Tamil

texish sets the Tamil script with the syllable shaping it needs. Like Devanagari, Bengali, Gurmukhi and Telugu it is written left to right, so it needs no bidirectional reordering; what it needs is work inside each syllable. The engine segments a run into orthographic syllables and shapes each one — splitting the two-part vowel signs, moving the pre-base signs to where they are drawn, and positioning the marks — all automatically from text typed in the usual Unicode order.

The Tamil font

A Tamil face, Noto Serif Tamil, is bundled in a regular and a bold cut. Select it with \font tamil:

{\font tamil 14 regular
தமிழ் ஒரு அழகான மொழி.
}

{\font tamil 18 bold வணக்கம்!}

(As with the bundled Hebrew, Arabic, CJK and other Indic faces, the in-browser Scala.js build does not ship the Tamil font, to keep the download small.)

What the engine shapes

Everything below happens automatically for any run set in a Tamil font — you type the characters in reading order and the engine does the rest.

A silenced consonant keeps its pulli. Tamil does not fold a virama-joined consonant into a half-form as Devanagari does, nor subjoin it beneath the base as Telugu does. The virama stays on the page as a dot above the letter — the pulli — and the consonant carrying it is a complete letter in its own right: க், ம், ன்.

A pre-base vowel sign moves before its own consonant, and no further. Three signs are drawn to the left of the consonant they belong to (ெ, ே, ை), and because a silenced consonant ahead of the base is a letter rather than half of a conjunct, the sign stops at the pulli instead of leaping over it. க்கே sets ka, pulli, the ee sign, ka — the vowel is drawn between the two consonants. This is the one place Tamil parts company with Devanagari and Bengali, whose pre-base signs go to the very front of the syllable.

Two-part signs are split. Three vowel signs are one character in memory and two signs on the page: ொ and ோ as an e or ee sign before the base and an aa sign after it, ௌ as an e sign before and a length mark after. The engine splits each so both parts land where the font expects them.

The u signs fuse, and the i signs take width-matched forms. கு and கூ are drawn as single glyphs rather than a letter and a mark, and the short-i and long-i signs come in variants sized to the consonant they follow — தி and மி use different forms of the same sign. Both are the font’s own post-base substitutions, and neither needs anything from the document.

Only two conjuncts survive. Tamil borrowed a handful of Sanskrit conjuncts and kept two in ordinary use: க்ஷ and the sacred ஸ்ரீ. The font draws each as a single ligature, and because the ligature swallows the pulli there is nothing left to stop a pre-base sign, so in க்ஷே the vowel does reach the front.

There is no reph. Unlike Devanagari and Bengali, a word-initial ra with a virama does not rise as a mark above the syllable. In ர்க the ra keeps its pulli and stays an ordinary letter.

The short-i sign does not reorder. Bengali’s ি is drawn before its consonant; Tamil’s ி is drawn above and to the right of it and stays where it is typed. So does ீ.

Setting a paragraph

Tamil sets in ordinary running text: the line breaker finds its breaks at the interword spaces exactly as for Latin, and only the shaping within each word is special. A whole Tamil paragraph needs nothing beyond selecting the font.

{\font tamil 12 regular
இது தமிழ் எழுத்தில் எழுதப்பட்டுள்ளது. ஒவ்வொரு எழுத்தும் ஒரு மெய்யெழுத்தும் அதனுடன்
சேர்ந்த உயிரெழுத்தும் இணைந்து உருவாகிறது.
}

Because Tamil sets left to right like a Latin text face, it uses the same TeX-style punctuation shorthands: `` and '' become curly quotes, --/--- en/em dashes, and ... an ellipsis. (The right-to-left Hebrew and Arabic faces do not — their quote conventions differ, so those documents type the marks literally.)

The bundled scripts/tamil-demo.script sets a full page covering each of these features.