selima-perl/lib/php/monica/pinyin.inc.php

<?php
// File name:	pinyin.inc.php
// Description:	PHP subroutines to convert Traditional Chinese to Pinyin
// Date:	2004-04-14
// Author:	imacat <imacat@pristine.com.tw>
// Copyright:	Copyright (C) 2004-2007 Pristine Communications

// Set the include path
if (!defined("INCPATH_SET")) {
    require_once dirname(__FILE__) . "/incpath.inc.php";
}
// Referenced subroutines
require_once "monica/chkwrite.inc.php";
require_once "monica/mkalldir.inc.php";
require_once "monica/xfileio.inc.php";

// Settings
if (!defined("DBMDIR")) {
    define("DBMDIR", dirname(dirname(dirname(__FILE__))) . "/" . php_uname("m"));
}
define("_PINYIN_PINYIN2BIG5", dirname(__FILE__) . "/pinyin2big5");
define("_PINYIN_B52PY_DB", DBMDIR . "/b52py.db");

$_PINYIN_DB = null;
define("_PINYIN_DBTYPE_DBA", 1);
define("_PINYIN_DBTYPE_ARRAY", 2);
$_PINYIN_SUP = array(
    "　"   => " ",
    "＃"   => "#",
    "﹟"   => "#",
    "兙"   => "shi2 ke4",
    "兛"   => "qian1 ke4",
    "兞"   => "mao2 ke4",
    "兝\"   => "fen1 ke4",
    "兡"   => "bai3 ke4",
    "兣"   => "li2 ke4",
    "嗧"   => "jia1 lun2",
    "瓩"   => "qian1 wa3",
    "糎"   => "li2 mi3",
);

// b52py: Convert Traditional Chinese to Pinyin
function b52py($big5)
{
    global $_PINYIN_DB;
    // Initialize the database
    _init_b52py();
    
    // Split into pieces
    $pieces = array();
    $remains = $big5;
    while (preg_match("/^([\\x01-\\x7F]+|(?:[\\x80-\\xFE].)+)(.*)$/s", $remains, $m)) {
        $pieces[] = $m[1];
        $remains = $m[2];
    }
    
    // Process each piece
    switch (_PINYIN_DBTYPE) {
    // Using a DB file
    case _PINYIN_DBTYPE_DBA:
        for ($i = 0; $i < count($pieces); $i++) {
            // Big5 piece
            if (preg_match("/^[\\x80-\\xFF]/", $pieces[$i])) {
                for ($j = 0, $words = array(); $j < strlen($pieces[$i]); $j += 2) {
                    $char = substr($pieces[$i], $j, 2);
                    if (dba_exists($char, $_PINYIN_DB)) {
                        $words[] = dba_fetch($char, $_PINYIN_DB);
                    } else {
                        $words[] = $char;
                    }
                }
                $pieces[$i] = implode(" ", $words);
            }
        }
        break;
    
    // Using an array
    case _PINYIN_DBTYPE_ARRAY:
        for ($i = 0; $i < count($pieces); $i++) {
            // Big5 piece
            if (preg_match("/^[\\x80-\\xFF]/", $pieces[$i])) {
                for ($j = 0, $words = array(); $j < strlen($pieces[$i]); $j += 2) {
                    $char = substr($pieces[$i], $j, 2);
                    if (array_key_exists($char, $_PINYIN_DB)) {
                        $words[] = $_PINYIN_DB[$char];
                    } else {
                        $words[] = $char;
                    }
                }
                $pieces[$i] = implode(" ", $words);
            }
        }
        break;
    }
    
    $pinyin = $pieces[0];
    for ($i = 1; $i < count($pieces); $i++) {
        // Insert a space
        if (    !preg_match("/\s$/", $pieces[$i-1])
                && !preg_match("/^\s/", $pieces[$i])) {
            $pinyin .= " ";
        }
        $pinyin .= $pieces[$i];
    }
    
    return $pinyin;
}

// _init_b52py: Initialize the Big5 to Pinyin database
function _init_b52py()
{
    global $_PINYIN_DB, $_PINYIN_SUP;
    // Already initialized
    if (defined("_PINYIN_DBTYPE")) {
        return;
    }
    
    // Decide the action to do
    if (file_exists(_PINYIN_B52PY_DB)) {
        // Not a file
        if (!is_file(_PINYIN_B52PY_DB)) {
            define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_ARRAY);
        // Not readable
        } elseif (!is_readable(_PINYIN_B52PY_DB)) {
            define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_ARRAY);
        // Not writable -- read only
        } elseif (!is_writable(_PINYIN_B52PY_DB)) {
            $_PINYIN_DB = dba_open(_PINYIN_B52PY_DB, "r", "gdbm");
            // No data
            if (dba_firstkey($_PINYIN_DB) === false) {
                dba_close($_PINYIN_DB);
                define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_ARRAY);
            // OK
            } else {
                define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_DBA);
                return;
            }
        // Writable
        } else {
            $_PINYIN_DB = dba_open(_PINYIN_B52PY_DB, "w", "gdbm");
            // No data - initialize the data
            if (dba_firstkey($_PINYIN_DB) === false) {
                define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_DBA);
            // OK
            } else {
                define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_DBA);
                return;
            }
        }
    // File does not exist
    } else {
        // Not creatable
        $error = check_writable(_PINYIN_B52PY_DB);
        if (!is_null($error)) {
            define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_ARRAY);
        // Creatable
        } else {
            mkalldir(dirname(_PINYIN_B52PY_DB));
            $_PINYIN_DB = dba_open(_PINYIN_B52PY_DB, "c", "gdbm");
            define("_PINYIN_DBTYPE", _PINYIN_DBTYPE_DBA);
        }
    }
    
    // Initialize the database
    switch (_PINYIN_DBTYPE) {
    // Initialize it as a DB file
    case _PINYIN_DBTYPE_DBA:
        $lines = explode("\n", xfread(_PINYIN_PINYIN2BIG5));
        foreach ($lines as $line) {
            // Skip comments
            if (substr($line, 0, 1) == "#") {
                continue;
            }
            // Skip empty lines
            if (!preg_match("/\S/", $line)) {
                continue;
            }
            $chars = explode(" ", $line);
            // First item is pinyin
            $pinyin = array_shift($chars);
            foreach ($chars as $char) {
                if (!dba_exists($char, $_PINYIN_DB)) {
                    dba_insert($char, $pinyin, $_PINYIN_DB);
                }
            }
        }
        // Special meta characters
        foreach (array_keys($_PINYIN_SUP) as $char) {
            if (!dba_exists($char, $_PINYIN_DB)) {
                dba_insert($char, $_PINYIN_SUP[$char], $_PINYIN_DB);
            }
        }
        break;
    
    // Initialize it as an array
    case _PINYIN_DBTYPE_ARRAY:
        $_PINYIN_DB = array();
        $lines = explode("\n", xfread(_PINYIN_PINYIN2BIG5));
        foreach ($lines as $line) {
            // Skip comments
            if (substr($line, 0, 1) == "#") {
                continue;
            }
            // Skip empty lines
            if (!preg_match("/\S/", $line)) {
                continue;
            }
            $chars = explode(" ", $line);
            // First item is pinyin
            $pinyin = array_shift($chars);
            foreach ($chars as $char) {
                $_PINYIN_DB[$char] = $pinyin;
            }
        }
        // Special meta characters
        foreach (array_keys($_PINYIN_SUP) as $char) {
            $_PINYIN_DB[$char] = $_PINYIN_SUP[$char];
        }
        break;
    }
    
    return;
}

?>