PDFToolkit object by object

Inspecting fonts

From a TrueType program to PDF objects

Optional module · pdftoolkit/pdf-fonts 0.1.7

English | Français

Two formats, two responsibilities

ISO 32000-1 describes /FontFile2 as a stream containing a TrueType program, a descriptor as a dictionary of metrics in PDF units, and a composite font as an object graph: /Type0, /CIDFontType2, /CIDToGIDMap, /W and /ToUnicode. The core can construct each of these objects from known values. It does not decode the sfnt tables of an external font.

The TrueType format defines its tables and the rules required to read metrics and outlines. This package's classes perform that reading, then pass only normalised values to the core.

Example steps

  1. The application supplies bytes to the FontProgram constructor, which validates offsets, sizes, metrics and embedding permissions.
  2. CIDFontProgram preserves all glyphs and rebuilds the program intended for /FontFile2. It removes cmap, as required by section 9.9 for a /CIDFontType2, and the DSIG signature invalidated by the conversion. The sfnt directory and checksums are recalculated.
  3. The core receives these bytes in FontFile2. It calculates /Length1 without reading or validating TrueType tables; PDF filters encode the stream.
  4. UnicodeText declares the repertoire before the document is built. CharacterMapFactory finds GIDs in cmap, assigns distinct CIDs to Unicode characters and converts advances into 1,000-unit PDF metrics. Two characters sharing a glyph retain distinct CIDs and /ToUnicode mappings.
  5. FontDescriptorFactory converts dimensions, the bounding box, angle and heights into core NumericType values and a Rectangle. StemV and flags are supplied explicitly because the format does not provide a universal measurement of stem thickness. CapHeight can be supplied explicitly if the font does not declare it and the height of the H glyph is insufficient.
  6. CIDWidths, CIDToGIDMap, ToUnicodeCMap, CIDFontType2 and Type0Font then compose only PDF objects. They know nothing about FontProgram, cmap or units per em.
  7. TextEncoder converts previously declared characters into two-byte codes. A character absent from the repertoire raises an exception; the encoder does not modify the font without the application's knowledge.

Both factories verify that the supplied FontFile2 bytes correspond to the CIDFontProgram in use. Core objects then verify the identity of the embedded stream, the font name and the shared repertoire.

Limits

The example sets French and Greek text without contextual shaping. It does not claim to handle Arabic, scripts that require shaping, ligatures, kerning, variations or vertical writing automatically.

The demonstration Noto Sans font is redistributed under its OFL licence in examples/assets/fonts/OFL.txt. Files are read by the application; the package's classes have no implicit access to the file system.

Guide source at the documented commit ↗