Merge lp:~ldolse/calibre/heuristics into lp:calibre

[Lee]

Discovered one other change in the European character patch that would cause false positive line unwrapping, updated the pattern to revert to the original behavior favoring of false negatives.

=== modified file 'src/calibre/ebooks/conversion/preprocess.py'
--- src/calibre/ebooks/conversion/preprocess.py	2012-04-13 15:23:43 +0000
+++ src/calibre/ebooks/conversion/preprocess.py	2012-04-20 17:04:32 +0000
@@ -559,7 +559,7 @@
                 end_rules.append((re.compile(u'(?<=.{%i}[–—])\s*<p>\s*(?=[[a-z\d])' % length), lambda match: ''))
                 end_rules.append(
                     # Un wrap using punctuation
-                    (re.compile(u'(?<=.{%i}([a-zäëïöüàèìòùáćéíĺóŕńśúýâêîôûçąężıãõñæøþðßěľščťžňďřů,:“”)\IA\u00DF]|(?<!\&\w{4});))\s*(?P<ital></(i|b|u)>)?\s*(</p>\s*<p>\s*)+\s*(?=(<(i|b|u)>)?\s*[\w\d$(])' % length, re.UNICODE), wrap_lines),
+                    (re.compile(u'(?<=.{%i}([a-zäëïöüàèìòùáćéíĺóŕńśúýâêîôûçąężıãõñæøþðßěľščťžňďřů,:)\IA\u00DF]|(?<!\&\w{4});))\s*(?P<ital></(i|b|u)>)?\s*(</p>\s*<p>\s*)+\s*(?=(<(i|b|u)>)?\s*[\w\d$(])' % length, re.UNICODE), wrap_lines),
                 )
 
         for rule in self.PREPROCESS + start_rules:

=== modified file 'src/calibre/ebooks/conversion/utils.py'
--- src/calibre/ebooks/conversion/utils.py	2012-04-20 13:52:57 +0000
+++ src/calibre/ebooks/conversion/utils.py	2012-04-20 17:04:32 +0000
@@ -316,10 +316,18 @@
         '''
         Unwraps lines based on line length and punctuation
         supports a range of html markup and text files
+        
+        the lookahead regex below is meant look for any non-full stop characters - punctuation
+        characters which can be used as a full stop should *not* be added below - e.g. ?!“”. etc
+        the reason for this is to prevent false positive wrapping.  False positives are more
+        difficult to detect than false negatives during a manual review of the doc
+        
+        This function intentionally leaves hyphenated content alone as that is handled by the 
+        dehyphenate routine in a separate step
         '''
+
         # define the pieces of the regex
-
-        lookahead = "(?<=.{"+str(length)+u"}([a-zäëïöüàèìòùáćéíĺóŕńśúýâêîôûçąężıãõñæøþðßěľščťžňďřů,:“”)\IA\u00DF]|(?<!\&\w{4});))" # (?<!\&\w{4});) is a semicolon not part of an entity
+        lookahead = "(?<=.{"+str(length)+u"}([a-zäëïöüàèìòùáćéíĺóŕńśúýâêîôûçąężıãõñæøþðßěľščťžňďřů,:)\IA\u00DF]|(?<!\&\w{4});))" # (?<!\&\w{4});) is a semicolon not part of an entity
         em_en_lookahead = "(?<=.{"+str(length)+u"}[\u2013\u2014])"
         soft_hyphen = u"\xad"
         line_ending = "\s*</(span|[iubp]|div)>\s*(</(span|[iubp]|div)>)?"