NullPointerException in ListUtils.isContainsHeading during hybrid mode postProcess

[odinsseo]

Bug

When using hybrid mode (--hybrid docling-fast, --hybrid-mode auto), the Java CLI crashes with a NullPointerException during the post-processing stage. The docling-fast backend processes pages successfully (returns 200 OK), but the Java process fails in HybridDocumentProcessor.postProcess() and produces no JSON output.

The crash occurs in the verapdf list processing code: same area as #134 but a different method (ListUtils.isContainsHeading vs ListLabelsUtils.haveDifferentSuffixChars). The defensive try-catch added in #134 does not cover this code path.

Stack trace:

SEVERE: Exception during processing file /tmp/tmp0i17_px2/chunk.pdf: null
java.lang.NullPointerException
at org.verapdf.wcag.algorithms.semanticalgorithms.utils.ListUtils.isContainsHeading(ListUtils.java:211)
at org.verapdf.wcag.algorithms.semanticalgorithms.utils.ListUtils.checkChildrenListInterval(ListUtils.java:157)
at org.verapdf.wcag.algorithms.semanticalgorithms.utils.ListUtils.getChildrenListIntervals(ListUtils.java:130)
at org.opendataloader.pdf.processors.ListProcessor.processListsFromTextNodes(ListProcessor.java:350)
at org.opendataloader.pdf.processors.HybridDocumentProcessor.postProcess(HybridDocumentProcessor.java:422)
at org.opendataloader.pdf.processors.HybridDocumentProcessor.processDocument(HybridDocumentProcessor.java:166)
at org.opendataloader.pdf.processors.HybridDocumentProcessor.processDocument(HybridDocumentProcessor.java:78)
at org.opendataloader.pdf.processors.DocumentProcessor.processFile(DocumentProcessor.java:73)
at org.opendataloader.pdf.api.OpenDataLoaderPDF.processFile(OpenDataLoaderPDF.java:32)
at org.opendataloader.pdf.cli.CLIMain.processFile(CLIMain.java:113)
at org.opendataloader.pdf.cli.CLIMain.processPath(CLIMain.java:92)
at org.opendataloader.pdf.cli.CLIMain.main(CLIMain.java:64)

The Java process logs SEVERE but exits with code 0, so the calling Python code receives no error. It only discovers the problem when no JSON output file exists. Reproducible on every document tested (50-page technical manuals with tables and lists). The triage routes ~24 pages to Java and ~26 to docling-fast.

...

Steps to reproduce

1. Start the hybrid backend:
   opendataloader-pdf-hybrid --port 5002

2. Convert a PDF with hybrid mode:
   opendataloader-pdf --hybrid docling-fast --hybrid-url http://localhost:5002 --hybrid-timeout 60000 --hybrid-fallback --table-method cluster -f json -o output/ input.pdf

   Or via Python:
   import opendataloader_pdf
   opendataloader_pdf.convert(
   input_path="document.pdf",
   output_dir="output/",
   format="json",
   hybrid="docling-fast",
   hybrid_url="http://localhost:5002",
   hybrid_timeout="60000",
   hybrid_fallback=True,
   table_method="cluster",
   )

3. The docling-fast backend processes successfully (200 OK in logs).

4. Java crashes in postProcess with NullPointerException: no JSON output is produced.

...

Version

1.10.1 (pip install opendataloader-pdf[hybrid])

...

Java version

OpenJDK 11 (default-jre-headless, from pytorch/pytorch:2.5.1-cuda12.4-cudnn9-runtime base image, Ubuntu 22.04)

...

[hnc-leebd]

Thanks for the detailed report and the clear stack trace — really helpful for pinpointing the issue.

Confirmed: this is a related but distinct code path from #134. The NullPointerException in ListUtils.isContainsHeading isn't covered by the defensive try-catch added there, so we need to extend the same protection to this call chain.

We've logged this internally and are actively working on a fix. Will update here once it's ready.

[sipa-echo-rcco]

I am running into this exact same issue on a 200-page technical manual.
The triage routed 45 pages to Java and 155 to docling-fast (running locally via Docker).

I wanted to share my stack trace because my Java environment provided a more detailed NullPointerException message.
It explicitly names INode.getCorrectSemanticScore() as the method returning the null value that causes the crash on line 211.

SEVERE: Exception during processing file .../placeholder.pdf: Cannot invoke "java.lang.Double.doubleValue()" because the return value of "org.verapdf.wcag.algorithms.entities.INode.getCorrectSemanticScore()" is null
java.lang.NullPointerException: Cannot invoke "java.lang.Double.doubleValue()" because the return value of "org.verapdf.wcag.algorithms.entities.INode.getCorrectSemanticScore()" is null
	at org.verapdf.wcag.algorithms.semanticalgorithms.utils.ListUtils.isContainsHeading(ListUtils.java:211)
	at org.verapdf.wcag.algorithms.semanticalgorithms.utils.ListUtils.checkChildrenListInterval(ListUtils.java:157)
	at org.verapdf.wcag.algorithms.semanticalgorithms.utils.ListUtils.getChildrenListIntervals(ListUtils.java:130)
	at org.opendataloader.pdf.processors.ListProcessor.processListsFromTextNodes(ListProcessor.java:350)
	at org.opendataloader.pdf.processors.HybridDocumentProcessor.postProcess(HybridDocumentProcessor.java:422)
	at org.opendataloader.pdf.processors.HybridDocumentProcessor.processDocument(HybridDocumentProcessor.java:166)