidt
The javadoc of EmbeddedLimits says maxDepth is the maximum nesting depth of
embedded documents that are still parsed (maxDepth=1: depth 1 parsed, depth 2
not). In practice maxDepth=2 parses depth 1 only; depth 2 needs maxDepth=3.
Reproduced with RecursiveParserWrapper on a DOCX whose EMF thumbnail is
rendered (the rendering is a depth-2 embedded document):
- maxDepth 1: depth 1 only, tk:exception:embedded-depth-limit-reached set
- maxDepth 2: depth 1 only, limit reached set
- maxDepth 3: depth 1 and 2
- maxDepth -1: everything
ParseRecord.depth is incremented in CompositeParser.parse for every composite
layer a parse passes through; with AutoDetectParser wrapping DefaultParser that
is two increments per document level, and ParsingEmbeddedDocumentExtractor
compares that counter with maxDepth + 1. The limit should be checked against
the embedding depth (tk:embedded-depth of the child), independent of how
parsers are composed. The same off-by-one applies to /unpack with
embedded-limits in tika-server.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
