utfnormal.php

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

								$head = '';
								continue;
							}
						} else {
							# Slower, but rarer checks...
							$n = ord( $head );
							if(
								# "Overlong sequences" are those that are syntactically
								# correct but use more UTF-8 bytes than are necessary to
								# encode a character. Na茂ve string comparisons can be
								# tricked into failing to see a match for an ASCII
								# character, for instance, which can be a security hole
								# if blacklist checks are being used.
							       ($n  < 0xc2 && $sequence <= UTF8_OVERLONG_A)
								|| ($n == 0xe0 && $sequence <= UTF8_OVERLONG_B)
								|| ($n == 0xf0 && $sequence <= UTF8_OVERLONG_C)

								# U+FFFE and U+FFFF are explicitly forbidden in Unicode.
								|| ($n == 0xef &&
									   ($sequence == UTF8_FFFE)
									|| ($sequence == UTF8_FFFF) )

								# Unicode has been limited to 21 bits; longer
								# sequences are not allowed.
								|| ($n >= 0xf0 && $sequence > UTF8_MAX) ) {

								$replace[] = array( UTF8_REPLACEMENT,
								                    $base + $i + 1 - strlen( $sequence ),
								                    strlen( $sequence ) );
								$head = '';
								continue;
							}
						}
					}

					if( isset( $utfCheckOrCombining[$sequence] ) ) {
						# If it's NO or MAYBE, we'll have to rip
						# the string apart and put it back together.
						# That's going to be mighty slow.
						$looksNormal = false;
					}

					# The sequence is legal!
					$head = '';
				} elseif( $c < "\x80" ) {
					# ASCII byte.
					$head = '';
				} elseif( $c < "\xc0" ) {
					# Illegal tail bytes
					if( $head == '' ) {
						# Out of the blue!
						$replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
					} else {
						# Don't add if we're continuing a broken sequence;
						# we already put a replacement character when we looked
						# at the broken sequence.
						$replace[] = array( '', $base + $i, 1 );
					}
				} else {
					# Miscellaneous freaks.
					$replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
					$head = '';
				}
			}
			$base += $chunk;
		}
		if( count( $replace ) ) {
			# There were illegal UTF-8 sequences we need to fix up.
			$out = '';
			$last = 0;
			foreach( $replace as $rep ) {
				list( $replacement, $start, $length ) = $rep;
				if( $last < $start ) {
					$out .= substr( $string, $last, $start - $last );
				}
				$out .= $replacement;
				$last = $start + $length;
			}
			if( $last < strlen( $string ) ) {
				$out .= substr( $string, $last );
			}
			$string = $out;
		}
		return $looksNormal;
	}

	# These take a string and run the normalization on them, without
	# checking for validity or any optimization etc. Input must be
	# VALID UTF-8!
	/**
	 * @param string $string
	 * @return string
	 * @private
	 */
	function NFC( $string ) {
		return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
	}

	/**
	 * @param string $string
	 * @return string
	 * @private
	 */
	function NFD( $string ) {
		UtfNormal::loadData();
		global $utfCanonicalDecomp;
		return UtfNormal::fastCombiningSort(
			UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
	}

	/**
	 * @param string $string
	 * @return string
	 * @private
	 */
	function NFKC( $string ) {
		return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
	}

	/**
	 * @param string $string
	 * @return string
	 * @private
	 */
	function NFKD( $string ) {
		global $utfCompatibilityDecomp;
		if( !isset( $utfCompatibilityDecomp ) ) {
			require_once( 'UtfNormalDataK.inc' );
		}
		return UtfNormal::fastCombiningSort(
			UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
	}


	/**
	 * Perform decomposition of a UTF-8 string into either D or KD form
	 * (depending on which decomposition map is passed to us).
	 * Input is assumed to be *valid* UTF-8. Invalid code will break.
	 * @private
	 * @param string $string Valid UTF-8 string
	 * @param array $map hash of expanded decomposition map
	 * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
	 */
	function fastDecompose( $string, &$map ) {
		UtfNormal::loadData();
		$len = strlen( $string );
		$out = '';
		for( $i = 0; $i < $len; $i++ ) {
			$c = $string{$i};
			$n = ord( $c );
			if( $n < 0x80 ) {
				# ASCII chars never decompose
				# THEY ARE IMMORTAL
				$out .= $c;
				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( $map[$c] ) ) {
				$out .= $map[$c];
				continue;
			} else {
				if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
					# Decompose a hangul syllable into jamo;
					# hardcoded for three-byte UTF-8 sequence.
					# A lookup table would be slightly faster,
					# but adds a lot of memory & disk needs.
					#
					$index = ( (ord( $c{0} ) & 0x0f) << 12
					         | (ord( $c{1} ) & 0x3f) <<  6
					         | (ord( $c{2} ) & 0x3f) )
					       - UNICODE_HANGUL_FIRST;
					$l = intval( $index / UNICODE_HANGUL_NCOUNT );
					$v = intval( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
					$t = $index % UNICODE_HANGUL_TCOUNT;
					$out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
					if( $t >= 25 ) {
						$out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
					} elseif( $t ) {
						$out .= "\xe1\x86" . chr( 0xa7 + $t );
					}
					continue;
				}
			}
			$out .= $c;
		}
		return $out;
	}

	/**
	 * Sorts combining characters into canonical order. This is the
	 * final step in creating decomposed normal forms D and KD.
	 * @private
	 * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
	 * @return string a UTF-8 string with combining characters sorted in canonical order
	 */
	function fastCombiningSort( $string ) {
		UtfNormal::loadData();
		global $utfCombiningClass;
		$len = strlen( $string );
		$out = '';
		$combiners = array();
		$lastClass = -1;
		for( $i = 0; $i < $len; $i++ ) {
			$c = $string{$i};
			$n = ord( $c );
			if( $n >= 0x80 ) {
				if( $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( $utfCombiningClass[$c] ) ) {
					$lastClass = $utfCombiningClass[$c];
					@$combiners[$lastClass] .= $c;
					continue;
				}
			}
			if( $lastClass ) {
				ksort( $combiners );
				$out .= implode( '', $combiners );
				$combiners = array();
			}
			$out .= $c;
			$lastClass = 0;
		}
		if( $lastClass ) {
			ksort( $combiners );
			$out .= implode( '', $combiners );
		}
		return $out;
	}

	/**
	 * Produces canonically composed sequences, i.e. normal form C or KC.
	 *
	 * @private
	 * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
	 * @return string a UTF-8 string with canonical precomposed characters used where possible
	 */
	function fastCompose( $string ) {
		UtfNormal::loadData();
		global $utfCanonicalComp, $utfCombiningClass;
		$len = strlen( $string );
		$out = '';
		$lastClass = -1;
		$lastHangul = 0;
		$startChar = '';
		$combining = '';
		$x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
		$x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
		for( $i = 0; $i < $len; $i++ ) {
			$c = $string{$i};
			$n = ord( $c );
			if( $n < 0x80 ) {
				# No combining characters here...
				$out .= $startChar;
				$out .= $combining;
				$startChar = $c;
				$combining = '';
				$lastClass = 0;
				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++;
			}
			$pair = $startChar . $c;
			if( $n > 0x80 ) {
				if( isset( $utfCombiningClass[$c] ) ) {
					# A combining char; see what we can do with it
					$class = $utfCombiningClass[$c];
					if( !empty( $startChar ) &&
						$lastClass < $class &&
						$class > 0 &&
						isset( $utfCanonicalComp[$pair] ) ) {
						$startChar = $utfCanonicalComp[$pair];
						$class = 0;
					} else {
						$combining .= $c;
					}
					$lastClass = $class;
					$lastHangul = 0;
					continue;
				}
			}
			# New start char
			if( $lastClass == 0 ) {
				if( isset( $utfCanonicalComp[$pair] ) ) {
					$startChar = $utfCanonicalComp[$pair];
					$lastHangul = 0;
					continue;
				}
				if( $n >= $x1 && $n <= $x2 ) {
					# WARNING: Hangul code is painfully slow.
					# I apologize for this ugly, ugly code; however
					# performance is even more teh suck if we call
					# out to nice clean functions. Lookup tables are
					# marginally faster, but require a lot of space.
					#
					if( $c >= UTF8_HANGUL_VBASE &&
						$c <= UTF8_HANGUL_VEND &&
						$startChar >= UTF8_HANGUL_LBASE &&
						$startChar <= UTF8_HANGUL_LEND ) {
						#
						#$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
						#$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
						$lIndex = ord( $startChar{2} ) - 0x80;
						$vIndex = ord( $c{2}         ) - 0xa1;

						$hangulPoint = UNICODE_HANGUL_FIRST +
							UNICODE_HANGUL_TCOUNT *
							(UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);

						# Hardcode the limited-range UTF-8 conversion:
						$startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
									 chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
									 chr( $hangulPoint       & 0x3f | 0x80 );
						$lastHangul = 0;
						continue;
					} elseif( $c >= UTF8_HANGUL_TBASE &&
							  $c <= UTF8_HANGUL_TEND &&
							  $startChar >= UTF8_HANGUL_FIRST &&
							  $startChar <= UTF8_HANGUL_LAST &&
							  !$lastHangul ) {
						# $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
						$tIndex = ord( $c{2} ) - 0xa7;
						if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);

						# Increment the code point by $tIndex, without
						# the function overhead of decoding and recoding UTF-8
						#
						$tail = ord( $startChar{2} ) + $tIndex;
						if( $tail > 0xbf ) {
							$tail -= 0x40;
							$mid = ord( $startChar{1} ) + 1;
							if( $mid > 0xbf ) {
								$startChar{0} = chr( ord( $startChar{0} ) + 1 );
								$mid -= 0x40;
							}
							$startChar{1} = chr( $mid );
						}
						$startChar{2} = chr( $tail );

						# If there's another jamo char after this, *don't* try to merge it.
						$lastHangul = 1;
						continue;
					}
				}
			}
			$out .= $startChar;
			$out .= $combining;
			$startChar = $c;
			$combining = '';
			$lastClass = 0;
			$lastHangul = 0;
		}
		$out .= $startChar . $combining;
		return $out;
	}

	/**
	 * This is just used for the benchmark, comparing how long it takes to
	 * interate through a string without really doing anything of substance.
	 * @param string $string
	 * @return string
	 */
	function placebo( $string ) {
		$len = strlen( $string );
		$out = '';
		for( $i = 0; $i < $len; $i++ ) {
			$out .= $string{$i};
		}
		return $out;
	}
}

?>