pub fn parse(data: &[u8], fmt: Option<Format>) -> Result<Document>Expand description
Parse data into the unified ir::Document.
fmt forces a format; None auto-detects from magic bytes. The core makes no
I/O assumptions — it only borrows &[u8], so each binding feeds it bytes
however it likes (Python bytes, Node Buffer, WASM Uint8Array).
M0 uses lopdf’s naive text extraction (one paragraph per page, no per-char
coordinates). PER-36 replaces the PDF path with the self-written content-stream
interpreter that fills ir::Page::chars with precise bboxes.