If you have troubles, maybe the comments from dzuchara can help:
#15 (comment)



Diff, Match and Patch Library (PHP Version)

  The Diff Match and Patch library offer robust algorithms to perform the operations required for synchronizing
  plain text.

  PHP port written 2010 by Tobias Buschor <shwups.ch>

  Diff, Match and Patch Library

  Written by Neil Fraser
  Copyright 2006 Google Inc.
  <http://code.google.com/p/google-diff-match-patch/>


  API   : <http://code.google.com/p/google-diff-match-patch/wiki/API>

Introduction

  This library is available in multiple languages. Regardless of the language used, the interface for using it is the
  same. This page describes the API for the public functions. For further examples, see the relevant test harness.

Initialization

  The first step is to create a new diff_match_patch object. This object contains various properties which set the
  behaviour of the algorithms, as well as the following methods/functions:

diff_main(text1, text2) => diffs

  An array of differences is computed which describe the transformation of text1 into text2. Each difference is an
  array (JavaScript, Lua) or tuple (Python) or Diff object (C++, C#, Java). The first element specifies if it is an
  insertion (1), a deletion (-1) or an equality (0). The second element specifies the affected text.

  diff_main("Good dog", "Bad dog") => [(-1, "Goo"), (1, "Ba"), (0, "d dog")]

  Despite the large number of optimisations used in this function, diff can take a while to compute. The
  diff_match_patch.Diff_Timeout property is available to set how many seconds any diff's exploration phase may take.
  The default value is 1.0. A value of 0 disables the timeout and lets diff run until completion. Should diff timeout,
  the return value will still be a valid difference, though probably non-optimal.

diff_cleanupSemantic(diffs) => null

  A diff of two unrelated texts can be filled with coincidental matches. For example, the diff of "mouse" and "sofas"
  is [(-1, "m"), (1, "s"), (0, "o"), (-1, "u"), (1, "fa"), (0, "s"), (-1, "e")]. While this is the optimum diff, it is
  difficult for humans to understand. Semantic cleanup rewrites the diff, expanding it into a more intelligible format.
  The above example would become: [(-1, "mouse"), (1, "sofas")]. If a diff is to be human-readable, it should be passed
  to diff_cleanupSemantic.

diff_cleanupEfficiency(diffs) => null

  This function is similar to diff_cleanupSemantic, except that instead of optimising a diff to be human-readable, it
  optimises the diff to be efficient for machine processing. The results of both cleanup types are often the same.

  The efficiency cleanup is based on the observation that a diff made up of large numbers of small diffs edits may take
  longer to process (in downstream applications) or take more capacity to store or transmit than a smaller number of
  larger diffs. The diff_match_patch.Diff_EditCost property sets what the cost of handling a new edit is in terms of
  handling extra characters in an existing edit. The default value is 4, which means if expanding the length of a diff
  by three characters can eliminate one edit, then that optimisation will reduce the total costs.

diff_levenshtein(diffs) => int

  Given a diff, measure its Levenshtein distance in terms of the number of inserted, deleted or substituted characters.
  The minimum distance is 0 which means equality, the maximum distance is the length of the longer string.

diff_prettyHtml(diffs) => html

  Takes a diff array and returns a pretty HTML sequence. This function is mainly intended as an example from which to
  write ones own display functions.

match_main(text, pattern, loc) => location

  Given a text to search, a pattern to search for and an expected location in the text near which to find the pattern,
  return the location which matches closest. The function will search for the best match based on both the number of
  character errors between the pattern and the potential match, as well as the distance between the expected location
  and the potential match.

  The following example is a classic dilemma. There are two potential matches, one is close to the expected location but
  contains a one character error, the other is far from the expected location but is exactly the pattern sought after:
  match_main("abc12345678901234567890abbc", "abc", 26) Which result is returned (0 or 24) is determined by the
  diff_match_patch.Match_Distance property. An exact letter match which is 'distance' characters away from the fuzzy
  location would score as a complete mismatch. For example, a distance of '0' requires the match be at the exact
  location specified, whereas a threshold of '1000' would require a perfect match to be within 800 characters of the
  expected location to be found using a 0.8 threshold (see below). The larger Match_Distance is, the
  slower match_main() may take to compute. This variable defaults to 1000.

  Another property is diff_match_patch.Match_Threshold which determines the cut-off value for a valid match. If
  Match_Threshold is closer to 0, the requirements for accuracy increase. If Match_Threshold is closer to 1 then it is
  more likely that a match will be found. The larger Match_Threshold is, the slower match_main() may take to compute.
  This variable defaults to 0.5. If no match is found, the function returns -1.

patch_make(text1, text2) => patches

patch_make(diffs) => patches

patch_make(text1, diffs) => patches

  Given two texts, or an already computed list of differences, return an array of patch objects. The third form
  (text1, diffs) is preferred, use it if you happen to have that data available, otherwise this function will compute
  the missing pieces.

patch_toText(patches) => text

  Reduces an array of patch objects to a block of text which looks extremely similar to the standard GNU diff/patch
  format. This text may be stored or transmitted.

patch_fromText(text) => patches

  Parses a block of text (which was presumably created by the patch_toText function) and returns an array of patch
  objects.

patch_apply(patches, text1) => results

  Applies a list of patches to text1. The first element of the return value is the newly patched text. The second
  element is an array of true/false values indicating which of the patches were successfully applied. [Note that this
  second element is not too useful since large patches may get broken up internally, resulting in a longer results list
  than the input with no way to figure out which patch succeeded or failed. A more informative API is in development.]

  The previously mentioned Match_Distance and Match_Threshold properties are used to evaluate patch application on text
  which does not match exactly. In addition, the diff_match_patch.Patch_DeleteThreshold property determines how closely
  the text within a major (~64 character) delete needs to match the expected text. If Patch_DeleteThreshold is closer to
  0, then the deleted text must match the expected text more closely. If Patch_DeleteThreshold is closer to 1, then the
  deleted text may contain anything. In most use cases Patch_DeleteThreshold should just be set to the same value as
  Match_Threshold.