The plumbing for images in DOCX Output

src/calibre/ebooks/docx/writer/container.py

@@ -6,7 +6,7 @@ from __future__ import (unicode_literals, division, absolute_import,
 __license__ = 'GPL v3'
 __copyright__ = '2013, Kovid Goyal <kovid at kovidgoyal.net>'
 
-import textwrap
+import textwrap, os
 from io import BytesIO
 
 from lxml import etree
@@ -14,7 +14,7 @@ from lxml.builder import ElementMaker
 
 from calibre import guess_type
 from calibre.constants import numeric_version, __appname__
-from calibre.ebooks.docx.names import namespaces, STYLES, WEB_SETTINGS
+from calibre.ebooks.docx.names import namespaces, STYLES, WEB_SETTINGS, IMAGES
 from calibre.ebooks.metadata import authors_to_string
 from calibre.ebooks.metadata.opf2 import OPF as ReadOPF
 from calibre.ebooks.oeb.base import OPF, OPF2_NS
@@ -51,7 +51,6 @@ class DocumentRelationships(object):
 
     def __init__(self):
         self.rmap = {}
-        self.counter = 0
         for typ, target in {
                 STYLES: 'styles.xml',
                 WEB_SETTINGS: 'webSettings.xml',
@@ -64,11 +63,13 @@ class DocumentRelationships(object):
     def add_relationship(self, target, rtype, target_mode=None):
         ans = self.get_relationship_id(target, rtype, target_mode)
         if ans is None:
-            self.counter += 1
+            ans = 'rId%d' % (len(self.rmap) + 1)
-            ans = 'rId%d' % self.counter
             self.rmap[(target, rtype, target_mode)] = ans
         return ans
 
+    def add_image(self, target):
+        return self.add_relationship(target, IMAGES)
+
     def serialize(self):
         E = ElementMaker(namespace=namespaces['pr'], nsmap={None:namespaces['pr']})
         relationships = E.Relationships()
@@ -113,8 +114,13 @@ class DOCX(object):
         }.iteritems():
             added.add(ext)
             types.append(E.Default(Extension=ext, ContentType=mt))
-        # TODO: Iterate over all resources and add mimetypes for any that are
+        for fname in self.images:
-        # not already added
+            ext = fname.rpartition(os.extsep)[-1]
+            if ext not in added:
+                added.add(ext)
+                mt = guess_type('a.' + ext)[0]
+                if mt:
+                    types.append(E.Default(Extension=ext, ContentType=mt))
         return xml2str(types)
 
     @property
@@ -176,6 +182,8 @@ class DOCX(object):
             zf.writestr('word/document.xml', xml2str(self.document))
             zf.writestr('word/styles.xml', xml2str(self.styles))
             zf.writestr('word/_rels/document.xml.rels', self.document_relationships.serialize())
+            for fname, data_getter in self.images.iteritems():
+                zf.writestr(fname, data_getter())
 
 if __name__ == '__main__':
     d = DOCX(None, None)

src/calibre/ebooks/docx/writer/from_html.py

@@ -13,6 +13,7 @@ from lxml.builder import ElementMaker
 
 from calibre.ebooks.docx.names import namespaces
 from calibre.ebooks.docx.writer.styles import w, StylesManager
+from calibre.ebooks.docx.writer.images import ImagesManager
 from calibre.ebooks.oeb.stylizer import Stylizer as Sz, Style as St
 from calibre.ebooks.oeb.base import XPath, barename
 from calibre.ebooks.pdf.render.common import PAPER_SIZES
@@ -151,18 +152,26 @@ class Convert(object):
 
     def __call__(self):
         from calibre.ebooks.oeb.transforms.rasterize import SVGRasterizer
-        SVGRasterizer()(self.oeb, self.opts)
+        self.svg_rasterizer = SVGRasterizer()
+        self.svg_rasterizer(self.oeb, self.opts)
 
         self.styles_manager = StylesManager()
+        self.images_manager = ImagesManager(self.oeb, self.docx.document_relationships)
 
-        for item in self.oeb.spine:
+        try:
-            self.process_item(item)
+            for item in self.oeb.spine:
+                self.process_item(item)
 
-        self.styles_manager.finalize(self.blocks)
+            self.styles_manager.finalize(self.blocks)
-        self.write()
+            self.write()
+        finally:
+            self.images_manager.cleanup()
 
     def process_item(self, item):
-        stylizer = Stylizer(item.data, item.href, self.oeb, self.opts, self.opts.output_profile)
+        stylizer = self.svg_rasterizer.stylizer_cache.get(item)
+        if stylizer is None:
+            stylizer = Stylizer(item.data, item.href, self.oeb, self.opts, self.opts.output_profile)
+        self.abshref = self.images_manager.abshref = item.abshref
 
         is_first_block = True
         for body in XPath('//h:body')(item.data):
@@ -177,21 +186,24 @@ class Convert(object):
         block_style = stylizer.style(html_block)
         if block_style.is_hidden:
             return
-        if html_block.text:
+        if html_block.tag.endswith('}img'):
-            docx_block.add_text(html_block.text, block_style, ignore_leading_whitespace=True, is_parent_style=True)
+            b = Block(self.styles_manager, html_block, stylizer.style(html_block))
+            self.blocks.append(b)
+            self.images_manager.add_image(html_block, b, stylizer)
+        else:
+            if html_block.text:
+                docx_block.add_text(html_block.text, block_style, ignore_leading_whitespace=True, is_parent_style=True)
 
-        for child in html_block.iterchildren(etree.Element):
+            for child in html_block.iterchildren(etree.Element):
-            tag = barename(child.tag)
+                tag = barename(child.tag)
-            style = stylizer.style(child)
+                style = stylizer.style(child)
-            display = style._get('display')
+                display = style._get('display')
-            if tag == 'img':
+                if display == 'block' and tag != 'br':
-                pass  # TODO: Handle images
+                    b = Block(self.styles_manager, child, style)
-            if display == 'block' and tag != 'br':
+                    self.blocks.append(b)
-                b = Block(self.styles_manager, child, style)
+                    self.process_block(child, b, stylizer)
-                self.blocks.append(b)
+                else:
-                self.process_block(child, b, stylizer)
+                    self.process_inline(child, self.blocks[-1], stylizer)
-            else:
-                self.process_inline(child, self.blocks[-1], stylizer)
 
         if ignore_tail is False and html_block.tail and html_block.tail.strip():
             b = docx_block
@@ -211,7 +223,7 @@ class Convert(object):
             if html_child.tail or html_child is not html_child.getparent()[-1]:
                 docx_block.add_break(clear={'both':'all', 'left':'left', 'right':'right'}.get(style['clear'], 'none'))
         elif tag == 'img':
-            return  # TODO: Handle images
+            self.images_manager.add_image(html_child, docx_block, stylizer)
         else:
             if html_child.text:
                 docx_block.add_text(html_child.text, style, html_parent=html_child)
@@ -249,7 +261,7 @@ class Convert(object):
             E.docGrid(**{w('linePitch'):"360"}),
         ))
 
-        dn = {k:v for k, v in namespaces.iteritems() if k in 'wr'}
+        dn = {k:v for k, v in namespaces.iteritems() if k in tuple('wra') + ('wp',)}
         E = ElementMaker(namespace=dn['w'], nsmap=dn)
         self.docx.styles = E.styles(
             E.docDefaults(
@@ -268,4 +280,6 @@ class Convert(object):
                 )
             )
         )
+        self.docx.images = {}
         self.styles_manager.serialize(self.docx.styles)
+        self.images_manager.serialize(self.docx.images)

src/calibre/ebooks/docx/writer/images.py

@@ -0,0 +1,78 @@
+#!/usr/bin/env python2
+# vim:fileencoding=utf-8
+from __future__ import (unicode_literals, division, absolute_import,
+                        print_function)
+
+__license__ = 'GPL v3'
+__copyright__ = '2015, Kovid Goyal <kovid at kovidgoyal.net>'
+
+import os
+import shutil, posixpath
+from collections import namedtuple
+from functools import partial
+
+from calibre.ebooks.oeb.base import urlunquote
+from calibre.ptempfile import PersistentTemporaryDirectory
+from calibre.utils.filenames import ascii_filename
+from calibre.utils.magick.draw import identify_data
+
+Image = namedtuple('Image', 'rid fname width height fmt item')
+
+class ImagesManager(object):
+
+    def __init__(self, oeb, document_relationships):
+        self.oeb, self.log = oeb, oeb.log
+        self.images = {}
+        self.seen_filenames = set()
+        self.document_relationships = document_relationships
+        self._tdir = None
+
+    @property
+    def tdir(self):
+        if self._tdir is None:
+            self._tdir = PersistentTemporaryDirectory(suffix='_docx_output_images')
+        return self._tdir
+
+    def cleanup(self):
+        if self._tdir is not None:
+            shutil.rmtree(self._tdir)
+            self._tdir = None
+
+    def add_image(self, img, block, stylizer):
+        src = img.get('src')
+        if not src:
+            return
+        href = self.abshref(src)
+        if href not in self.images:
+            item = self.oeb.manifest.hrefs.get(href)
+            if item is None or not isinstance(item.data, bytes):
+                return
+            width, height, fmt = identify_data(item.data)
+            image_fname = 'media/' + self.create_filename(href, fmt)
+            image_rid = self.document_relationships.add_image(image_fname)
+            self.images[href] = Image(image_rid, image_fname, width, height, fmt, item)
+            item.unload_data_from_memory()
+        return self.images[href].rid
+
+    def create_filename(self, href, fmt):
+        fname = ascii_filename(urlunquote(posixpath.basename(href)))
+        fname = posixpath.splitext(fname)[0]
+        fname = fname[:75].rstrip('.') or 'image'
+        num = 0
+        base = fname
+        while fname.lower() in self.seen_filenames:
+            num += 1
+            fname = base + str(num)
+        self.seen_filenames.add(fname.lower())
+        fname += os.extsep + fmt.lower()
+        return fname
+
+    def serialize(self, images_map):
+        for img in self.images.itervalues():
+            images_map['word/' + img.fname] = partial(self.get_data, img.item)
+
+    def get_data(self, item):
+        try:
+            return item.data
+        finally:
+            item.unload_data_from_memory(False)