greek-fontenc

T7

The [encguide] reserves the name T7 for a Greek standard font encoding. However, up to now, there is no agreement on an implementation because the restrictions for general text encodings are too severe for typesetting polytonic Greek.

LGR

The LGR font encoding is the de-facto standard for typesetting Greek with (8-bit) LaTeX. greek-fontenc provides a comprehensive LGR font encoding definition file.

Fonts in this encoding include the CB fonts (matching CM), grtimes (Greek Times), Kerkis (matching URW Bookman), DejaVu, Libertine GC, and the GFS fonts. Setup of these fonts as Greek variant to matching Latin fonts is facilitated by the substitutefont package.

The LGR font encoding allows to access Greek characters via an ASCII transliteration. This enables simple input with a Latin keyboard. Characters with diacritics can be selected by ligature definitions in the font (see [greek-usage], [teubner-doc], [cbfonts]).

A major drawback of the transliteration is, that you cannot access Latin letters if LGR is the active font encoding (e.g. in documents or parts of documents given the Babel language greek or polutionikogreek). This means that for every Latin-written word or acronym an explicit language-switch is required. This problem can only be solved via a font-encoding comprising Latin and Greek like the envisaged T7 or Unicode (with XeTeX or LuaTeX).

LGI

The ‘Ibycus’ fonts from the package ibygrk implement an alternative transliteration scheme (also explained in [babel-patch]). It is currently not supported by greek-fontenc.

The font encoding file lgienc.def from ibycus-babel provides a basic setup (without any LICR macros or composite definitions).

TU

Standard Unicode font encoding for XeTeX and LuaTeX loaded by fontspec (since v2.5a) rsp. the LaTeX kernel since 2017/01/01 [LaTeX2e News Issue 26]_. greek-fontenc adds support for the Greek script (see tuenc-greek).

Xe/LuaTeX works with any system-wide installed OpenType font. Suitable fonts supporting Greek include CM Unicode, Deja Vu, EB Garamond, the GFS fonts, Libertine OTF, Libertinus, Old Standard, Tempora, and UM Typewriter (all available on CTAN) but also many commercial fonts. Unfortunately, the fontspec default, Latin Modern misses most Greek characters.

XeTeX uses the Unicode NFC normalization, so that combining characters are merged with the base character if a pre-composed character exists. This results in better looking output for characters with multiple diacritics. Unfortunately, LuaTeX does not apply the NFC normalization. This leads to suboptimal placing of some diactritics, especially the sub-iota (becoming unintelligable in combination with small letter eta).

TODO: The lua-uni-algos package may be helpfull to implement a NTC normalization to Greek text in LuaTeX.

The legacy Unicode font encodings EU1 and EU2 for XeTeX and LuaTeX respectively were superseded by TU in the 2017 fontspec release.

PU

The package hyperref defines the PU font encoding for use in PDF strings (ToC, bookmarks) which supports monotonic Greek. greek-fontenc adds support for polytonic Greek and some archaic characters also supported in LGR and TU (see hyperref-with-greek.tex, hyperref-with-greek.pdf).

letters and symbols

• The fntguide (section 6.4 Naming conventions) recommends:

  Where possible, text symbols should be named as \text followed by the Adobe glyph name: for example \textonequarter or \textsterling. Similarly, math symbols should be named as \math followed by the glyph name, for example \mathonequarter or \mathsterling.

  Problem:

  The Adobe Glyph List For New Fonts has names for many glyphs in the Greek and Coptic Unicode block, but not for Greek extended. The Adobe Glyph List (for existing fonts) lists additional glyph names used in older fonts. However, these are not intended for active use.

• If there exists a math-mode macro for a symbol, the corresponding text macro could be formed by prepending text.

  Example:

  The glyph name for the GREEK SMALL LETTER FINAL SIGMA is sigma1, the corresponding math-macro is \varsigma. The text symbol is made available as \textvarsigma.

  Problem:

  The math macros for the symbol variants \varepsilon and \varphi map to characters named “GREEK SMALL LETTER …”, while \vartheta, \varkappa, \varrho, and \varpi map to “GREEK … SYMBOL” Unicode characters. (See also section 5.5.3 of the unicode-math documentation.)

• The Unicode names list provides standardized descriptive names for all Unicode characters that use only capital letters of the Latin alphabet. While not suited for direct use in LICR macros, they can be either

  1. used as inspiration for new LICR macro names or

  2. converted to LICR macro names via a defined set of transformation rules.

  Example:

  \textfinalsigma is a descriptive alias for GREEK SMALL LETTER FINAL SIGMA derived via the rules:

  • drop “LETTER” if the name remains unique,

  • drop “GREEK” if the name remains unique,

  • use capitalized name for capital letters, lowercase for “SMALL” letters and drop “SMALL”,

  • concatenate

• Omit the “text” prefix for macros that do not have a math counterpart?

  Pro:

  • Simpler,

  • ease of use (less typing, better readability of source text),

  • many established text macro names without “text”,

  • text prefix does not mark a macro as encoding-specific or “inserting a glyph”. There are e.g. font-changing macros (\textbf, \textit) and encoding-changing macros (\textgreek, \textcyr).

  • There are examples of encoding-specific macros without the text-prefix, especially for letters, see encguide.

  Contra:

  • Less consistent,

  • possible name clashes

  • text prefix marks a macro as confined to text (as opposed to math) mode,

  The font encoding definition files use the text prefix for symbols. Aliases (short forms, compatibility defs, etc.) are defined in additional packages (e.g. alphabeta.sty, babel-greek, or teubner)

accent macros

• standard accent macros (\DeclareTextAccent definitions in latex/base/...) are one-symbol macros (\' \" ... \u \v ...) .

• tipa.sty, xunicode, and ucs use the “text” prefix also for accents.

  However, the Adobe Glyph List For New Fonts maps, e.g., “tonos” and “dieresistonos” to the spacing characters GREEK TONOS and GREEK DIALYTIKA TONOS, hence texttonos and textdiaresistonos should be spacing characters.

• textcomp (ts1enc.def) defines \capital... accents (i.e. without text prefix).

Currently, greek-fontenc uses for diacritics:

• Greek names like in Unicode, and ucsencs.def, and

• the prefix \acc to distinguish the macros as TextAaccent and reduce the risc of name clashes (cf. \@tabacckludge).

For the end-user “symbol macros” (\~ \' \` \" \< \> \"' ...) are provided. (The non-standard macros \< and \> only with textalpha or alphabeta.)

symbol variants

See also http://en.wikipedia.org/wiki/Greek_alphabet#Glyph_variants

Mathematical notation distinguishes variant shapes for beta (β|ϐ), theta (θ|ϑ), phi (φ|ϕ), pi (π|ϖ), kappa (κ|ϰ), rho (ρ|ϱ), Theta (Θ|ϴ), and epsilon (ε|ϵ). The variations have no syntactic meaning in Greek text and Greek text fonts use the shape variants indiscriminately.

Unicode defines separate code points for the symbol variants for use in mathematical context. However, they are sometimes also used in place of the corresponding letter characters in Unicode-encoded text.

The variant shapes are not given separate code-points in the LGR font encoding.

In mathematical mode, TeX supports the distinction between θ|ϑ, π|ϖ, φ|ϕ, ρ|ϱ, and ε|ϵ with \var<lettername> macros. However, the mapping of letter/symbol in Unicode to “normal”/variant in TeX is inconsistent and variant macros for ϴ ϐ, and ϰ are not available without additional packages (e.g. amssymb provides ϰ as \varkappa).

greek-fontenc provides \text<lettername>symbol LICR macros for these characters:

• With Unicode fonts, the macros select the GREEK <lettername> SYMBOL``.

• With LGR encoded fonts, they report an error by default and are mapped to the corresponding letter with the normalize-symbols option of textalpha and alphabeta (loosing the distinction between the shape variants).