DOCX Input: Add support for framed text created by setting the same border style on neighboring paragraphs. Fixes #1526663 [DOCX Conversion- Border to a bunch of paras](https://bugs.launchpad.net/calibre/+bug/1526663)

src/calibre/ebooks/docx/block_styles.py

@@ -74,6 +74,7 @@ LINE_STYLES = {  # {{{
 # Read from XML {{{
 
 border_props = ('padding_%s', 'border_%s_width', 'border_%s_style', 'border_%s_color')
+border_edges = ('left', 'top', 'right', 'bottom', 'between')
 
 def read_single_border(parent, edge, XPath, get):
     color = style = width = padding = None
@@ -94,15 +95,12 @@ def read_single_border(parent, edge, XPath, get):
         if sz is not None:
             # we dont care about art borders (they are only used for page borders)
             try:
-                # WebKit needs at least 1pt to render borders
-                width = min(96, max(8, float(sz))) / 8
+                width = min(96, max(2, float(sz))) / 8
             except (ValueError, TypeError):
                 pass
-    if style == 'double' and width is not None and 0 < width < 3:
-        width = 3  # WebKit needs 3pts to render double borders
     return {p:v for p, v in zip(border_props, (padding, width, style, color))}
 
-def read_border(parent, dest, XPath, get, border_edges=('left', 'top', 'right', 'bottom'), name='pBdr'):
+def read_border(parent, dest, XPath, get, border_edges=border_edges, name='pBdr'):
     vals = {k % edge:inherit for edge in border_edges for k in border_props}
 
     for border in XPath('./w:' + name)(parent):
@@ -114,6 +112,22 @@ def read_border(parent, dest, XPath, get, border_edges=('left', 'top', 'right',
     for key, val in vals.iteritems():
         setattr(dest, key, val)
 
+def border_to_css(edge, style, css):
+    bs = getattr(style, 'border_%s_style' % edge)
+    bc = getattr(style, 'border_%s_color' % edge)
+    bw = getattr(style, 'border_%s_width' % edge)
+    if isinstance(bw, (float, int, long)):
+        # WebKit needs at least 1pt to render borders and 3pt to render double borders
+        bw = max(bw, (3 if bs == 'double' else 1))
+    if bs is not inherit and bs is not None:
+        css['border-%s-style' % edge] = bs
+    if bc is not inherit and bc is not None:
+        css['border-%s-color' % edge] = bc
+    if bw is not inherit and bw is not None:
+        if isinstance(bw, (int, float, long)):
+            bw = '%.3gpt' % bw
+        css['border-%s-width' % edge] = bw
+
 def read_indent(parent, dest, XPath, get):
     padding_left = padding_right = text_indent = inherit
     for indent in XPath('./w:ind')(parent):
@@ -304,6 +318,7 @@ class ParagraphStyle(object):
         'border_top_width', 'border_top_style', 'border_top_color', 'padding_top',
         'border_right_width', 'border_right_style', 'border_right_color', 'padding_right',
         'border_bottom_width', 'border_bottom_style', 'border_bottom_color', 'padding_bottom',
+        'border_between_width', 'border_between_style', 'border_between_color', 'padding_between',
         'margin_left', 'margin_top', 'margin_right', 'margin_bottom',
 
         # Misc.
@@ -336,6 +351,7 @@ class ParagraphStyle(object):
             self.font_family = self.font_size = self.color = inherit
 
         self._css = None
+        self._border_key = None
 
     def update(self, other):
         for prop in self.all_properties:
@@ -362,13 +378,7 @@ class ParagraphStyle(object):
             if self.keepNext is True:
                 c['page-break-after'] = 'avoid'
             for edge in ('left', 'top', 'right', 'bottom'):
-                val = getattr(self, 'border_%s_width' % edge)
-                if val is not inherit:
-                    c['border-left-width'] = '%.3gpt' % val
-                for x in ('style', 'color'):
-                    val = getattr(self, 'border_%s_%s' % (edge, x))
-                    if val is not inherit:
-                        c['border-%s-%s' % (edge, x)] = val
+                border_to_css(edge, self, c)
                 val = getattr(self, 'padding_%s' % edge)
                 if val is not inherit:
                     c['padding-%s' % edge] = '%.3gpt' % val
@@ -388,3 +398,40 @@ class ParagraphStyle(object):
 
         return self._css
 
+    @property
+    def border_key(self):
+        if self._border_key is None:
+            k = []
+            for edge in border_edges:
+                for prop in border_props:
+                    prop = prop % edge
+                    k.append(getattr(self, prop))
+            self._border_key = tuple(k)
+        return self._border_key
+
+    def has_identical_borders(self, other_style):
+        return self.border_key == getattr(other_style, 'border_key', None)
+
+    def clear_borders(self):
+        for edge in border_edges[:-1]:
+            for prop in ('width', 'color', 'style'):
+                setattr(self, 'border_%s_%s' % (edge, prop), inherit)
+
+    def clone_border_styles(self):
+        style = ParagraphStyle(self.namespace)
+        for edge in border_edges[:-1]:
+            for prop in ('width', 'color', 'style'):
+                attr = 'border_%s_%s' % (edge, prop)
+                setattr(style, attr, getattr(self, attr))
+        return style
+
+    def apply_between_border(self):
+        for prop in ('width', 'color', 'style'):
+            setattr(self, 'border_bottom_%s' % prop, getattr(self, 'border_between_%s' % prop))
+
+    def has_visible_border(self):
+        for edge in border_edges[:-1]:
+            bw, bs = getattr(self, 'border_%s_width' % edge), getattr(self, 'border_%s_style' % edge)
+            if bw is not inherit and bw and bs is not inherit and bs != 'none':
+                return True
+        return False

src/calibre/ebooks/docx/styles.py

@@ -439,7 +439,7 @@ class Styles(object):
             body { font-family: %s; font-size: %s; color: %s }
 
             /* In word all paragraphs have zero margins unless explicitly specified in a style */
-            p, h1, h2, h3, h4, h5, h6 { margin: 0; padding: 0 }
+            p, h1, h2, h3, h4, h5, h6, div { margin: 0; padding: 0 }
             /* In word headings only have bold font if explicitly specified */
             h1, h2, h3, h4, h5, h6 { font-weight: normal }
             /* Setting padding-left to zero breaks rendering of lists, so we only set the other values to zero and leave padding-left for the user-agent */

src/calibre/ebooks/docx/tables.py

@@ -8,7 +8,7 @@ __copyright__ = '2013, Kovid Goyal <kovid at kovidgoyal.net>'
 
 from lxml.html.builder import TABLE, TR, TD
 
-from calibre.ebooks.docx.block_styles import inherit, read_shd as rs, read_border, binary_property, border_props, ParagraphStyle
+from calibre.ebooks.docx.block_styles import inherit, read_shd as rs, read_border, binary_property, border_props, ParagraphStyle, border_to_css
 from calibre.ebooks.docx.char_styles import RunStyle
 
 # Read from XML {{{
@@ -177,14 +177,10 @@ class Style(object):
     def convert_border(self):
         c = {}
         for x in edges:
-            for prop in border_props:
-                prop = prop % x
-                if prop.startswith('border'):
-                    val = getattr(self, prop)
+            border_to_css(x, self, c)
+            val = getattr(self, 'padding_%s' % x)
             if val is not inherit:
-                        if isinstance(val, (int, float)):
-                            val = '%.3gpt' % val
-                        c[prop.replace('_', '-')] = val
+                c['padding-%s' % x] = '%.3gpt' % val
         return c
 
 class RowStyle(Style):

src/calibre/ebooks/docx/to_html.py

@@ -100,10 +100,12 @@ class Convert(object):
         self.images(relationships_by_id)
         self.layers = OrderedDict()
         self.framed = [[]]
+        self.frame_map = {}
         self.framed_map = {}
         self.anchor_map = {}
         self.link_map = defaultdict(list)
         self.link_source_map = {}
+        self.block_runs = []
         paras = []
 
         self.log.debug('Converting Word markup to HTML')
@@ -119,6 +121,7 @@ class Convert(object):
 
         self.read_block_anchors(doc)
         self.styles.apply_contextual_spacing(paras)
+        self.mark_block_runs(paras)
         # Apply page breaks at the start of every section, except the first
         # section (since that will be the start of the file)
         self.styles.apply_section_page_breaks(self.section_starts[1:])
@@ -147,6 +150,7 @@ class Convert(object):
                         dl[-1].append(p)
                         paras.append(wp)
                 self.styles.apply_contextual_spacing(paras)
+                self.mark_block_runs(paras)
 
         for p, wp in self.object_map.iteritems():
             if len(p) > 0 and not p.text and len(p[0]) > 0 and not p[0].text and p[0][0].get('class', None) == 'tab':
@@ -391,6 +395,7 @@ class Convert(object):
         self.object_map[dest] = p
         style = self.styles.resolve_paragraph(p)
         self.layers[p] = []
+        self.frame_map[p] = style.frame
         self.add_frame(dest, style.frame)
 
         current_anchor = None
@@ -680,6 +685,62 @@ class Convert(object):
             self.framed_map[frame] = css = style.css(self.page_map[self.object_map[paras[0]]])
             self.styles.register(css, 'frame')
 
+        if not self.block_runs:
+            return
+        rmap = {v:k for k, v in self.object_map.iteritems()}
+        for border_style, blocks in self.block_runs:
+            paras = tuple(rmap[p] for p in blocks)
+            parent = paras[0].getparent()
+            idx = parent.index(paras[0])
+            frame = DIV(*paras)
+            parent.insert(idx, frame)
+            self.framed_map[frame] = css = border_style.css
+            self.styles.register(css, 'frame')
+
+    def mark_block_runs(self, paras):
+
+        def process_run(run):
+            max_left = max_right = 0
+            has_visible_border = None
+            for p in run:
+                style = self.styles.resolve_paragraph(p)
+                if has_visible_border is None:
+                    has_visible_border = style.has_visible_border()
+                max_left, max_right = max(style.margin_left, max_left), max(style.margin_right, max_right)
+                if has_visible_border:
+                    style.margin_left = style.margin_right = inherit
+                if p is not run[0]:
+                    style.padding_top = 0
+                else:
+                    border_style = style.clone_border_styles()
+                    if has_visible_border:
+                        border_style.margin_top, style.margin_top = style.margin_top, inherit
+                if p is not run[-1]:
+                    style.padding_bottom = 0
+                else:
+                    if has_visible_border:
+                        border_style.margin_bottom, style.margin_bottom = style.margin_bottom, inherit
+                style.clear_borders()
+                if p is not run[-1]:
+                    style.apply_between_border()
+            if has_visible_border:
+                border_style.margin_left, border_style.margin_right = max_left,max_right
+                self.block_runs.append((border_style, run))
+
+        run = []
+        for p in paras:
+            if run and self.frame_map.get(p) == self.frame_map.get(run[-1]):
+                style = self.styles.resolve_paragraph(p)
+                last_style = self.styles.resolve_paragraph(run[-1])
+                if style.has_identical_borders(last_style):
+                    run.append(p)
+                    continue
+            if len(run) > 1:
+                process_run(run)
+            run = [p]
+        if len(run) > 1:
+            process_run(run)
+
 if __name__ == '__main__':
     import shutil
     from calibre.utils.logging import default_log