Special syntax

Special syntax lets you control the conversion of individual words directly in the input text — without writing a custom config. It is honored by all three builtin pipelines (tarask, alphabetic, phonetic).

The three controls

	fix (convert normally)	no fix (leave as-is)	change only alphabet
brackets	<,Планета>	<Планета>	<*Планета>
no brackets	Планета	<.Планета>	<*.Планета>

By default a word is converted, so “fix” with brackets is only needed to override a word that would otherwise be left alone. The three markers are:

• <.X> / <*.X> — no fix. Leave X untouched (don’t convert it). With the leading * it still transliterates the alphabet but skips taraskevization.
• <,X> — force fix. Convert X even if it would normally be escaped.
• <*X> — alphabet only. Change only the alphabet of X, skipping taraskevization.

Bracketed vs. unbracketed forms

The bracketed form <…> is the most explicit and works anywhere, including mid-word. The unbracketed form uses a leading marker character and is handy for whole words:

import { pipelines, TaraskConfig, dicts } from "taraskevizer";

const cfg = new TaraskConfig({ abc: dicts.alphabets.latin });

pipelines.tarask("Гэта <Планета>", cfg);
// "Heta <Планета>"

pipelines.tarask("плянэта <.Планета>", cfg);
// "planeta Планета"

pipelines.tarask("<*Гродна>", cfg);
// <Hrodna>

Escaping angle brackets

Inside the markers, a literal > is written as \> so it is not treated as the marker’s closing bracket:

<тэкст з \> стрэлкай>

How it works under the hood

Marked parts are extracted before conversion and stored, then re-inserted afterwards using the noFixPlaceholder. The steps involved are resolveSpecialSyntax (which captures the parts) and applyNoFix (which puts them back). You normally never call these directly — they run as part of every builtin pipeline.