mirror of
https://github.com/arsvendg/Stirling-PDF.git
synced 2026-07-14 10:34:06 +02:00
Anthony StirlingandGitHub
56ff1e5092
## Summary Migrates `SplitPDFController`, `SplitPdfByChaptersController`, `SplitPdfBySizeController` from PDFBox to JPDFium. `SplitPdfBySectionsController` and `AutoSplitPdfController` are intentionally left on PDFBox (require JPDFium 1.0.2 features that don't exist yet). ## Benchmark (audited on `audit/jpdfium-split`, file `app/core/src/test/java/stirling/software/SPDF/bench/SplitBenchmark.java`) | Workload | PDFBox heap | JPDFium heap | PDFBox wall | JPDFium wall | |---|---|---|---|---| | 100 pp, chunk 10 | +21-26 MB | **+0.02 MB** | 80-106 ms | **25 ms** | | 300 pp, chunk 10 | +59 MB | **+1.0 MB** | 232 ms | **76 ms** | **98-99.9% heap reduction. 3-4.2x faster wall.** ## Hybrid - AcroForm-bearing splits keep PDFBox `FormUtils.pruneOrphanedFormFields` post-pass (FPDF_ImportPagesByIndex drops AcroForm dict). Sub-bench shows +1.0 MB / +27 ms - tightly bounded. - Metadata extraction stays on PDFBox. - `SplitPdfBySectionsController` - JPDFium `PdfPageSplitter` only does 2-up halving, not arbitrary MxN. - `AutoSplitPdfController` - needs PDFRenderer + zxing for QR markers. ## Test plan - [ ] 30/30 unit tests pass with PDFBox `Loader.loadPDF` as the oracle (assert page counts + document totals) - [ ] Existing cucumber feature `split.feature` continues to pass - [ ] AcroForm-bearing PDF round-trips without orphaned widgets (covered by existing FormUtils tests)