Don't copy styles when copying boxes, improve memory management

Style is not copied anymore when boxes are duplicated. Style dicts are not
modified anymore during the layout, as it was before for some properties:

- margins, borders and paddings when the box was split between two
  pages (useless as these computed values are stored directly in the box),
- top borders were changed in tables (useless for the same reason),
- bookmark labels and string sets are now stored in the box.

This commit can introduce very subtle bugs that are hard to debug. In the
future, we should try to freeze the style dicts before the layout.

Related to #70.

weasyprint/document.py

@@ -95,11 +95,11 @@ def _gather_links_and_bookmarks(box, bookmarks, links, anchors, matrix):
     if transform:
         matrix = transform * matrix if matrix else transform
 
-    bookmark_label = box.style.bookmark_label
-    if box.style.bookmark_level == 'none':
+    bookmark_label = box.bookmark_label
+    if box.style['bookmark_level'] == 'none':
         bookmark_level = None
     else:
-        bookmark_level = box.style.bookmark_level
+        bookmark_level = box.style['bookmark_level']
     link = box.style.link
     anchor_name = box.style.anchor
     has_bookmark = bookmark_label and bookmark_level

weasyprint/formatting_structure/boxes.py

@@ -62,7 +62,6 @@
 import itertools
 
 from ..compat import unichr, xrange
-from ..css.computed_values import ZERO_PIXELS
 
 # The *Box classes have many attributes and methods, but that's the way it is
 # pylint: disable=R0904,R0902
@@ -80,6 +79,8 @@ class Box(object):
     is_table_wrapper = False
     is_for_root_element = False
     transformation_matrix = None
+    bookmark_label = None
+    string_set = None
 
     # Default, overriden on some subclasses
     def all_children(self):
@@ -98,18 +99,15 @@ def anonymous_from(cls, parent, *args, **kwargs):
         return cls(
             parent.element_tag, parent.style.inherit_from(), *args, **kwargs)
 
-    def copy(self, copy_style=True):
+    def copy(self):
         """Return shallow copy of the box."""
         cls = type(self)
         # Create a new instance without calling __init__: initializing
         # styles may be kinda expensive, no need to do it again.
         new_box = cls.__new__(cls)
         # Copy attributes
         new_box.__dict__.update(self.__dict__)
-        if copy_style:
-            new_box.style = self.style.copy()
-        else:
-            new_box.style = self.style
+        new_box.style = self.style
         return new_box
 
     def translate(self, dx=0, dy=0):
@@ -299,10 +297,6 @@ def _reset_spacing(self, side):
         setattr(self, 'padding_%s' % side, 0)
         setattr(self, 'border_%s_width' % side, 0)
 
-        self.style['margin_%s' % side] = ZERO_PIXELS
-        self.style['padding_%s' % side] = ZERO_PIXELS
-        self.style['border_%s_width' % side] = 0
-
     def _remove_decoration(self, start, end):
         if start:
             self._reset_spacing('top')
@@ -456,7 +450,7 @@ def __init__(self, element_tag, style, text):
     def copy_with_text(self, text):
         """Return a new TextBox identical to this one except for the text."""
         assert text
-        new_box = self.copy(copy_style=False)
+        new_box = self.copy()
         new_box.text = text
         return new_box
 

weasyprint/formatting_structure/build.py

@@ -180,7 +180,7 @@ def element_to_box(element, style_for, get_image_from_uri, base_url,
 
     box.children = children
     box.first_letter_style = style_for(element, 'first-letter')
-    replace_content_lists(element, box, style, counter_values)
+    set_content_lists(element, box, style, counter_values)
 
     # Specific handling for the element. (eg. replaced element)
     return html.handle_element(element, box, get_image_from_uri, base_url)
@@ -289,25 +289,25 @@ def compute_content_list_string(element, box, counter_values, content_list):
     return string
 
 
-def replace_content_lists(element, box, style, counter_values):
-    """Replace the content-lists by strings.
+def set_content_lists(element, box, style, counter_values):
+    """Set the content-lists by strings.
 
     These content-lists are used in GCPM properties like ``string-set`` and
     ``bookmark-label``.
 
     """
     string_set = []
-    if style.string_set != 'none':
-        for i, (string_name, string_values) in enumerate(style.string_set):
+    if style['string_set'] != 'none':
+        for i, (string_name, string_values) in enumerate(style['string_set']):
             string_set.append((string_name, compute_content_list_string(
                 element, box, counter_values, string_values)))
-    box.style['string_set'] = string_set
+    box.string_set = string_set
 
-    if style.bookmark_label == 'none':
-        box.style['bookmark_label'] = ''
+    if style['bookmark_label'] == 'none':
+        box.bookmark_label = ''
     else:
-        box.style['bookmark_label'] = compute_content_list_string(
-            element, box, counter_values, style.bookmark_label)
+        box.bookmark_label = compute_content_list_string(
+            element, box, counter_values, style['bookmark_label'])
 
 
 def update_counters(state, style):

weasyprint/layout/pages.py

@@ -526,7 +526,7 @@ def make_page(context, root_box, page_type, resume_at, content_empty,
     page.children = [root_box]
     descendants = page.descendants()
     for child in descendants:
-        string_sets = child.style.string_set
+        string_sets = child.string_set
         if string_sets and string_sets != 'none':
             for string_set in string_sets:
                 string_name, text = string_set

weasyprint/layout/tables.py

@@ -13,7 +13,6 @@
 from __future__ import division, unicode_literals
 
 from ..compat import xrange
-from ..css.properties import Dimension
 from ..formatting_structure import boxes
 from ..logger import LOGGER
 from .percentages import resolve_one_percentage, resolve_percentages
@@ -58,7 +57,7 @@ def table_layout(context, table, max_position_y, skip_stack,
             skipped_rows = 0
         _, horizontal_borders = table.collapsed_border_grid
         if horizontal_borders:
-            table.style['border_top_width'] = table.border_top_width = max(
+            table.border_top_width = max(
                 width for _, (_, width, _)
                 in horizontal_borders[skipped_rows]) / 2
 
@@ -740,7 +739,6 @@ def table_wrapper_width(context, wrapper, containing_block):
         auto_table_layout(context, wrapper, containing_block)
 
     wrapper.width = table.border_width()
-    wrapper.style['width'] = Dimension(wrapper.width, 'px')
 
 
 def cell_baseline(cell):