source/data/translit/t_Latn_NPinyn.txt

 // -*- Coding: utf-8; -*-
//--------------------------------------------------------------------
// Copyright (c) 1999-2004, International Business Machines
// Corporation and others.  All Rights Reserved.
//--------------------------------------------------------------------
// THIS IS A MACHINE-GENERATED FILE
// Tool: \icu4j\src\com\ibm\icu\dev\tool\translit\dumpICURules.bat
// Source: \icu4j\src\com\ibm\icu\impl\data/Transliterator_Latin_NumericPinyin.txt
// Date: Fri May 28 17:07:31 2004
//--------------------------------------------------------------------

// Latin_NumericPinyin

t_Latn_NPinyn {
  Rule {
//--------------------------------------------------------------------
//--------------------------------------------------------------------

// According to the pinyin definitions I've been able to find:
// 'a', 'e' are the preferred bases
// otherwise 'o'
// otherwise last vowel

// The trailing form of syllables are the following:
//         "a", "ai", "ao", "an", "ang",
//         "o", "ou", "ong",
//         "e", "ei", "er", "en", "eng",
//         "i", "ia", "iao", "ie", "iu", "ian", "in", "iang", "ing", "iong",
//         "u", "ua", "uo", "uai", "ui", "uan", "un", "uang", "ueng",
//         "ü", "üe", "üan", "ün"
// so the letters the tone will 'hop' are:

"::NFD (NFC);"
"$tone = [\u0304\u0301\u030C\u0300\u0306] ;"

// Move the tone to the end of a syllable, and convert to number
"e {($tone) r} > r &tone-digit($1);"
"($tone) ( [i o n u {o n} {n g}]) > $2 &tone-digit($1);"
"($tone) > &tone-digit($1);"

// The following backs up until it finds the right vowel, then deposits the tone

"$vowel = [aAeEiIoOuUüÜ];"
"$consonant = [[a-z A-Z] - [$vowel]];"
"$digit = [1-5];"
"$1 &digit-tone($3) $2 < ([aAeE]) ($vowel* $consonant*) ($digit);"
"$1 &digit-tone($3) $2 < ([oO]) ([$vowel-[aeAE]]* $consonant*) ($digit);"
"$1 &digit-tone($3) $2 < ($vowel) ($consonant*) ($digit);"
"&digit-tone($1) < [:letter:] {($digit)};"

"::NFC (NFD);"



  }
}