Skip to contents

Per-character text extraction

Source: R/page_extras.R
pdf_text_chars.Rd

Returns one tibble row per character on the page, with the character's Unicode codepoint and UTF-8 form, glyph bounding box, effective font size, and two PDF flags indicating "generated" characters (whitespace PDFium inferred between positioned glyphs) and end-of-line "soft" hyphens. Wraps FPDFText_LoadPage plus FPDFText_CountChars / _GetUnicode / _GetCharBox / _GetFontSize / _IsGenerated / _IsHyphen.

Usage

pdf_text_chars(page, page_num = 1L)

Arguments

page

A pdfium_page from pdf_page_load(), or a pdfium_doc.

page_num

One-based page index. Only used when page is a pdfium_doc. Ignored otherwise.

Value

A tibble with columns:

  • char_index integer - 1-based position in the page's character stream.

  • codepoint integer - Unicode code point.

  • char character - UTF-8 character; empty for surrogate halves or PDFium's NUL sentinel.

  • bounds_left, bounds_bottom, bounds_right, bounds_top - glyph bounding box in PDF user space.

  • font_size numeric - effective glyph height in user-space points (the run's font size times the text matrix scale).

  • is_generated logical - TRUE for whitespace PDFium synthesised between positioned glyphs (the source PDF does not carry a character there; PDFium infers one for text-extraction consumers).

  • is_hyphen logical - TRUE for end-of-line soft hyphens.

  • origin_x, origin_y - the character's glyph origin point in PDF user space (FPDFText_GetCharOrigin). Distinct from the bounding-box corners; for many fonts the origin is at the baseline left of the glyph.

  • loose_left, loose_bottom, loose_right, loose_top - the "loose" bounding box covering the entire glyph cell (font ascent / descent included), not just the glyph outline. Use these when you need consistent line heights; use bounds_* for the tight glyph extent.

  • unicode_map_error logical - TRUE when PDFium detected that the character's ToUnicode CMap is malformed for this glyph (the codepoint reported may be the PDF's ` fallback rather than the intended character).

  • text_index integer - 0-based position in the extractable text string (i.e. the linear pdf_doc_text() output) for this character, or NA for synthesised whitespace and other characters that don't appear in the extracted text.

  • char_font_name character - the font name PDFium reports for this specific character (via FPDFText_GetFontInfo). Per-character because pages can mix fonts within a single text run after PDFium re-flows characters during extraction.

  • char_font_flags integer - the PDF Font Descriptor /Flags bitmask for this character's font (PDF spec Table 121). Useful for detecting /Symbolic (bit 3) or /AllCap (bit 17) fonts whose ToUnicode mapping may be unreliable.

Returns a 0-row tibble of the same schema when the page has no text.

Details

This is the per-character analog of pdf_text_runs() (per-text-object) and pdf_doc_text() (per-page). The three coexist: use pdf_doc_text() when you just want the strings, pdf_text_runs() for object-level positions, and pdf_text_chars() when you need glyph-level geometry (e.g. word segmentation, character-by-character layout analysis).

See also

pdf_text_runs(), pdf_doc_text().