Large document performance & low-memory PDF generation | Voters

Large document performance & low-memory PDF generation

Saurabh Choudhary

Improve vue3-pdf-export for large and complex documents, especially 50–100+ page PDFs with images, headers, footers, watermarks, security, and mixed content.
Planned work includes:
benchmark 10 / 25 / 50 / 100 page documents
measure generation time, PDF size, and browser memory usage
detect and prevent memory leaks during repeated generation
reduce peak memory usage for canvas rendering
explore page-by-page / low-memory rendering
improve image optimization and repeated asset caching
add cancellation for long-running PDF generation
expose current-page / total-page progress
warn developers when a document may exceed safe browser limits
improve browser responsiveness during long generation
stress-test complex HTML layouts
expand Playwright E2E coverage for large documents and cleanup
The goal is to make large PDF generation safer, faster, and more reliable across desktop and lower-memory devices without changing the existing HTML-to-PDF developer experience.

6 days ago

Saurabh Choudhary

Large-document PDF generation has been significantly improved in vue3-pdf-export, with a focus on performance, browser memory usage, responsiveness, cancellation, progress reporting, and reliability for complex documents.
The main improvement is incremental page-by-page Canvas rendering. Instead of relying on one very large full-document canvas, paginated documents can now be rendered one page at a time. In our benchmark fixture, the peak capture surface dropped from about 99.96 million pixels to about 1.74 million pixels per page, reducing peak capture-surface geometry by about 98.26%.
Large-document regression coverage now includes 10, 25, 40, 50, 60, 75, and 100-page documents. Canvas generation has been verified with 50-page and 100-page PDFs. 100 pages is the current tested range, not a hard package limit.
Static HTML footers were also optimized so they can be rendered once and reused across pages. In benchmark fixtures, footer generation improved from 7.82s to 1.43s without a logo and from 11.81s to 1.29s with a logo. Footer-heavy PDF size also dropped from about 46.7 MB to about 1.0 MB in the tested fixture.
Repeated images in selectable-text mode are now cached within a single generation, reducing unnecessary image loading and encoding work.
Temporary DOM nodes, canvases, encoded image data, and per-generation caches are cleaned up earlier to reduce retained resources after generation.
Long-running generation can now be cancelled using cancelGeneration(), with recovery support so another PDF can be generated after cancellation.
Progress reporting now includes current page, total pages, percentage, and processing stages such as pagination, images, rendering, watermark, header, footer, metadata, serialization, and completion.
The package also now supports selectable/searchable text rendering alongside the existing Canvas mode. Text mode can be especially useful for text-heavy documents and produces real selectable PDF text.
Additional stress and regression tests now cover large documents, image-heavy PDFs, repeated generation, cleanup, cancellation, security, repeated images, and footer optimization.
Overall, large-document PDF generation is now safer, more observable, more responsive, and substantially more robust while preserving the existing Vue developer experience and Canvas-based behavior.

Photo Viewer