utfnormal.php

来自「php 开发的内容管理系统」· PHP 代码 · 共 793 行 · 第 1/2 页

<?php
# Copyright (C) 2004 Brion Vibber <[e-mail]>
# http://www.mediawiki.org/
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
# http://www.gnu.org/copyleft/gpl.html

/**
 * Unicode normalization routines for working with UTF-8 strings.
 * Currently assumes that input strings are valid UTF-8!
 *
 * Not as fast as I'd like, but should be usable for most purposes.
 * UtfNormal::toNFC() will bail early if given ASCII text or text
 * it can quickly deterimine is already normalized.
 *
 * All functions can be called static.
 *
 * See description of forms at http://www.unicode.org/reports/tr15/
 *
 * @package UtfNormal
 */

/** */
require_once 'UtfNormalUtil.php';

global $utfCombiningClass, $utfCanonicalComp, $utfCanonicalDecomp;
$utfCombiningClass = NULL;
$utfCanonicalComp = NULL;
$utfCanonicalDecomp = NULL;

# Load compatibility decompositions on demand if they are needed.
global $utfCompatibilityDecomp;
$utfCompatibilityDecomp = NULL;

define( 'UNICODE_HANGUL_FIRST', 0xac00 );
define( 'UNICODE_HANGUL_LAST',  0xd7a3 );

define( 'UNICODE_HANGUL_LBASE', 0x1100 );
define( 'UNICODE_HANGUL_VBASE', 0x1161 );
define( 'UNICODE_HANGUL_TBASE', 0x11a7 );

define( 'UNICODE_HANGUL_LCOUNT', 19 );
define( 'UNICODE_HANGUL_VCOUNT', 21 );
define( 'UNICODE_HANGUL_TCOUNT', 28 );
define( 'UNICODE_HANGUL_NCOUNT', UNICODE_HANGUL_VCOUNT * UNICODE_HANGUL_TCOUNT );

define( 'UNICODE_HANGUL_LEND', UNICODE_HANGUL_LBASE + UNICODE_HANGUL_LCOUNT - 1 );
define( 'UNICODE_HANGUL_VEND', UNICODE_HANGUL_VBASE + UNICODE_HANGUL_VCOUNT - 1 );
define( 'UNICODE_HANGUL_TEND', UNICODE_HANGUL_TBASE + UNICODE_HANGUL_TCOUNT - 1 );

define( 'UNICODE_SURROGATE_FIRST', 0xd800 );
define( 'UNICODE_SURROGATE_LAST', 0xdfff );
define( 'UNICODE_MAX', 0x10ffff );
define( 'UNICODE_REPLACEMENT', 0xfffd );


define( 'UTF8_HANGUL_FIRST', "\xea\xb0\x80" /*codepointToUtf8( UNICODE_HANGUL_FIRST )*/ );
define( 'UTF8_HANGUL_LAST', "\xed\x9e\xa3" /*codepointToUtf8( UNICODE_HANGUL_LAST )*/ );

define( 'UTF8_HANGUL_LBASE', "\xe1\x84\x80" /*codepointToUtf8( UNICODE_HANGUL_LBASE )*/ );
define( 'UTF8_HANGUL_VBASE', "\xe1\x85\xa1" /*codepointToUtf8( UNICODE_HANGUL_VBASE )*/ );
define( 'UTF8_HANGUL_TBASE', "\xe1\x86\xa7" /*codepointToUtf8( UNICODE_HANGUL_TBASE )*/ );

define( 'UTF8_HANGUL_LEND', "\xe1\x84\x92" /*codepointToUtf8( UNICODE_HANGUL_LEND )*/ );
define( 'UTF8_HANGUL_VEND', "\xe1\x85\xb5" /*codepointToUtf8( UNICODE_HANGUL_VEND )*/ );
define( 'UTF8_HANGUL_TEND', "\xe1\x87\x82" /*codepointToUtf8( UNICODE_HANGUL_TEND )*/ );

define( 'UTF8_SURROGATE_FIRST', "\xed\xa0\x80" /*codepointToUtf8( UNICODE_SURROGATE_FIRST )*/ );
define( 'UTF8_SURROGATE_LAST', "\xed\xbf\xbf" /*codepointToUtf8( UNICODE_SURROGATE_LAST )*/ );
define( 'UTF8_MAX', "\xf4\x8f\xbf\xbf" /*codepointToUtf8( UNICODE_MAX )*/ );
define( 'UTF8_REPLACEMENT', "\xef\xbf\xbd" /*codepointToUtf8( UNICODE_REPLACEMENT )*/ );
#define( 'UTF8_REPLACEMENT', '!' );

define( 'UTF8_OVERLONG_A', "\xc1\xbf" );
define( 'UTF8_OVERLONG_B', "\xe0\x9f\xbf" );
define( 'UTF8_OVERLONG_C', "\xf0\x8f\xbf\xbf" );

# These two ranges are illegal
define( 'UTF8_FDD0', "\xef\xb7\x90" /*codepointToUtf8( 0xfdd0 )*/ );
define( 'UTF8_FDEF', "\xef\xb7\xaf" /*codepointToUtf8( 0xfdef )*/ );
define( 'UTF8_FFFE', "\xef\xbf\xbe" /*codepointToUtf8( 0xfffe )*/ );
define( 'UTF8_FFFF', "\xef\xbf\xbf" /*codepointToUtf8( 0xffff )*/ );

define( 'UTF8_HEAD', false );
define( 'UTF8_TAIL', true );


/**
 * For using the ICU wrapper
 */
define( 'UNORM_NONE', 1 );
define( 'UNORM_NFD',  2 );
define( 'UNORM_NFKD', 3 );
define( 'UNORM_NFC',  4 );
define( 'UNORM_DEFAULT', UNORM_NFC );
define( 'UNORM_NFKC', 5 );
define( 'UNORM_FCD',  6 );

define( 'NORMALIZE_ICU', function_exists( 'utf8_normalize' ) );

/**
 *
 * @package MediaWiki
 */
class UtfNormal {
	/**
	 * The ultimate convenience function! Clean up invalid UTF-8 sequences,
	 * and convert to normal form C, canonical composition.
	 *
	 * Fast return for pure ASCII strings; some lesser optimizations for
	 * strings containing only known-good characters. Not as fast as toNFC().
	 *
	 * @param string $string a UTF-8 string
	 * @return string a clean, shiny, normalized UTF-8 string
	 */
	function cleanUp( $string ) {
		if( NORMALIZE_ICU ) {
			# We exclude a few chars that ICU would not.
			$string = preg_replace(
				'/[\x00-\x08\x0b\x0c\x0e-\x1f]/',
				UTF8_REPLACEMENT,
				$string );
			$string = str_replace( UTF8_FFFE, UTF8_REPLACEMENT, $string );
			$string = str_replace( UTF8_FFFF, UTF8_REPLACEMENT, $string );

			# UnicodeString constructor fails if the string ends with a
			# head byte. Add a junk char at the end, we'll strip it off.
			return rtrim( utf8_normalize( $string . "\x01", UNORM_NFC ), "\x01" );
		} elseif( UtfNormal::quickIsNFCVerify( $string ) ) {
			# Side effect -- $string has had UTF-8 errors cleaned up.
			return $string;
		} else {
			return UtfNormal::NFC( $string );
		}
	}

	/**
	 * Convert a UTF-8 string to normal form C, canonical composition.
	 * Fast return for pure ASCII strings; some lesser optimizations for
	 * strings containing only known-good characters.
	 *
	 * @param string $string a valid UTF-8 string. Input is not validated.
	 * @return string a UTF-8 string in normal form C
	 */
	function toNFC( $string ) {
		if( NORMALIZE_ICU )
			return utf8_normalize( $string, UNORM_NFC );
		elseif( UtfNormal::quickIsNFC( $string ) )
			return $string;
		else
			return UtfNormal::NFC( $string );
	}

	/**
	 * Convert a UTF-8 string to normal form D, canonical decomposition.
	 * Fast return for pure ASCII strings.
	 *
	 * @param string $string a valid UTF-8 string. Input is not validated.
	 * @return string a UTF-8 string in normal form D
	 */
	function toNFD( $string ) {
		if( NORMALIZE_ICU )
			return utf8_normalize( $string, UNORM_NFD );
		elseif( preg_match( '/[\x80-\xff]/', $string ) )
			return UtfNormal::NFD( $string );
		else
			return $string;
	}

	/**
	 * Convert a UTF-8 string to normal form KC, compatibility composition.
	 * This may cause irreversible information loss, use judiciously.
	 * Fast return for pure ASCII strings.
	 *
	 * @param string $string a valid UTF-8 string. Input is not validated.
	 * @return string a UTF-8 string in normal form KC
	 */
	function toNFKC( $string ) {
		if( NORMALIZE_ICU )
			return utf8_normalize( $string, UNORM_NFKC );
		elseif( preg_match( '/[\x80-\xff]/', $string ) )
			return UtfNormal::NFKC( $string );
		else
			return $string;
	}

	/**
	 * Convert a UTF-8 string to normal form KD, compatibility decomposition.
	 * This may cause irreversible information loss, use judiciously.
	 * Fast return for pure ASCII strings.
	 *
	 * @param string $string a valid UTF-8 string. Input is not validated.
	 * @return string a UTF-8 string in normal form KD
	 */
	function toNFKD( $string ) {
		if( NORMALIZE_ICU )
			return utf8_normalize( $string, UNORM_NFKD );
		elseif( preg_match( '/[\x80-\xff]/', $string ) )
			return UtfNormal::NFKD( $string );
		else
			return $string;
	}

	/**
	 * Load the basic composition data if necessary
	 * @private
	 */
	function loadData() {
		# fixme : are $utfCanonicalComp, $utfCanonicalDecomp really used?
		global $utfCombiningClass, $utfCanonicalComp, $utfCanonicalDecomp;
		if( !isset( $utfCombiningClass ) ) {
			require_once( 'UtfNormalData.inc' );
		}
	}

	/**
	 * Returns true if the string is _definitely_ in NFC.
	 * Returns false if not or uncertain.
	 * @param string $string a valid UTF-8 string. Input is not validated.
	 * @return bool
	 */
	function quickIsNFC( $string ) {
		# ASCII is always valid NFC!
		# If it's pure ASCII, let it through.
		if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;

		UtfNormal::loadData();
		global $utfCheckNFC, $utfCombiningClass;
		$len = strlen( $string );
		for( $i = 0; $i < $len; $i++ ) {
			$c = $string{$i};
			$n = ord( $c );
			if( $n < 0x80 ) {
				continue;
			} elseif( $n >= 0xf0 ) {
				$c = substr( $string, $i, 4 );
				$i += 3;
			} elseif( $n >= 0xe0 ) {
				$c = substr( $string, $i, 3 );
				$i += 2;
			} elseif( $n >= 0xc0 ) {
				$c = substr( $string, $i, 2 );
				$i++;
			}
			if( isset( $utfCheckNFC[$c] ) ) {
				# If it's NO or MAYBE, bail and do the slow check.
				return false;
			}
			if( isset( $utfCombiningClass[$c] ) ) {
				# Combining character? We might have to do sorting, at least.
				return false;
			}
		}
		return true;
	}

	/**
	 * Returns true if the string is _definitely_ in NFC.
	 * Returns false if not or uncertain.
	 * @param string $string a UTF-8 string, altered on output to be valid UTF-8 safe for XML.
	 */
	function quickIsNFCVerify( &$string ) {
		# Screen out some characters that eg won't be allowed in XML
		$string = preg_replace( '/[\x00-\x08\x0b\x0c\x0e-\x1f]/', UTF8_REPLACEMENT, $string );

		# ASCII is always valid NFC!
		# If we're only ever given plain ASCII, we can avoid the overhead
		# of initializing the decomposition tables by skipping out early.
		if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;

		static $checkit = null, $tailBytes = null, $utfCheckOrCombining = null;
		if( !isset( $checkit ) ) {
			# Load/build some scary lookup tables...
			UtfNormal::loadData();
			global $utfCheckNFC, $utfCombiningClass;

			$utfCheckOrCombining = array_merge( $utfCheckNFC, $utfCombiningClass );

			# Head bytes for sequences which we should do further validity checks
			$checkit = array_flip( array_map( 'chr',
					array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
						   0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
						   0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) ) );

			# Each UTF-8 head byte is followed by a certain
			# number of tail bytes.
			$tailBytes = array();
			for( $n = 0; $n < 256; $n++ ) {
				if( $n < 0xc0 ) {
					$remaining = 0;
				} elseif( $n < 0xe0 ) {
					$remaining = 1;
				} elseif( $n < 0xf0 ) {
					$remaining = 2;
				} elseif( $n < 0xf8 ) {
					$remaining = 3;
				} elseif( $n < 0xfc ) {
					$remaining = 4;
				} elseif( $n < 0xfe ) {
					$remaining = 5;
				} else {
					$remaining = 0;
				}
				$tailBytes[chr($n)] = $remaining;
			}
		}

		# Chop the text into pure-ASCII and non-ASCII areas;
		# large ASCII parts can be handled much more quickly.
		# Don't chop up Unicode areas for punctuation, though,
		# that wastes energy.
		preg_match_all(
			'/([\x00-\x7f]+|[\x80-\xff][\x00-\x40\x5b-\x5f\x7b-\xff]*)/',
			$string, $matches );

		$looksNormal = true;
		$base = 0;
		$replace = array();
		foreach( $matches[1] as $str ) {
			$chunk = strlen( $str );

			if( $str{0} < "\x80" ) {
				# ASCII chunk: guaranteed to be valid UTF-8
				# and in normal form C, so skip over it.
				$base += $chunk;
				continue;
			}

			# We'll have to examine the chunk byte by byte to ensure
			# that it consists of valid UTF-8 sequences, and to see
			# if any of them might not be normalized.
			#
			# Since PHP is not the fastest language on earth, some of
			# this code is a little ugly with inner loop optimizations.

			$head = '';
			$len = $chunk + 1; # Counting down is faster. I'm *so* sorry.

			for( $i = -1; --$len; ) {
				if( $remaining = $tailBytes[$c = $str{++$i}] ) {
					# UTF-8 head byte!
					$sequence = $head = $c;
					do {
						# Look for the defined number of tail bytes...
						if( --$len && ( $c = $str{++$i} ) >= "\x80" && $c < "\xc0" ) {
							# Legal tail bytes are nice.
							$sequence .= $c;
						} else {
							if( 0 == $len ) {
								# Premature end of string!
								# Drop a replacement character into output to
								# represent the invalid UTF-8 sequence.
								$replace[] = array( UTF8_REPLACEMENT,
													$base + $i + 1 - strlen( $sequence ),
													strlen( $sequence ) );
								break 2;
							} else {
								# Illegal tail byte; abandon the sequence.
								$replace[] = array( UTF8_REPLACEMENT,
													$base + $i - strlen( $sequence ),
													strlen( $sequence ) );
								# Back up and reprocess this byte; it may itself
								# be a legal ASCII or UTF-8 sequence head.
								--$i;
								++$len;
								continue 2;
							}
						}
					} while( --$remaining );

					if( isset( $checkit[$head] ) ) {
						# Do some more detailed validity checks, for
						# invalid characters and illegal sequences.
						if( $head == "\xed" ) {
							# 0xed is relatively frequent in Korean, which
							# abuts the surrogate area, so we're doing
							# this check separately to speed things up.

							if( $sequence >= UTF8_SURROGATE_FIRST ) {
								# Surrogates are legal only in UTF-16 code.
								# They are totally forbidden here in UTF-8
								# utopia.
								$replace[] = array( UTF8_REPLACEMENT,
								             $base + $i + 1 - strlen( $sequence ),
								             strlen( $sequence ) );