tballison commented on PR #3146:
URL: https://github.com/apache/tika/pull/3146#issuecomment-5604983907
From my :robot:, let me know what you think.
```
Bug (must fix): RENDER_PAGES_BEFORE_PARSE with maxRenderedPages above the
page count renders nothing.
PDFParser.renderPDF now sends PageRangeRequest(1, N), and
PDFBoxRenderer.renderRange loops 1..N without clamping to getNumberOfPages().
Page nPages+1
throws IndexOutOfBoundsException, which is a RuntimeException so it passes
the per-page IOException catch, the renderer closes its results, and
renderPagesBeforeParse records it as a warning and returns with zero
renderings. Verified with a probe test on the 2-page fixture, maxRenderedPages:
5:
RENDER_PAGES_BEFORE_PARSE: renderings=0
tk:exception:warn=java.lang.IndexOutOfBoundsException: 1-based index out
of bounds: 3
RENDER_PAGES_AT_PAGE_END: renderings=2 (fine; per-page requests never
exceed the doc)
So the thumbnail case (maxRenderedPages: 1) works, but any N ≥ 2 silently
loses every rendering on documents shorter than N. No temp-file leak: the
renderer's catch (Throwable) closes the partial results.
Suggested fix: clamp to in PDFBoxRenderer.processRequest (Math.min(to,
pdDocument.getNumberOfPages())). That matches PopplerRenderer, since pdftoppm
-f 1
-l 5 on the same fixture writes exactly 2 PNGs. Keep from beyond the last
page throwing, because PDFBoxRendererTest.testFailedRenderLeavesNoTempFiles uses
PageRangeRequest(9999, 9999) as its vehicle. Add the failing test: both
strategies, limit above the page count, assert nPages renderings. The PR's tests
only cover maxRenderedPages: 1 on a 2-page doc, which can't reach this.
Hygiene
-
tika-parsers-standard-integration-tests/.../config-examples/pdf-parser-full.json
is the "every knob" example; add "maxRenderedPages": -1 next to
maxPages with a one-line comment.
- The PR description's "used to require maxPages: 1" is only true for
AT_PAGE_END; BEFORE_PARSE never honored maxPages at all. Wording only.
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
