A Formatter can now decide how (or whether) to order the attributes

   inside a tag. [bug=1812422]

3 files changed, 37 insertions, 3 deletions

diff --git a/CHANGELOG b/CHANGELOG
index b60b5b5..019ace4 100644
--- a/CHANGELOG
+++ b/CHANGELOG
@@ -6,6 +6,9 @@
    attributes are treated -- you can do this with the
    `multi_valued_attributes` argument. [bug=1832978]
 
+* A Formatter can now decide how (or whether) to order the attributes
+   inside a tag. [bug=1812422]
+
 * ' (which is valid in XML and XHTML, but not HTML 4) is now
    recognized as a named entity and converted to a single quote. [bug=1818721]
 
diff --git a/bs4/element.py b/bs4/element.py
index 1183f77..e8e48df 100644
--- a/bs4/element.py
+++ b/bs4/element.py
@@ -142,15 +142,20 @@ class Formatter(object):
     # By default, represent void elements as <tag/> rather than <tag>
     void_element_close_prefix = '/'
 
-    def substitute_entities(self, *args, **kwargs):
+    def substitute(self, *args, **kwargs):
         """Transform certain characters into named entities."""
         raise NotImplementedError()
+    
+    def sort_attributes(self, attributes):
+        """Reorder a tag's attributes however you want."""
+        return sorted(attributes.items())
+
 
 class HTMLFormatter(Formatter):
     """The default HTML formatter."""
     def substitute(self, *args, **kwargs):
         return HTMLAwareEntitySubstitution.substitute_html(*args, **kwargs)
-
+    
 class MinimalHTMLFormatter(Formatter):
     """A minimal HTML formatter."""
     def substitute(self, *args, **kwargs):
@@ -1157,7 +1162,11 @@ class Tag(PageElement):
             formatter = self._formatter_for_name(formatter)
         attrs = []
         if self.attrs:
-            for key, val in sorted(self.attrs.items()):
+            if isinstance(formatter, Formatter):
+                sorted_attrs = formatter.sort_attributes(self.attrs)
+            else:
+                sorted_attrs = self.attrs.items()
+            for key, val in sorted_attrs:
                 if val is None:
                     decoded = key
                 else:
diff --git a/bs4/tests/test_tree.py b/bs4/tests/test_tree.py
index a14928e..f7c5e2f 100644
--- a/bs4/tests/test_tree.py
+++ b/bs4/tests/test_tree.py
@@ -24,6 +24,7 @@ from bs4.element import (
     CData,
     Comment,
     Declaration,
+    MinimalHTMLFormatter,
     Doctype,
     NavigableString,
     SoupStrainer,
@@ -1683,6 +1684,27 @@ class TestEncoding(SoupTest):
         else:
             self.assertEqual(b'<b>\\u2603</b>', repr(soup))
 
+class TestFormatter(SoupTest):
+
+    def test_sort_attributes(self):
+        class UnsortedFormatter(MinimalHTMLFormatter):
+            def sort_attributes(self, attributes):
+                self.called_with = attributes
+                for k, v in sorted(attributes.items()):
+                    if k == 'ignore':
+                        continue
+                    yield k,v
+
+        soup = self.soup('<p cval="1" aval="2" ignore="ignored"></p>')
+        formatter = UnsortedFormatter()
+        decoded = soup.decode(formatter=formatter)
+
+        # sort_attributes() was called with all three attributes. It removed one and
+        # sorted the other two.
+        self.assertEquals(formatter.called_with, dict(cval="1", aval="2", ignore="ignored"))
+        self.assertEquals(u'<p aval="2" cval="1"></p>', decoded)
+
+
 class TestNavigableStringSubclasses(SoupTest):
 
     def test_cdata(self):