Wikimedia\Parsoid\Language Namespace Reference

A bidirectional Language Converter, capable of round-tripping variant conversion. More...

Classes
class	ConstantLanguageGuesser
	A simple LanguageGuesser that returns the same "source language" for every node. More...
class	ConversionTraverser
class	CrhConverter
class	EnConverter
class	KuConverter
class	Language
	Base class for Language objects. More...
class	LanguageConverter
	Base class for language variant conversion. More...
class	LanguageCrh
	Crimean Tatar (Qırımtatarca) conversion code. More...
class	LanguageEn
	English ( / Pig Latin) conversion code. More...
class	LanguageGuesser
	An oracle that gives you a predicted "source language" for every node in a DOM, which is used when converting the result back to the source language during round-tripping. More...
class	LanguageKu
	Kurdish conversion code. More...
class	LanguageSr
	Serbian (Српски / Srpski) specific code. More...
class	LanguageZh
	Chinese conversion code. More...
class	MachineLanguageGuesser
	Use a {@Link ReplacementMachine} to predict the best "source language" for every node in a DOM. More...
class	SrConverter
class	ZhConverter

Detailed Description

A bidirectional Language Converter, capable of round-tripping variant conversion.

Language conversion is as DOMPostProcessor pass, run over the Parsoid-format HTML output, which may have embedded language converter rules. We first assign a (guessed) wikitext variant to each DOM node, the variant we expect the original wikitext was written in, which will be used when round-tripping the result back to the original wikitext variant. Then for each applicable text node in the DOM, we first "bracket" the text, splitting it into cleanly round-trippable segments and lossy/unclean segments. For the lossy segments we add additional metadata to the output to record the original text used in the wikitext to allow round-tripping (and variant-aware editing).

Note that different wikis have different policies for wikitext variant: in some wikis all articles are authored in one particular variant, by convention. In others, it's a "first author gets to choose the variant" situation. In both cases, a constant/per-article "wikitext variant" may be specified via some as-of-yet-unimplemented mechanism; either part of the site configuration, or per-article metadata like pageLanguage. In other wikis (like zhwiki) the text is a random mix of variants; in these cases the "wikitext variant" will be null/unspecified, and we'll dynamically pick the most likely wikitext variant for each subtree.

Each individual language has a dynamically-loaded subclass of Language, which may also have a LanguageConverter subclass to load appropriate ReplacementMachines and do other language-specific customizations.