finediff.php

<?php
/**
 * FINE granularity DIFF
 *
 * Computes a set of instructions to convert the content of
 * one string into another.
 *
 * Copyright (c) 2011 Raymond Hill (http://raymondhill.net/blog/?p=441)
 *
 * Licensed under The MIT License
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 *
 * @copyright Copyright 2011 (c) Raymond Hill (http://raymondhill.net/blog/?p=441)
 * @link http://www.raymondhill.net/finediff/
 * @version 0.6
 * @license MIT License (http://www.opensource.org/licenses/mit-license.php)
 *
 * 10-Dec-2011 (Christoph Mewes):
 *   - added UTF-8 support, fixed strange usage of htmlentities
 */

mb_internal_encoding('UTF-8');

/**
 * Usage (simplest):
 *
 *   include 'finediff.php';
 *
 *   // for the stock stack, granularity values are:
 *   // FineDiff::$paragraphGranularity = paragraph/line level
 *   // FineDiff::$sentenceGranularity = sentence level
 *   // FineDiff::$wordGranularity = word level
 *   // FineDiff::$characterGranularity = character level [default]
 *
 *   $opcodes = FineDiff::getDiffOpcodes($from_text, $to_text [, $granularityStack = null] );
 *   // store opcodes for later use...
 *
 *   ...
 *
 *   // restore $to_text from $from_text + $opcodes
 *   include 'finediff.php';
 *   $to_text = FineDiff::renderToTextFromOpcodes($from_text, $opcodes);
 *
 *   ...
 */

/**
 * Persisted opcodes (string) are a sequence of atomic opcode.
 * A single opcode can be one of the following:
 *   c | c{n} | d | d{n} | i:{c} | i{length}:{s}
 *   'c'        = copy one character from source
 *   'c{n}'     = copy n characters from source
 *   'd'        = skip one character from source
 *   'd{n}'     = skip n characters from source
 *   'i:{c}     = insert character 'c'
 *   'i{n}:{s}' = insert string s, which is of length n
 *
 * Do not exist as of now, under consideration:
 *   'm{n}:{o}  = move n characters from source o characters ahead.
 *   It would be essentially a shortcut for a delete->copy->insert
 *   command (swap) for when the inserted segment is exactly the same
 *   as the deleted one, and with only a copy operation in between.
 *   TODO: How often this case occurs? Is it worth it? Can only
 *   be done as a postprocessing method (->optimize()?)
 */
abstract class FineDiffOp {
    abstract public function getFromLen();
    abstract public function getToLen();
    abstract public function getOpcode();
}

class FineDiffDeleteOp extends FineDiffOp {
    public function __construct($len) {
        $this->fromLen = $len;
    }
    public function getFromLen() {
        return $this->fromLen;
    }
    public function getToLen() {
        return 0;
    }
    public function getOpcode() {
        if ( $this->fromLen === 1 ) {
            return 'd';
        }
        return "d{$this->fromLen}";
    }
}

class FineDiffInsertOp extends FineDiffOp {
    public $text;

    public function __construct($text) {
        $this->text = $text;
    }
    public function getFromLen() {
        return 0;
    }
    public function getToLen() {
        return mb_strlen($this->text);
    }
    public function getText() {
        return $this->text;
    }
    public function getOpcode() {
        $to_len = mb_strlen($this->text);
        if ( $to_len === 1 ) {
            return "i:{$this->text}";
        }
        return "i{$to_len}:{$this->text}";
    }
}

class FineDiffReplaceOp extends FineDiffOp {
    public $fromLen;
    public $text;

    public function __construct($fromLen, $text) {
        $this->fromLen = $fromLen;
        $this->text = $text;
    }
    public function getFromLen() {
        return $this->fromLen;
    }
    public function getToLen() {
        return mb_strlen($this->text);
    }
    public function getText() {
        return $this->text;
    }
    public function getOpcode() {
        if ( $this->fromLen === 1 ) {
            $del_opcode = 'd';
        }
        else {
            $del_opcode = "d{$this->fromLen}";
        }
        $to_len = mb_strlen($this->text);
        if ( $to_len === 1 ) {
            return "{$del_opcode}i:{$this->text}";
        }
        return "{$del_opcode}i{$to_len}:{$this->text}";
    }
}

class FineDiffCopyOp extends FineDiffOp {
    public $len;

    public function __construct($len) {
        $this->len = $len;
    }
    public function getFromLen() {
        return $this->len;
    }
    public function getToLen() {
        return $this->len;
    }
    public function getOpcode() {
        if ( $this->len === 1 ) {
            return 'c';
        }
        return "c{$this->len}";
    }
    public function increase($size) {
        return $this->len += $size;
    }
}

/**
 * FineDiff ops
 *
 * Collection of ops
 */
class FineDiffOps {
    public $edits = array();

    public function appendOpcode($opcode, $from, $from_offset, $from_len) {
        if ( $opcode === 'c' ) {
            $edits[] = new FineDiffCopyOp($from_len);
        }
        else if ( $opcode === 'd' ) {
            $edits[] = new FineDiffDeleteOp($from_len);
        }
        else /* if ( $opcode === 'i' ) */ {
            $edits[] = new FineDiffInsertOp(mb_substr($from, $from_offset, $from_len));
        }
    }
}

/**
 * FineDiff class
 *
 * TODO: Document
 *
 */
class FineDiff {
    /**
     * Stock granularity stacks and delimiters
     */

    const paragraphDelimiters = "\n\r";
    public static $paragraphGranularity = array(
        FineDiff::paragraphDelimiters
    );
    const sentenceDelimiters = ".\n\r";
    public static $sentenceGranularity = array(
        FineDiff::paragraphDelimiters,
        FineDiff::sentenceDelimiters
    );
    const wordDelimiters = " \t.\n\r";
    public static $wordGranularity = array(
        FineDiff::paragraphDelimiters,
        FineDiff::sentenceDelimiters,
        FineDiff::wordDelimiters
    );
    const characterDelimiters = "";
    public static $characterGranularity = array(
        FineDiff::paragraphDelimiters,
        FineDiff::sentenceDelimiters,
        FineDiff::wordDelimiters,
        FineDiff::characterDelimiters
    );

    public static $textStack = array(
        ".",
        " \t.\n\r",
        ""
    );

    public $granularityStack;
    public $edits;
    public $from_text;
    public $last_edit;
    public $stackpointer;
    public $from_offset;
    public $insertions_count;
    public $deletions_count;

    /**------------------------------------------------------------------------
     *
     * Public section
     *
     */

    /**
     * Constructor
     * ...
     * The $granularityStack allows FineDiff to be configurable so that
     * a particular stack tailored to the specific content of a document can
     * be passed.
     */
    public function __construct($from_text = '', $to_text = '', $granularityStack = null) {
        // setup stack for generic text documents by default
        $this->insertions_count = 0;
        $this->deletions_count = 0;
        $this->granularityStack = $granularityStack ? $granularityStack : FineDiff::$characterGranularity;
        $this->edits = array();
        $this->from_text = $from_text;
        $this->doDiff($from_text, $to_text);
    }

    /**
     * Magic __toString() method
     * ...
     * Assuming `$diff = new FineDiff($from, $to);`, `echo $diff` will print the HTML diff.
     */
    public function __toString() {
        return $this->renderDiffToHTML();
    }

    public function getOps() {
        return $this->edits;
    }

    public function getOpcodes() {
        $opcodes = array();
        foreach ( $this->edits as $edit ) {
            $opcodes[] = $edit->getOpcode();
        }
        return implode('', $opcodes);
    }

    public function renderDiffToHTML() {
        $in_offset = 0;
        ob_start();
        foreach ( $this->edits as $edit ) {
            $n = $edit->getFromLen();
            if ( $edit instanceof FineDiffCopyOp ) {
                FineDiff::renderDiffToHTMLFromOpcode('c', $this->from_text, $in_offset, $n);
            }
            else if ( $edit instanceof FineDiffDeleteOp ) {
                FineDiff::renderDiffToHTMLFromOpcode('d', $this->from_text, $in_offset, $n);
            }
            else if ( $edit instanceof FineDiffInsertOp ) {
                FineDiff::renderDiffToHTMLFromOpcode('i', $edit->getText(), 0, $edit->getToLen());
            }
            else /* if ( $edit instanceof FineDiffReplaceOp ) */ {
                FineDiff::renderDiffToHTMLFromOpcode('d', $this->from_text, $in_offset, $n);
                FineDiff::renderDiffToHTMLFromOpcode('i', $edit->getText(), 0, $edit->getToLen());
            }
            $in_offset += $n;
        }
        $result = ob_get_clean();
        return ($result !== false) ? $result : '';
    }

    /**------------------------------------------------------------------------
     * Return an opcodes string describing the diff between a "From" and a
     * "To" string
     */
    public static function getDiffOpcodes($from, $to, $granularities = null) {
        return (new self($from, $to, $granularities))->getOpcodes();
    }

    /**------------------------------------------------------------------------
     * Return an iterable collection of diff ops from an opcodes string
     */
    public static function getDiffOpsFromOpcodes($opcodes) {
        $diffops = new FineDiffOps();
        self::renderFromOpcodes(null, $opcodes, array($diffops,'appendOpcode'));
        return $diffops->edits;
    }

    /**------------------------------------------------------------------------
     * Re-create the "To" string from the "From" string and an "Opcodes" string
     */
    public static function renderToTextFromOpcodes($from, $opcodes) {
        ob_start();
        self::renderFromOpcodes($from, $opcodes, array('FineDiff','renderToTextFromOpcode'));
        return ob_get_clean();
    }

    /**------------------------------------------------------------------------
     * Render the diff to an HTML string
     */
    public static function renderDiffToHTMLFromOpcodes($from, $opcodes) {
        ob_start();
        self::renderFromOpcodes($from, $opcodes, array('FineDiff','renderDiffToHTMLFromOpcode'));
        return ob_get_clean();
    }

    /**------------------------------------------------------------------------
     * Generic opcodes parser, user must supply callback for handling
     * single opcode
     */
    public static function renderFromOpcodes($from, $opcodes, $callback) {
        if ( !is_callable($callback) ) {
            return;
        }
        $opcodes_len = mb_strlen($opcodes);
        $from_offset = $opcodes_offset = 0;
        while ( $opcodes_offset <  $opcodes_len ) {
            $opcode = mb_substr($opcodes, $opcodes_offset, 1);
            $opcodes_offset++;
            $n = intval(mb_substr($opcodes, $opcodes_offset));
            if ( $n ) {
                $opcodes_offset += mb_strlen(strval($n));
            }
            else {
                $n = 1;
            }
            if ( $opcode === 'c' ) { // copy n characters from source
                call_user_func($callback, 'c', $from, $from_offset, $n, '');
                $from_offset += $n;
            }
            else if ( $opcode === 'd' ) { // delete n characters from source
                call_user_func($callback, 'd', $from, $from_offset, $n, '');
                $from_offset += $n;
            }
            else /* if ( $opcode === 'i' ) */ { // insert n characters from opcodes
                call_user_func($callback, 'i', $opcodes, $opcodes_offset + 1, $n);
                $opcodes_offset += 1 + $n;
            }
        }
    }

    /**------------------------------------------------------------------------
     *
     * Private section
     *
     */

    /**
     * Entry point to compute the diff.
     */
    private function doDiff($from_text, $to_text) {
        $this->last_edit = false;
        $this->stackpointer = 0;
        $this->from_text = $from_text;
        $this->from_offset = 0;
        // can't diff without at least one granularity specifier
        if ( empty($this->granularityStack) ) {
            return;
        }
        $this->_processGranularity($from_text, $to_text);
    }

    /**
     * This is the recursive function which is responsible for
     * handling/increasing granularity.
     *
     * Incrementally increasing the granularity is key to compute the
     * overall diff in a very efficient way.
     */
    private function _processGranularity($from_segment, $to_segment) {
        $delimiters = $this->granularityStack[$this->stackpointer++];
        $has_next_stage = $this->stackpointer < count($this->granularityStack);
        foreach ( FineDiff::doFragmentDiff($from_segment, $to_segment, $delimiters) as $fragment_edit ) {
            // increase granularity
            if ( $fragment_edit instanceof FineDiffReplaceOp && $has_next_stage ) {
                $this->_processGranularity(
                    mb_substr($this->from_text, $this->from_offset, $fragment_edit->getFromLen()),
                    $fragment_edit->getText()
                );
            }
            // fuse copy ops whenever possible
            else if ( $fragment_edit instanceof FineDiffCopyOp && $this->last_edit instanceof FineDiffCopyOp ) {
                $this->edits[count($this->edits)-1]->increase($fragment_edit->getFromLen());
                $this->from_offset += $fragment_edit->getFromLen();
            }
            else {
                /* $fragment_edit instanceof FineDiffCopyOp */
                /* $fragment_edit instanceof FineDiffDeleteOp */
                /* $fragment_edit instanceof FineDiffInsertOp */
                $this->edits[] = $this->last_edit = $fragment_edit;
                $this->from_offset += $fragment_edit->getFromLen();

                // increase insertion count
                if ( $fragment_edit instanceof fineDiffInsertOp ) {
                    $this->insertions_count++;
                }
                // increase deletion count
                else if ( $fragment_edit instanceof fineDiffDeleteOp ) {
                    $this->deletions_count++;
                }
                // increase insertion and deletion count
                else if ( $fragment_edit instanceof fineDiffReplaceOp ) {
                    $this->insertions_count++;
                    $this->deletions_count++;
                }
            }
        }
        $this->stackpointer--;
    }

    /**
     * This is the core algorithm which actually perform the diff itself,
     * fragmenting the strings as per specified delimiters.
     *
     * This function is naturally recursive, however for performance purpose
     * a local job queue is used instead of outright recursivity.
     */
    private static function doFragmentDiff($from_text, $to_text, $delimiters) {
        // Empty delimiter means character-level diffing.
        // In such case, use code path optimized for character-level
        // diffing.
        if ( empty($delimiters) ) {
            return FineDiff::doCharDiff($from_text, $to_text);
        }

        $result = array();

        // fragment-level diffing
        $from_text_len = mb_strlen($from_text);
        $to_text_len = mb_strlen($to_text);
        $from_fragments = FineDiff::extractFragments($from_text, $delimiters);
        $to_fragments = FineDiff::extractFragments($to_text, $delimiters);

        $jobs = array(array(0, $from_text_len, 0, $to_text_len));

        $cached_array_keys = array();

        while ( $job = array_pop($jobs) ) {

            // get the segments which must be diff'ed
            list($from_segment_start, $from_segment_end, $to_segment_start, $to_segment_end) = $job;

            // catch easy cases first
            $from_segment_length = $from_segment_end - $from_segment_start;
            $to_segment_length = $to_segment_end - $to_segment_start;
            if ( !$from_segment_length || !$to_segment_length ) {
                if ( $from_segment_length ) {
                    $result[$from_segment_start * 4] = new FineDiffDeleteOp($from_segment_length);
                }
                else if ( $to_segment_length ) {
                    $result[$from_segment_start * 4 + 1] = new FineDiffInsertOp(mb_substr($to_text, $to_segment_start, $to_segment_length));
                }
                continue;
            }

            // find longest copy operation for the current segments
            $best_copy_length = 0;

            $from_base_fragment_index = $from_segment_start;

            $cached_array_keys_for_current_segment = array();

            while ( $from_base_fragment_index < $from_segment_end ) {
                $from_base_fragment = $from_fragments[$from_base_fragment_index];
                $from_base_fragment_length = mb_strlen($from_base_fragment);
                // performance boost: cache array keys
                if ( !isset($cached_array_keys_for_current_segment[$from_base_fragment]) ) {
                    if ( !isset($cached_array_keys[$from_base_fragment]) ) {
                        $to_all_fragment_indices = $cached_array_keys[$from_base_fragment] = array_keys($to_fragments, $from_base_fragment, true);
                    }
                    else {
                        $to_all_fragment_indices = $cached_array_keys[$from_base_fragment];
                    }
                    // get only indices which falls within current segment
                    if ( $to_segment_start > 0 || $to_segment_end < $to_text_len ) {
                        $to_fragment_indices = array();
                        foreach ( $to_all_fragment_indices as $to_fragment_index ) {
                            if ( $to_fragment_index < $to_segment_start ) {
                                continue;
                            }
                            if ( $to_fragment_index >= $to_segment_end ) {
                                break;
                            }
                            $to_fragment_indices[] = $to_fragment_index;
                        }
                        $cached_array_keys_for_current_segment[$from_base_fragment] = $to_fragment_indices;
                    }
                    else {
                        $to_fragment_indices = $to_all_fragment_indices;
                    }
                }
                else {
                    $to_fragment_indices = $cached_array_keys_for_current_segment[$from_base_fragment];
                }
                // iterate through collected indices
                foreach ( $to_fragment_indices as $to_base_fragment_index ) {
                    $fragment_index_offset = $from_base_fragment_length;
                    // iterate until no more match
                    for (;;) {
                        $fragment_from_index = $from_base_fragment_index + $fragment_index_offset;
                        if ( $fragment_from_index >= $from_segment_end ) {
                            break;
                        }
                        $fragment_to_index = $to_base_fragment_index + $fragment_index_offset;
                        if ( $fragment_to_index >= $to_segment_end ) {
                            break;
                        }
                        if ( $from_fragments[$fragment_from_index] !== $to_fragments[$fragment_to_index] ) {
                            break;
                        }
                        $fragment_length = mb_strlen($from_fragments[$fragment_from_index]);
                        $fragment_index_offset += $fragment_length;
                    }
                    if ( $fragment_index_offset > $best_copy_length ) {
                        $best_copy_length = $fragment_index_offset;
                        $best_from_start = $from_base_fragment_index;
                        $best_to_start = $to_base_fragment_index;
                    }
                }
                $from_base_fragment_index += mb_strlen($from_base_fragment);
                // If match is larger than half segment size, no point trying to find better
                // TODO: Really?
                if ( $best_copy_length >= $from_segment_length / 2) {
                    break;
                }
                // no point to keep looking if what is left is less than
                // current best match
                if ( $from_base_fragment_index + $best_copy_length >= $from_segment_end ) {
                    break;
                }
            }

            if ( $best_copy_length ) {
                $jobs[] = array($from_segment_start, $best_from_start, $to_segment_start, $best_to_start);
                $result[$best_from_start * 4 + 2] = new FineDiffCopyOp($best_copy_length);
                $jobs[] = array($best_from_start + $best_copy_length, $from_segment_end, $best_to_start + $best_copy_length, $to_segment_end);
            }
            else {
                $result[$from_segment_start * 4 ] = new FineDiffReplaceOp($from_segment_length, mb_substr($to_text, $to_segment_start, $to_segment_length));
            }
        }

        ksort($result, SORT_NUMERIC);
        return array_values($result);
    }

    /**
     * Perform a character-level diff.
     *
     * The algorithm is quite similar to doFragmentDiff(), except that
     * the code path is optimized for character-level diff -- strpos() is
     * used to find out the longest common subequence of characters.
     *
     * We try to find a match using the longest possible subsequence, which
     * is at most the length of the shortest of the two strings, then incrementally
     * reduce the size until a match is found.
     *
     * I still need to study more the performance of this function. It
     * appears that for long strings, the generic doFragmentDiff() is more
     * performant. For word-sized strings, doCharDiff() is somewhat more
     * performant.
     */
    private static function doCharDiff($from_text, $to_text) {
        $result = array();
        $jobs = array(array(0, mb_strlen($from_text), 0, mb_strlen($to_text)));
        while ( $job = array_pop($jobs) ) {
            // get the segments which must be diff'ed
            list($from_segment_start, $from_segment_end, $to_segment_start, $to_segment_end) = $job;
            $from_segment_len = $from_segment_end - $from_segment_start;
            $to_segment_len = $to_segment_end - $to_segment_start;

            // catch easy cases first
            if ( !$from_segment_len || !$to_segment_len ) {
                if ( $from_segment_len ) {
                    $result[$from_segment_start * 4 + 0] = new FineDiffDeleteOp($from_segment_len);
                }
                else if ( $to_segment_len ) {
                    $result[$from_segment_start * 4 + 1] = new FineDiffInsertOp(mb_substr($to_text, $to_segment_start, $to_segment_len));
                }
                continue;
            }
            if ( $from_segment_len >= $to_segment_len ) {
                $copy_len = $to_segment_len;
                while ( $copy_len ) {
                    $to_copy_start = $to_segment_start;
                    $to_copy_start_max = $to_segment_end - $copy_len;
                    while ( $to_copy_start <= $to_copy_start_max ) {
                        $from_copy_start = mb_strpos(mb_substr($from_text, $from_segment_start, $from_segment_len), mb_substr($to_text, $to_copy_start, $copy_len));
                        if ( $from_copy_start !== false ) {
                            $from_copy_start += $from_segment_start;
                            break 2;
                        }
                        $to_copy_start++;
                    }
                    $copy_len--;
                }
            }
            else {
                $copy_len = $from_segment_len;
                while ( $copy_len ) {
                    $from_copy_start = $from_segment_start;
                    $from_copy_start_max = $from_segment_end - $copy_len;
                    while ( $from_copy_start <= $from_copy_start_max ) {
                        $to_copy_start = mb_strpos(mb_substr($to_text, $to_segment_start, $to_segment_len), mb_substr($from_text, $from_copy_start, $copy_len));
                        if ( $to_copy_start !== false ) {
                            $to_copy_start += $to_segment_start;
                            break 2;
                        }
                        $from_copy_start++;
                    }
                    $copy_len--;
                }
            }
            // match found
            if ( $copy_len ) {
                $jobs[] = array($from_segment_start, $from_copy_start, $to_segment_start, $to_copy_start);
                $result[$from_copy_start * 4 + 2] = new FineDiffCopyOp($copy_len);
                $jobs[] = array($from_copy_start + $copy_len, $from_segment_end, $to_copy_start + $copy_len, $to_segment_end);
            }
            // no match,  so delete all, insert all
            else {
                $result[$from_segment_start * 4] = new FineDiffReplaceOp($from_segment_len, mb_substr($to_text, $to_segment_start, $to_segment_len));
            }
        }
        ksort($result, SORT_NUMERIC);
        return array_values($result);
    }

    /**
     * Efficiently fragment the text into an array according to
     * specified delimiters.
     * No delimiters means fragment into single character.
     * The array indices are the offset of the fragments into
     * the input string.
     * A sentinel empty fragment is always added at the end.
     * Careful: No check is performed as to the validity of the
     * delimiters.
     */
    private static function extractFragments($text, $delimiters) {
        // special case: split into characters
        if ( empty($delimiters) ) {
            $chars = self::splitToChars($text, 1);
            $chars[] = '';
            return $chars;
        }
        $fragments = array();
        $start = $end = 0;
        for (;;) {
            $end += self::mb_strcspn($text, $delimiters, $end);
            $end += self::mb_strspn($text, $delimiters, $end);
            if ( $end === $start ) {
                break;
            }
            $fragments[$start] = mb_substr($text, $start, $end - $start);
            $start = $end;
        }
        $fragments[$start] = '';
        return $fragments;
    }

    /**
     * Stock opcode renderers
     */
    private static function renderToTextFromOpcode($opcode, $from, $from_offset, $from_len) {
        if ( $opcode === 'c' || $opcode === 'i' ) {
            echo mb_substr($from, $from_offset, $from_len);
        }
    }

    private static function renderDiffToHTMLFromOpcode($opcode, $from, $from_offset, $from_len) {
        if ( $opcode === 'c' ) {
            echo htmlspecialchars(mb_substr($from, $from_offset, $from_len), ENT_QUOTES);
        }
        else if ( $opcode === 'd' ) {
            $deletion = mb_substr($from, $from_offset, $from_len);
            if ( strcspn($deletion, " \n\r") === 0 ) {
                // no mb_ here is okay
                $deletion = str_replace(array("\n","\r"), array('\n','\r'), $deletion);
            }
            echo '<del>', htmlspecialchars($deletion), '</del>';
        }
        else /* if ( $opcode === 'i' ) */ {
            echo '<ins>', htmlspecialchars(mb_substr($from, $from_offset, $from_len), ENT_QUOTES), '</ins>';
        }
    }


    private static function splitToChars($str) {
        preg_match_all('/./us', $str, $matches);
        $matches = $matches[0];

        if (count($matches) === 0) return array('');
        return $matches;
    }

    private static function mb_strcspn($str, $delimiters, $start) {
        $dels = self::splitToChars($delimiters);
        $min  = mb_strlen($str);

        foreach ($dels as $del) {
            $pos = mb_strpos($str, $del, $start);
            if ($pos !== false && $pos < $min) $min = $pos;
        }

        return $min - $start;
    }

    private static function mb_strspn($str, $delimiters, $start) {
        $str  = mb_substr($str, $start);
        $dels = self::splitToChars($delimiters);

        foreach ($dels as $idx => $del) {
            // use of array_walk() here with preg_quote would likely be faster in the end
            $dels[$idx] = preg_quote($del, '/');
        }

        $dels = implode('|', $dels);

        preg_match("/^($dels)+/us", $str, $match);
        return $match ? mb_strlen($match[0]) : 0;
    }
}