Add html element to Box class

Unless the Box is a PageBox or a MarginBox the new attribute `element` points
to the html element (xml.etree.ElementTree.Element) where the Box stems from.
This enables e.g. post-processing that depends on Element attributes, like
'class' or 'id'.

Caution: Elements with no subelements will test as False! Test the element's
existence with `if box.element is not None:`.

Related to #434, #970.

weasyprint/formatting_structure/boxes.py

@@ -89,8 +89,9 @@ class Box(object):
     def all_children(self):
         return ()
 
-    def __init__(self, element_tag, style):
+    def __init__(self, element_tag, style, element):
         self.element_tag = element_tag
+        self.element = element
         self.style = style
 
     def __repr__(self):
@@ -101,7 +102,7 @@ def anonymous_from(cls, parent, *args, **kwargs):
         """Return an anonymous box that inherits from ``parent``."""
         style = computed_from_cascaded(
             cascaded={}, parent_style=parent.style, element=None)
-        return cls(parent.element_tag, style, *args, **kwargs)
+        return cls(parent.element_tag, style, parent.element, *args, **kwargs)
 
     def copy(self):
         """Return shallow copy of the box."""
@@ -297,8 +298,8 @@ def page_values(self):
 
 class ParentBox(Box):
     """A box that has children."""
-    def __init__(self, element_tag, style, children):
-        super(ParentBox, self).__init__(element_tag, style)
+    def __init__(self, element_tag, style, element, children):
+        super(ParentBox, self).__init__(element_tag, style, element)
         self.children = tuple(children)
 
     def all_children(self):
@@ -482,9 +483,9 @@ class TextBox(InlineLevelBox):
     ascii_to_wide = dict((i, chr(i + 0xfee0)) for i in range(0x21, 0x7f))
     ascii_to_wide.update({0x20: '\u3000', 0x2D: '\u2212'})
 
-    def __init__(self, element_tag, style, text):
+    def __init__(self, element_tag, style, element, text):
         assert text
-        super(TextBox, self).__init__(element_tag, style)
+        super(TextBox, self).__init__(element_tag, style, element)
         text_transform = style['text_transform']
         if text_transform != 'none':
             text = {
@@ -532,8 +533,8 @@ class ReplacedBox(Box):
     and is opaque from CSS’s point of view.
 
     """
-    def __init__(self, element_tag, style, replacement):
-        super(ReplacedBox, self).__init__(element_tag, style)
+    def __init__(self, element_tag, style, element, replacement):
+        super(ReplacedBox, self).__init__(element_tag, style, element)
         self.replacement = replacement
 
 
@@ -683,7 +684,7 @@ def __init__(self, page_type, style):
         self.page_type = page_type
         # Page boxes are not linked to any element.
         super(PageBox, self).__init__(
-            element_tag=None, style=style, children=[])
+            element_tag=None, style=style, element=None, children=[])
 
     def __repr__(self):
         return '<%s %s>' % (type(self).__name__, self.page_type)
@@ -695,7 +696,7 @@ def __init__(self, at_keyword, style):
         self.at_keyword = at_keyword
         # Margin boxes are not linked to any element.
         super(MarginBox, self).__init__(
-            element_tag=None, style=style, children=[])
+            element_tag=None, style=style, element=None, children=[])
 
     def __repr__(self):
         return '<%s %s>' % (type(self).__name__, self.at_keyword)

weasyprint/formatting_structure/build.py

@@ -80,9 +80,10 @@ def root_style_for(element, pseudo_type=None):
     return box
 
 
-def make_box(element_tag, style, content):
-    return BOX_TYPE_FROM_DISPLAY[style['display']](
-        element_tag, style, content)
+def make_box(element_tag, style, content, element):
+    box = BOX_TYPE_FROM_DISPLAY[style['display']](
+        element_tag, style, element, content)
+    return box
 
 
 def element_to_box(element, style_for, get_image_from_uri, base_url,
@@ -122,7 +123,7 @@ def element_to_box(element, style_for, get_image_from_uri, base_url,
     if display == 'none':
         return []
 
-    box = make_box(element.tag, style, [])
+    box = make_box(element.tag, style, [], element)
 
     if state is None:
         # use a list to have a shared mutable object
@@ -227,7 +228,7 @@ def before_after_to_box(element, pseudo_type, state, style_for,
     if 'none' in (display, content) or content in ('normal', 'inhibit'):
         return []
 
-    box = make_box('%s::%s' % (element.tag, pseudo_type), style, [])
+    box = make_box('%s::%s' % (element.tag, pseudo_type), style, [], element)
 
     quote_depth, counter_values, _counter_scopes = state
     update_counters(state, style)
@@ -264,7 +265,7 @@ def marker_to_box(element, state, parent_style, style_for, get_image_from_uri,
     # `content` where 'normal' computes as 'inhibit' for pseudo elements.
     quote_depth, counter_values, _counter_scopes = state
 
-    box = make_box('%s::marker' % element.tag, style, children)
+    box = make_box('%s::marker' % element.tag, style, children, element)
 
     if style['display'] == 'none':
         return

weasyprint/html.py

@@ -111,7 +111,7 @@ def make_replaced_box(element, box, image):
     else:
         # TODO: support images with 'display: table-cell'?
         type_ = boxes.InlineReplacedBox
-    new_box = type_(element.tag, box.style, image)
+    new_box = type_(element.tag, box.style, element, image)
     # TODO: check other attributes that need to be copied
     # TODO: find another solution
     new_box.string_set = box.string_set