.PDF File Format – Fonts

Font types

Every font dictionary has Type Font. CIDFontType0 and CIDFontType2 are never used directly: a Type 0 font refers to them.

Simple fonts: Type 1 and TrueType

A simple font has at most 256 glyphs, one for each byte of a string.

Up to PDF 1.7, FirstChar, LastChar, Widths and FontDescriptor are required except for the standard 14 fonts. In PDF 2.0 they are required for every simple font except Type 3’s descriptor.

TrueType

A TrueType font, from PDF 1.1, has the same keys as Type 1, with the font program in FontFile2. How a byte code finds a glyph depends on the font’s cmap tables and the flags. For a non-symbolic font the code is mapped through the encoding to a glyph name, then to Unicode, and looked up in the (3, 1) Microsoft Unicode cmap, or through the name in the (1, 0) Macintosh cmap. For a symbolic font the code is looked up directly in the (3, 0) symbol cmap, often at 0xF000 plus the code, or in (1, 0). Readers differ in the details, which is why writers usually embed TrueType fonts as CIDFontType2 with Identity mapping instead.

Type 3 fonts

A Type 3 font’s glyphs are content streams. Its keys:

Type 3 fonts hold bitmap fonts from TeX and similar systems, and shapes that are not outline fonts.

Type 0 (composite) fonts

A Type 0 font, from PDF 1.2, has many glyphs and multi-byte codes, for Chinese, Japanese and Korean, and for any font embedded with more than 256 glyphs.

CIDFonts

A CIDFont is a font whose glyphs are numbered by CID, a character identifier. It is the one font below a Type 0 font.

The usual way to embed any OpenType or TrueType font today: a Type 0 font with Encoding Identity-H, a CIDFontType2 with CIDToGIDMap Identity, two-byte codes in strings that are the glyph indices, and a ToUnicode CMap so that the text can be copied.

The standard 14 fonts

Fourteen Type 1 fonts may be used by name without being embedded: Times-Roman, Times-Bold, Times-Italic, Times-BoldItalic, Helvetica, Helvetica-Bold, Helvetica-Oblique, Helvetica-BoldOblique, Courier, Courier-Bold, Courier-Oblique, Courier-BoldOblique, Symbol and ZapfDingbats. Readers draw them with fonts of the same metrics, such as Nimbus or Liberation. For any other font left out, a reader substitutes a font with the flags and widths of the descriptor. PDF/A and PDF/X require every font to be embedded, the standard 14 included.

Encodings

A simple font’s Encoding maps one-byte codes to glyph names, which the font program then finds. It is a name, or a dictionary:

• StandardEncoding – Adobe’s standard Latin encoding, the built-in encoding of most Type 1 fonts; used when nothing else is given for a non-symbolic font
• MacRomanEncoding – the classic Mac OS character set
• WinAnsiEncoding – Windows code page 1252
• MacExpertEncoding – expert glyphs: small capitals, fractions, old-style figures
• An encoding dictionary – Type Encoding, BaseEncoding (one of the names above) and Differences, an array such as [39 /quotesingle 96 /grave 128 /Adieresis /Aring]: a code, then the names of that code and the codes after it

A symbolic font with no Encoding, such as Symbol or ZapfDingbats, uses the encoding built into its font program. PDFDocEncoding is for text strings only, never for fonts.

CMaps

A CMap, from PDF 1.2, maps the bytes of a string to codes and the codes to CIDs, for a Type 0 font. It decides how many bytes each code has, from its code space ranges. Two CMaps work with any CIDFont: Identity-H and Identity-V, two-byte codes that equal the CID, horizontal and vertical. Many others are predefined for each Asian character collection, such as:

The UTF-16 CMaps came with PDF 1.5. Each has a -V form for vertical writing. An embedded CMap is a stream with Type CMap, CMapName, CIDSystemInfo, WMode (0 horizontal, 1 vertical) and UseCMap, whose data is a CMap program in PostScript syntax.

ToUnicode

A ToUnicode stream, from PDF 1.2, maps a font’s codes to Unicode, so that text can be searched, copied and read aloud. It is a CMap program:

/CIDInit /ProcSet findresource begin
12 dict begin
begincmap
/CIDSystemInfo << /Registry (Adobe) /Ordering (UCS) /Supplement 0 >> def
/CMapName /Adobe-Identity-UCS def
/CMapType 2 def
1 begincodespacerange
<0000> <FFFF>
endcodespacerange
2 beginbfchar
<0003> <0020>
<0024> <0041>
endbfchar
1 beginbfrange
<0044> <005D> <0061>
endbfrange
endcmap
CMapName currentdict /CMap defineresource pop
end
end

bfchar maps single codes; bfrange maps a run, either to consecutive Unicode values or to an array of strings. A destination may be several UTF-16BE units, for a ligature such as “fi” or a character outside the Basic Multilingual Plane. Without a ToUnicode, a reader works from the encoding’s glyph names (/Adieresis is Ä) or from a predefined CMap’s collection.

The font descriptor

At most one of FontFile, FontFile2 and FontFile3 may be present.

Embedded font programs

Each may also carry Metadata (1.4). An OpenType font in FontFile3 can serve a TrueType, Type1, CIDFontType0 or CIDFontType2 font dictionary, as long as its outlines suit it. A font program embedded in a PDF is a file of its own format; see Font file formats.

Subsets

A writer that embeds only the glyphs used gives the font a name of six capital letters, a plus sign and the real name: EOODIA+Poetica. The tag differs for each subset, so that two subsets of the same font in one file, or in files joined together, are not taken for each other. A subset’s glyph numbering may differ from the full font, so the subset cannot be replaced by the installed font.

Widths

Glyph widths come from the PDF, not the font program; a reader uses them to place each glyph, even when it draws the glyph from another font. For a simple font Widths gives one width a code. A CIDFont gives DW, the default (1,000), and W, an array of two forms mixed:

/W [ 120 [400 325 500]  7080 8032 1000 ]

For vertical writing, DW2 (default [880 −1000]) and W2 give the vertical origin and advance in the same two forms, with three numbers a CID.