.PDF File Format – Fonts

Kinds of font

Every font dictionary has /Type /Font and a Subtype. The first four are simple fonts: a string’s bytes are codes 0 to 255, each one glyph.

Type 1 fonts

The widths in the PDF are the ones that count for placing glyphs, even if the embedded font program says otherwise. A code outside FirstChar to LastChar takes the descriptor’s MissingWidth, default 0.

7 0 obj
<< /Type /Font /Subtype /Type1 /BaseFont /ABCDEF+Minion-Regular
   /FirstChar 32 /LastChar 126 /Widths 8 0 R
   /Encoding /WinAnsiEncoding /FontDescriptor 9 0 R >>
endobj

TrueType fonts

From PDF 1.1. The dictionary has the same keys as a Type 1 font, with /Subtype /TrueType. BaseFont is the PostScript name from the font’s name table, or the name with spaces removed, and for a style made by the system a suffix such as ,Bold or ,Italic.

To find a glyph for a code, a reader uses the font’s cmap table. For a non-symbolic font with an encoding, the code is turned into a glyph name, the name into Unicode, and the (3, 1) Microsoft Unicode subtable consulted, or the name looked up in the (1, 0) Macintosh table. For a symbolic font the code is looked up directly in the (3, 0) subtable, often at 0xF000 plus the code, or in the (1, 0) table. The rules leave room for error, and readers differ on badly made fonts.

Type 3 fonts

A Type 3 font has no font program. Each glyph is a small content stream in the PDF itself, so it can draw anything, bitmaps included: TeX’s bitmap fonts often reach PDF this way.

Each glyph stream begins with d0 or d1; see Content streams.

The standard 14 fonts

Every reader is expected to have these, so in PDF 1.0 to 1.4 they may be used without embedding, and without Widths or a descriptor:

Readers draw them with their own copies or close substitutes, using the widths from Adobe’s font metrics files. Any other font that is not embedded is replaced by a reader’s guess from the descriptor, which is why the descriptor’s flags and metrics matter.

Encodings

A simple font’s encoding maps each code, 0 to 255, to a glyph name such as A, eacute or quotedblleft. The font program then finds the glyph by name. Encoding is a name or a dictionary:

Differences is a list in which a number sets the next code and each name after it takes one code in turn:

<< /Type /Encoding /BaseEncoding /WinAnsiEncoding
   /Differences [ 39 /quotesingle 96 /grave 128 /Euro /bullet ] >>

A string such as (Hello) shown in this font is the codes 72, 101, 108, 108, 111. Text in PDFs is only as readable as its encoding: copying text works when the glyph names are standard or a ToUnicode CMap is present.

Type 0 fonts

From PDF 1.2, for Chinese, Japanese and Korean text and any font with more than 256 glyphs. A Type 0 font joins a CMap, which turns the bytes of a string into character identifiers (CIDs), and a CIDFont, which holds the glyphs by CID.

CIDFonts

W mixes two forms: a CID and an array of widths for it and the CIDs that follow, or a first CID, a last CID and one width for them all:

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

CMaps

A CMap maps byte sequences to CIDs. It also defines the code space: which byte sequences are one, two, three or four bytes long, so a reader can split a string into codes. It is either one of Adobe’s predefined CMaps, named in Encoding, or a stream in the file holding a CMap program, with /Type /CMap, CMapName, CIDSystemInfo, WMode (0 horizontal, 1 vertical) and UseCMap.

Each predefined CMap ends in -H for horizontal or -V for vertical writing. The list grew with each version; a reader needs Adobe’s CMap files, published as open source, to use them.

ToUnicode

From PDF 1.2 a font can carry a ToUnicode stream: a CMap program that maps its codes to Unicode, so that text can be copied, searched and read aloud. It uses the bfchar and bfrange sections of a CMap, with UTF-16BE destinations:

/CIDInit /ProcSet findresource begin
12 dict begin
begincmap
/CMapName /Adobe-Identity-UCS def
/CMapType 2 def
1 begincodespacerange
<0000> <FFFF>
endcodespacerange
2 beginbfchar
<0003> <0020>
<0024> <0041>
endbfchar
1 beginbfrange
<0044> <0046> <0061>
endbfrange
endcmap
CMapName currentdict /CMap defineresource pop
end
end

A destination can be several code units, for ligatures: <0102> <00660069> maps one glyph to “fi”. Without ToUnicode, a reader falls back on glyph names and the font’s own tables, and may get text wrong.

Font descriptors

A font descriptor gives the metrics and style of a font, for substituting it when it is not embedded, and points at the embedded program when it is:

A CIDFont’s descriptor can also have Style (a Panose classification), FD (descriptors for parts of the font) and CIDSet (a stream with a bit for each CID present in a subset).

Embedded font programs

An embedded font is a stream named by one of the descriptor’s three keys:

A CIDFontType2 font embeds its TrueType program with FontFile2, and a CIDFontType0 font its CFF with FontFile3. Font streams may carry Metadata from PDF 1.4.