Optimize regular expression for identifying line breaks in comments.

sqlparse/filters/others.py

@@ -22,7 +22,10 @@ def get_next_comment():
         def _get_insert_token(token):
             """Returns either a whitespace or the line breaks from token."""
             # See issue484 why line breaks should be preserved.
-            m = re.search(r'((\r\n|\r|\n)+) *$', token.value)
+            # Note: The actual value for a line break is replaced by \n
+            # in SerializerUnicode which will be executed in the
+            # postprocessing state.
+            m = re.search(r'((\r|\n)+) *$', token.value)
             if m is not None:
                 return sql.Token(T.Whitespace.Newline, m.groups()[0])
             else:

tests/test_format.py

@@ -84,6 +84,23 @@ def test_strip_comments_multi(self):
         res = sqlparse.format(sql, strip_comments=True)
         assert res == 'select (select 2)'
 
+    def test_strip_comments_preserves_linebreak(self):
+        sql = 'select * -- a comment\r\nfrom foo'
+        res = sqlparse.format(sql, strip_comments=True)
+        assert res == 'select *\nfrom foo'
+        sql = 'select * -- a comment\nfrom foo'
+        res = sqlparse.format(sql, strip_comments=True)
+        assert res == 'select *\nfrom foo'
+        sql = 'select * -- a comment\rfrom foo'
+        res = sqlparse.format(sql, strip_comments=True)
+        assert res == 'select *\nfrom foo'
+        sql = 'select * -- a comment\r\n\r\nfrom foo'
+        res = sqlparse.format(sql, strip_comments=True)
+        assert res == 'select *\n\nfrom foo'
+        sql = 'select * -- a comment\n\nfrom foo'
+        res = sqlparse.format(sql, strip_comments=True)
+        assert res == 'select *\n\nfrom foo'
+
     def test_strip_ws(self):
         f = lambda sql: sqlparse.format(sql, strip_whitespace=True)
         s = 'select\n* from      foo\n\twhere  ( 1 = 2 )\n'