Poppler (via pdftools) reports PDF parsing problems by printing "PDF error: ..." lines directly to R's message connection rather than raising a condition on them, so they bypass ordinary error handling and, uncaptured, flood the console - one line per malformed glyph or object, sometimes hundreds per certificate. This splits captured poppler output into messages that mean the PDF is genuinely broken (unparsable, not really a PDF) versus cosmetic rendering quirks that poppler recovers from on its own (e.g. malformed embedded fonts) and that don't affect the resulting page images.

Poppler (via pdftools) reports PDF parsing problems by printing "PDF error: ..." lines directly to R's message connection rather than raising a condition on them, so they bypass ordinary error handling and, uncaptured, flood the console - one line per malformed glyph or object, sometimes hundreds per certificate. This splits captured poppler output into messages that mean the PDF is genuinely broken (unparsable, not really a PDF) versus cosmetic rendering quirks that poppler recovers from on its own (e.g. malformed embedded fonts) and that don't affect the resulting page images.

Usage

classify_poppler_log(lines)

Arguments

lines

Character vector of captured message-stream output; only lines starting with "PDF error" are considered, everything else is ignored.

Value

A list with `fatal` (unique messages indicating the PDF could not really be parsed, character(0) if none) and `cosmetic_count` (number of suppressed non-fatal poppler messages).