Use the 'flexContext' option for nicer fuzzy anchoring

* Updated dom-text-matcher to c93abc20
* Updated Annotator to 41b88f9f (maintenance branch)

This version of Annotator uses a new method for adjusting the results of the two-phase fuzzy matching algorythm (used for anchoring). For details, see here: https://github.com/hypothesis/dom-text-matcher/commit/c93abc20bedd4450fd4eefe7d7d487aba324a088

Fixes #681.

h/lib/annotator.js

 /*
-** Annotator 1.2.6-dev-5fafbdf
+** Annotator 1.2.6-dev-f85315e
 ** https://github.com/okfn/annotator/
 **
 ** Copyright 2012 Aron Carroll, Rufus Pollock, and Nick Stenning.
 ** Dual licensed under the MIT and GPLv3 licenses.
 ** https://github.com/okfn/annotator/blob/master/LICENSE
 **
-** Built at: 2013-08-28 00:28:44Z
+** Built at: 2013-08-28 02:11:19Z
 */
 
 
@@ -1102,7 +1102,8 @@
       options = {
         contextMatchDistance: this.domMapper.getDocLength() * 2,
         contextMatchThreshold: 0.5,
-        patternMatchThreshold: 0.5
+        patternMatchThreshold: 0.5,
+        flexContext: true
       };
       result = this.domMatcher.searchFuzzyWithContext(prefix, suffix, quote, expectedStart, expectedEnd, false, null, options);
       if (!result.matches.length) {
@@ -1110,7 +1111,7 @@
         return null;
       }
       match = result.matches[0];
-      console.log("Fuzzy found match:");
+      console.log("2-phase fuzzy found match:");
       console.log(match);
       browserRange = new Range.BrowserRange(match.realRange);
       normalizedRange = browserRange.normalize(this.wrapper[0]);
@@ -1146,7 +1147,7 @@
         return null;
       }
       match = result.matches[0];
-      console.log("Fuzzy found match:");
+      console.log("1-phase fuzzy found match:");
       console.log(match);
       browserRange = new Range.BrowserRange(match.realRange);
       normalizedRange = browserRange.normalize(this.wrapper[0]);

h/lib/dom_text_matcher.coffee

@@ -146,7 +146,8 @@ class window.DomTextMatcher
     # If the prefix is not found, give up
     unless prefixResult.length then return matches: []
 
-    # This is where the prefix ends
+    # This is where the prefix was found
+    prefixStart = prefixResult[0].start
     prefixEnd = prefixResult[0].end
 
     # Let's find out where do we expect to find the suffix!
@@ -176,9 +177,11 @@ class window.DomTextMatcher
     # If the suffix is not found, give up
     unless suffixResult.length then return matches: []
 
-    # This is where the suffix starts
+    # This is where the suffix was found
     suffixStart = prefixEnd + suffixResult[0].start
+    suffixEnd = prefixEnd + suffixResult[0].end
 
+    # This if the range between the prefix and the suffix
     charRange =
       start: prefixEnd
       end: suffixStart
@@ -189,11 +192,58 @@ class window.DomTextMatcher
     # See how good a match we have
     analysis = @analyzeMatch pattern, charRange, true
 
+    # Should we try to find a better match by moving the
+    # initial match around a little bit, even if this has
+    # a negative impact on the similarity of the context?
+    if pattern? and options.flexContext and not analysis.exact
+      # Do we have and exact match for the quote around here?
+
+      if not @pm then @pm = new window.DTM_ExactMatcher
+      @pm.setDistinct false
+      @pm.setCaseSensitive false
+
+      flexMatches = @pm.search @mapper.corpus[prefixStart..suffixEnd], pattern
+      delete candidate
+      bestError = 2
+
+      for flexMatch in flexMatches
+
+        # Calculate the range that matched the quote
+        flexRange =
+          start: prefixStart + flexMatch.start
+          end: prefixStart + flexMatch.end
+
+        # Check how the prefix would fare
+        prefixRange = start: prefixStart, end: flexRange.start
+        a1 = @analyzeMatch prefix, prefixRange, true
+        prefixError = if a1.exact then 0 else a1.comparison.errorLevel
+
+        # Check how the suffix would fare
+        suffixRange = start: flexRange.end, end: suffixEnd
+        a2 = @analyzeMatch suffix, suffixRange, true
+        suffixError = if a2.exact then 0 else a2.comparison.errorLevel
+
+        # Did we at least one match?
+        if a1.exact or a2.exact
+          # Yes, we did. Calculate the total error
+          totalError = prefixError + suffixError
+
+          # Is this better than our best bet?
+          if totalError < bestError
+            # This is our best candidate so far. Store it.
+            candidate = flexRange
+            bestError = totalError
+
+      if candidate?
+        console.log "flexContext adjustment: we found a better candidate!"
+        charRange = candidate
+        analysis = @analyzeMatch pattern, charRange, true
+
     # Do we have to compare what we found to a pattern?
     if (not pattern?) or # "No pattern, nothing to compare. Assume it's OK."
         analysis.exact or # "Found text matches exactly to pattern"
         (analysis.comparison.errorLevel <= matchThreshold) # still acceptable
-      mappings = @mapper.getMappingsForCharRange prefixEnd, suffixStart
+      mappings = @mapper.getMappingsForCharRange charRange.start, charRange.end
 
       # Collect the results
       match = {}