utf_normalizer.php

这些都是我以前学习是用到的源码

				}
				else
				{
					// The char is not decomposable
					$utf_seq = array($utf_char);
				}


				// STEP 2: Capture the starter

				// Check out the combining class of the first character of the UTF sequence
				$k = 0;
				if (isset($utf_combining_class[$utf_seq[0]]) || $qc[$utf_char] == UNICODE_QC_MAYBE)
				{
					// Not a starter, inspect previous characters
					// The last 8 characters are kept in a buffer so that we don't have to capture them everytime.
					// This is enough for all real-life strings but even if it wasn't, we can capture characters in backward mode,
					// although it is slower than this method.
					//
					// In the following loop, $j starts at the previous buffered character ($i - 1, because current character is
					// at offset $i) and process them in backward mode until we find a starter.
					//
					// $k is the index on each UTF character inside of our UTF sequence. At this time, $utf_seq contains one or more
					// characters numbered 0 to n. $k starts at 0 and for each char we prepend we pre-decrement it and for numbering
					$starter_found = 0;
					$j_min = max(1, $i - 7);

					for ($j = $i - 1; $j >= $j_min && $lpos > $tmp_pos; --$j)
					{
						$utf_char = $buffer[$j & 7];
						$lpos -= strlen($utf_char);

						if (isset($decomp_map[$utf_char]))
						{
							// The char is a composite, decompose for storage
							$decomp_seq = array();
							$_pos = 0;
							$_len = strlen($decomp_map[$utf_char]);

							do
							{
								$c = $decomp_map[$utf_char][$_pos];
								$_utf_len =& $utf_len_mask[$c & "\xF0"];

								if (isset($_utf_len))
								{
									$decomp_seq[] = substr($decomp_map[$utf_char], $_pos, $_utf_len);
									$_pos += $_utf_len;
								}
								else
								{
									$decomp_seq[] = $c;
									++$_pos;
								}
							}
							while ($_pos < $_len);

							// Prepend the UTF sequence with our decomposed sequence
							if (isset($decomp_seq[1]))
							{
								// The char expanded into several chars
								$decomp_cnt = sizeof($decomp_seq);

								foreach ($decomp_seq as $decomp_i => $decomp_char)
								{
									$utf_seq[$k + $decomp_i - $decomp_cnt] = $decomp_char;
								}
								$k -= $decomp_cnt;
							}
							else
							{
								// Decomposed to a single char, easier to prepend
								$utf_seq[--$k] = $decomp_seq[0];
							}
						}
						else
						{
							$utf_seq[--$k] = $utf_char;
						}

						if (!isset($utf_combining_class[$utf_seq[$k]]))
						{
							// We have found our starter
							$starter_found = 1;
							break;
						}
					}

					if (!$starter_found && $lpos > $tmp_pos)
					{
						// The starter was not found in the buffer, let's rewind some more
						do
						{
							// $utf_len_mask contains the masks of both leading bytes and trailing bytes. If $utf_en > 0 then it's a leading byte, otherwise it's a trailing byte.
							$c = $str[--$lpos];
							$c_mask = $c & "\xF0";

							if (isset($utf_len_mask[$c_mask]))
							{
								// UTF byte
								if ($utf_len = $utf_len_mask[$c_mask])
								{
									// UTF *leading* byte
									$utf_char = substr($str, $lpos, $utf_len);

									if (isset($decomp_map[$utf_char]))
									{
										// Decompose the character
										$decomp_seq = array();
										$_pos = 0;
										$_len = strlen($decomp_map[$utf_char]);

										do
										{
											$c = $decomp_map[$utf_char][$_pos];
											$_utf_len =& $utf_len_mask[$c & "\xF0"];

											if (isset($_utf_len))
											{
												$decomp_seq[] = substr($decomp_map[$utf_char], $_pos, $_utf_len);
												$_pos += $_utf_len;
											}
											else
											{
												$decomp_seq[] = $c;
												++$_pos;
											}
										}
										while ($_pos < $_len);

										// Prepend the UTF sequence with our decomposed sequence
										if (isset($decomp_seq[1]))
										{
											// The char expanded into several chars
											$decomp_cnt = sizeof($decomp_seq);
											foreach ($decomp_seq as $decomp_i => $utf_char)
											{
												$utf_seq[$k + $decomp_i - $decomp_cnt] = $utf_char;
											}
											$k -= $decomp_cnt;
										}
										else
										{
											// Decomposed to a single char, easier to prepend
											$utf_seq[--$k] = $decomp_seq[0];
										}
									}
									else
									{
										$utf_seq[--$k] = $utf_char;
									}
								}
							}
							else
							{
								// ASCII char
								$utf_seq[--$k] = $c;
							}
						}
						while ($lpos > $tmp_pos);
					}
				}


				// STEP 3: Capture following combining modifiers

				while ($pos < $len)
				{
					$c_mask = $str[$pos] & "\xF0";

					if (isset($utf_len_mask[$c_mask]))
					{
						if ($utf_len = $utf_len_mask[$c_mask])
						{
							$utf_char = substr($str, $pos, $utf_len);
						}
						else
						{
							// A trailing byte came out of nowhere
							// Trailing bytes are replaced with Unicode replacement chars, we will just ignore it for now, break out of the loop
							// as if it was a starter (replacement chars ARE starters) and let the next loop replace it
							break;
						}

						if (isset($utf_combining_class[$utf_char]) || isset($qc[$utf_char]))
						{
							// Combining character, add it to the sequence and move the cursor
							if (isset($decomp_map[$utf_char]))
							{
								// Decompose the character
								$_pos = 0;
								$_len = strlen($decomp_map[$utf_char]);

								do
								{
									$c = $decomp_map[$utf_char][$_pos];
									$_utf_len =& $utf_len_mask[$c & "\xF0"];

									if (isset($_utf_len))
									{
										$utf_seq[] = substr($decomp_map[$utf_char], $_pos, $_utf_len);
										$_pos += $_utf_len;
									}
									else
									{
										$utf_seq[] = $c;
										++$_pos;
									}
								}
								while ($_pos < $_len);
							}
							else
							{
								$utf_seq[] = $utf_char;
							}

							$pos += $utf_len;
						}
						else
						{
							// Combining class 0 and no QC, break out of the loop
							// Note: we do not know if that character is valid. If it's not, the next iteration will replace it
							break;
						}
					}
					else
					{
						// ASCII chars are starters
						break;
					}
				}


				// STEP 4: Sort and combine

				// Here we sort...
				$k_max = $k + sizeof($utf_seq);

				if (!$k && $k_max == 1)
				{
					// There is only one char in the UTF sequence, add it then jump to the next iteration of main loop
						// Note: the two commented lines below can be enabled under PHP5 for a very small performance gain in most cases
//						if (substr_compare($str, $utf_seq[0], $lpos, $pos - $lpos))
//						{
						$tmp .= substr($str, $tmp_pos, $lpos - $tmp_pos) . $utf_seq[0];
						$tmp_pos = $pos;
//						}

					continue;
				}

				// ...there we combine
				if (isset($utf_combining_class[$utf_seq[$k]]))
				{
					$starter = $nf_seq = '';
				}
				else
				{
					$starter = $utf_seq[$k++];
					$nf_seq = '';
				}
				$utf_sort = array();

				// We add an empty char at the end of the UTF char sequence. It will act as a starter and trigger the sort/combine routine
				// at the end of the string without altering it
				$utf_seq[] = '';

				do
				{
					$utf_char = $utf_seq[$k++];

					if (isset($utf_combining_class[$utf_char]))
					{
						$utf_sort[$utf_combining_class[$utf_char]][] = $utf_char;
					}
					else
					{
						if (empty($utf_sort))
						{
							// No combining characters... check for a composite of the two starters
							if (isset($utf_canonical_comp[$starter . $utf_char]))
							{
								// Good ol' composite character
								$starter = $utf_canonical_comp[$starter . $utf_char];
							}
							else if (isset($utf_jamo_type[$utf_char]))
							{
								// Current char is a composable jamo
								if (isset($utf_jamo_type[$starter]) && $utf_jamo_type[$starter] == UNICODE_JAMO_L && $utf_jamo_type[$utf_char] == UNICODE_JAMO_V)
								{
									// We have a L jamo followed by a V jamo, we are going to prefetch the next char to see if it's a T jamo
									if (isset($utf_jamo_type[$utf_seq[$k]]) && $utf_jamo_type[$utf_seq[$k]] == UNICODE_JAMO_T)
									{
										// L+V+T jamos, combine to a LVT Hangul syllable ($k is incremented)
										$cp = $utf_jamo_index[$starter] + $utf_jamo_index[$utf_char] + $utf_jamo_index[$utf_seq[$k]];
										++$k;
									}
									else
									{
										// L+V jamos, combine to a LV Hangul syllable
										$cp = $utf_jamo_index[$starter] + $utf_jamo_index[$utf_char];
									}

									$starter = chr(0xE0 | ($cp >> 12)) . chr(0x80 | (($cp >> 6) & 0x3F)) . chr(0x80 | ($cp & 0x3F));
								}
								else
								{
									// Non-composable jamo, just add it to the sequence
									$nf_seq .= $starter;
									$starter = $utf_char;
								}
							}
							else
							{
								// No composite, just add the first starter to the sequence then continue with the other one
								$nf_seq .= $starter;
								$starter = $utf_char;
							}
						}
						else
						{
							ksort($utf_sort);

							// For each class of combining characters
							foreach ($utf_sort as $cc => $utf_chars)
							{
								$j = 0;

								do
								{
									// Look for a composite
									if (isset($utf_canonical_comp[$starter . $utf_chars[$j]]))
									{
										// Found a composite, replace the starter
										$starter = $utf_canonical_comp[$starter . $utf_chars[$j]];
										unset($utf_sort[$cc][$j]);
									}
									else
									{
										// No composite, all following characters in that class are blocked
										break;
									}
								}
								while (isset($utf_sort[$cc][++$j]));
							}

							// Add the starter to the normalized sequence, followed by non-starters in canonical order
							$nf_seq .= $starter;

							foreach ($utf_sort as $utf_chars)
							{
								if (!empty($utf_chars))
								{
									$nf_seq .= implode('', $utf_chars);
								}
							}

							// Reset the array and go on
							$utf_sort = array();
							$starter = $utf_char;
						}
					}
				}
				while ($k <= $k_max);

				$tmp .= substr($str, $tmp_pos, $lpos - $tmp_pos) . $nf_seq;
				$tmp_pos = $pos;
			}
			else
			{
				// Only a ASCII char can make the program get here
				//
				// First we skip the current byte with ++$pos, then we quickly skip following ASCII chars with strspn().
				//
				// The first two "if"'s here can be removed, with the consequences of being faster on latin text (lots of ASCII) and slower on
				// multi-byte text (where the only ASCII chars are spaces and punctuation)
				if (++$pos != $len)
				{
					if ($str[$pos] < "\x80")
					{
						$pos += strspn($str, UTF8_ASCII_RANGE, ++$pos);
						$buffer[++$i & 7] = $str[$pos - 1];
					}
					else
					{
						$buffer[++$i & 7] = $c;
					}
				}
			}
		}
		while ($pos < $len);

		// Now is time to return the string
		if ($tmp_pos)
		{
			// If the $tmp_pos cursor is not at the beggining of the string then at least one character was not in normal form. Replace $str with the fixed version
			if ($tmp_pos == $len)
			{
				// The $tmp_pos cursor is at the end of $str, therefore $tmp holds the whole $str
				return $tmp;
			}
			else
			{
				// The rightmost chunk of $str has not been appended to $tmp yet
				return $tmp . substr($str, $tmp_pos);
			}
		}

		// The string was already in normal form
		return $str;
	}

	/**
	* Decompose a UTF string
	*
	* @param	string	$str			UTF string
	* @param	integer	$pos			Position of the first UTF char (in bytes)
	* @param	integer	$len			Length of the string (in bytes)
	* @param	array	&$decomp_map	Decomposition mapping, passed by reference but never modified
	* @return	string					The string, decomposed and sorted canonically
	*
	* @access	private
	*/
	function decompose($str, $pos, $len, &$decomp_map)
	{
		global $utf_combining_class, $phpbb_root_path;

		// Load some commonly-used tables
		if (!isset($utf_combining_class))
		{
			include($phpbb_root_path . 'includes/utf/data/utf_normalizer_common.php');
		}

		// UTF char length array
		$utf_len_mask = array(
			// Leading bytes masks
			"\xC0" => 2, "\xD0" => 2, "\xE0" => 3, "\xF0" => 4,
			// Trailing bytes masks
			"\x80" => 0, "\x90" => 0, "\xA0" => 0, "\xB0" => 0
		);

		// Some extra checks are triggered on the first byte of a UTF sequence
		$extra_check = array(
			"\xED" => 1, "\xEF" => 1, "\xC0" => 1, "\xC1" => 1, "\xE0" => 1, "\xF0" => 1,
			"\xF4" => 1, "\xF5" => 1, "\xF6" => 1, "\xF7" => 1, "\xF8" => 1, "\xF9" => 1,
			"\xFA" => 1, "\xFB" => 1, "\xFC" => 1, "\xFD" => 1, "\xFE" => 1, "\xFF" => 1
		);

		// These masks are used to check if a UTF sequence is well formed. Here are the only 3 lengths we acknowledge:
		//   - 2-byte: 110? ???? 10?? ????
		//   - 3-byte: 1110 ???? 10?? ???? 10?? ????
		//   - 4-byte: 1111 0??? 10?? ???? 10?? ???? 10?? ????
		// Note that 5- and 6- byte sequences are automatically discarded
		$utf_validation_mask = array(
			2	=> "\xE0\xC0",
			3	=> "\xF0\xC0\xC0",
			4	=> "\xF8\xC0\xC0\xC0"
		);

		$utf_validation_check = array(
			2	=> "\xC0\x80",
			3	=> "\xE0\x80\x80",
			4	=> "\xF0\x80\x80\x80"
		);

		$tmp = '';
		$starter_pos = $pos;
		$tmp_pos = $last_cc = $sort = $dump = 0;
		$utf_sort = array();


		// Main loop
		do
		{
			// STEP 0: Capture the current char

			$cur_mask = $str[$pos] & "\xF0";
			if (isset($utf_len_mask[$cur_mask]))
			{
				if ($utf_len = $utf_len_mask[$cur_mask])
				{
					// Multibyte char
					$utf_char = substr($str, $pos, $utf_len);
					$pos += $utf_len;
				}
				else
				{
					// A trailing byte came out of nowhere, we will treat it and all following trailing bytes as if each of them was a Unicode
					// replacement char and we will advance the cursor
					$spn = strspn($str, UTF8_TRAILING_BYTES, $pos);

					if ($dump)
					{
						$tmp .= substr($str, $tmp_pos, $starter_pos - $tmp_pos);

						// Dump combiners
						if (!empty($utf_sort))