Large PDF Analysis

Books, catalogues, and long reports can contain hundreds of pages. Running the deepest possible inspection on every page at once can slow down the browser, so NowToPrint uses a staged analysis model.

Large-document triage is owned by the Rust PDF crates: structural truth comes from ntp-pdf-core, preflight policy and sampling evidence come from ntp-pdf-preflight, and the WASM engine only bridges that evidence into browser, desktop, and sidecar workflows.

How Analysis Runs

1. File received: The PDF is loaded into a safe worker environment.
2. Single worker pass: Large documents are opened once and the first triage data is collected in one worker command.
3. Fast manifest: Page count, PDF version, and core production signals from representative pages are read first.
4. Manifest-first result: For 121+ page or 45 MB+ files, the first visible result is built from structural manifest evidence instead of waiting for a full blocking preflight pass.
5. Sampled preflight: For large documents, representative pages and risk signals are checked first.
6. Scope is disclosed: The report shows whether the result is full or sampled.
7. Deep verification when needed: High-value or critical work can continue through a deeper verification path.

Report Scope

Scope	Meaning
Full analysis	The whole file was checked within the current analysis budget.
Sampled analysis	Representative pages and risk signals were checked first.
Deep analysis recommended	A broader verification pass is recommended for production-critical jobs.

How Sample Pages Are Chosen

For large PDFs, NowToPrint first extracts a lightweight Rust preflight manifest. The manifest records page count, PDF version, output-intent presence, and a deterministic sample plan. The first screen does not wait for page-box inspection across all 461 pages; the manifest reads page-box and bleed signals from representative pages. Pdfium then uses that plan for the first deep pass instead of guessing a new page list in the browser.

The sample plan does not only pick first, middle, and final pages. Pages with different dimensions, missing TrimBox, missing BleedBox, or insufficient bleed can be prioritized as representative risk pages.

This keeps the first result fast and repeatable: the same large PDF gets the same initial sample set, and the report can say when the evidence came from a sampled pass rather than a full-page exhaustive pass. Counts on the first screen are sampled evidence; production-critical work can request full verification.

What Gets Checked

• page count and page-size consistency,
• PDF version,
• font embedding risk,
• low-resolution image signals,
• colour profile and output intent signals,
• bleed and page box state,
• missing technical information for quoting.

When To Request Deeper Verification

• The file has 100+ pages and a high print run.
• The job is a book, catalogue, or high-value report.
• The report flags font, colour, or bleed risk.
• Approval requires production evidence.
• There is supplier, customer, or dispute risk.

Related Guides