def read_page_properties(self, doc):
current = []
self.page_map = OrderedDict()
in_table = False
for p in descendants(doc, 'w:p', 'w:tbl'):
if p.tag.endswith('}tbl'):
in_table = True
self.tables.register(p)
continue
sect = tuple(descendants(p, 'w:sectPr'))
if sect:
pr = PageProperties(sect)
for x in current + [p]:
self.page_map[x] = pr
current = []
else:
current.append(p)
if in_table:
if ancestor(p, 'w:tbl') is not None:
self.tables.add(p)
else:
in_table = False
if current:
last = XPath('./w:body/w:sectPr')(doc)
pr = PageProperties(last)
for x in current:
self.page_map[x] = pr
def read_page_properties(self, doc):
current = []
self.page_map = OrderedDict()
in_table = False
for p in descendants(doc, "w:p", "w:tbl"):
if p.tag.endswith("}tbl"):
in_table = True
self.tables.register(p)
continue
sect = tuple(descendants(p, "w:sectPr"))
if sect:
pr = PageProperties(sect)
for x in current + [p]:
self.page_map[x] = pr
current = []
else:
current.append(p)
if in_table:
if ancestor(p, "w:tbl") is not None:
self.tables.add(p)
else:
in_table = False
if current:
last = XPath("./w:body/w:sectPr")(doc)
pr = PageProperties(last)
for x in current:
self.page_map[x] = pr
def from_toc(docx, link_map, styles, object_map):
toc_level = None
level = 0
TI = namedtuple('TI', 'text anchor indent')
toc = []
for tag in XPath('//*[(@w:fldCharType and name()="w:fldChar") or name()="w:hyperlink" or name()="w:instrText"]')(docx):
n = tag.tag.rpartition('}')[-1]
if n == 'fldChar':
t = get(tag, 'w:fldCharType')
if t == 'begin':
level += 1
elif t == 'end':
level -= 1
if toc_level is not None and level < toc_level:
break
elif n == 'instrText':
if level > 0 and tag.text and tag.text.strip().startswith('TOC '):
toc_level = level
elif n == 'hyperlink':
if toc_level is not None and level >= toc_level and tag in link_map:
a = link_map[tag]
href = a.get('href', None)
txt = link_to_txt(a, styles, object_map)
p = ancestor(tag, 'w:p')
if txt and href and p is not None:
ps = styles.resolve_paragraph(p)
try:
ml = int(ps.margin_left[:-2])
except (TypeError, ValueError, AttributeError):
ml = 0
if ps.text_align in {'center', 'right'}:
ml = 0
toc.append(TI(txt, href[1:], ml))
if toc:
return structure_toc(toc)
def from_toc(docx, link_map, styles, object_map, log):
toc_level = None
level = 0
TI = namedtuple('TI', 'text anchor indent')
toc = []
for tag in XPath(
'//*[(@w:fldCharType and name()="w:fldChar") or name()="w:hyperlink" or name()="w:instrText"]'
)(docx):
n = tag.tag.rpartition('}')[-1]
if n == 'fldChar':
t = get(tag, 'w:fldCharType')
if t == 'begin':
level += 1
elif t == 'end':
level -= 1
if toc_level is not None and level < toc_level:
break
elif n == 'instrText':
if level > 0 and tag.text and tag.text.strip().startswith('TOC '):
toc_level = level
elif n == 'hyperlink':
if toc_level is not None and level >= toc_level and tag in link_map:
a = link_map[tag]
href = a.get('href', None)
txt = link_to_txt(a, styles, object_map)
p = ancestor(tag, 'w:p')
if txt and href and p is not None:
ps = styles.resolve_paragraph(p)
try:
ml = int(ps.margin_left[:-2])
except (TypeError, ValueError, AttributeError):
ml = 0
if ps.text_align in {'center', 'right'}:
ml = 0
toc.append(TI(txt, href[1:], ml))
if toc:
log('Found Word Table of Contents, using it to generate the Table of Contents'
)
return structure_toc(toc)
def __call__(self):
doc = self.docx.document
relationships_by_id, relationships_by_type = self.docx.document_relationships
self.read_styles(relationships_by_type)
self.images(relationships_by_id)
self.layers = OrderedDict()
self.framed = [[]]
self.framed_map = {}
self.anchor_map = {}
self.link_map = defaultdict(list)
self.read_page_properties(doc)
for wp, page_properties in self.page_map.iteritems():
self.current_page = page_properties
p = self.convert_p(wp)
self.body.append(p)
notes_header = None
if self.footnotes.has_notes:
dl = DL()
dl.set("class", "notes")
self.body.append(H1(self.notes_text))
notes_header = self.body[-1]
notes_header.set("class", "notes-header")
self.body.append(dl)
for anchor, text, note in self.footnotes:
dl.append(DT("[", A("←" + text, href="#back_%s" % anchor, title=text), id=anchor))
dl[-1][0].tail = "]"
dl.append(DD())
in_table = False
for wp in note:
if wp.tag.endswith("}tbl"):
self.tables.register(wp)
in_table = True
continue
if in_table:
if ancestor(wp, "w:tbl") is not None:
self.tables.add(wp)
else:
in_table = False
p = self.convert_p(wp)
dl[-1].append(p)
self.resolve_links(relationships_by_id)
self.styles.cascade(self.layers)
self.tables.apply_markup(self.object_map)
numbered = []
for html_obj, obj in self.object_map.iteritems():
raw = obj.get("calibre_num_id", None)
if raw is not None:
lvl, num_id = raw.partition(":")[0::2]
try:
lvl = int(lvl)
except (TypeError, ValueError):
lvl = 0
numbered.append((html_obj, num_id, lvl))
self.numbering.apply_markup(numbered, self.body, self.styles, self.object_map)
self.apply_frames()
if len(self.body) > 0:
self.body.text = "\n\t"
for child in self.body:
child.tail = "\n\t"
self.body[-1].tail = "\n"
self.styles.generate_classes()
for html_obj, obj in self.object_map.iteritems():
style = self.styles.resolve(obj)
if style is not None:
css = style.css
if css:
cls = self.styles.class_name(css)
if cls:
html_obj.set("class", cls)
for html_obj, css in self.framed_map.iteritems():
cls = self.styles.class_name(css)
if cls:
html_obj.set("class", cls)
if notes_header is not None:
for h in self.body.iterchildren("h1", "h2", "h3"):
notes_header.tag = h.tag
cls = h.get("class", None)
if cls and cls != "notes-header":
notes_header.set("class", "%s notes-header" % cls)
break
return self.write()
def convert_p(self, p):
dest = P()
self.object_map[dest] = p
style = self.styles.resolve_paragraph(p)
self.layers[p] = []
self.add_frame(dest, style.frame)
current_anchor = None
current_hyperlink = None
for x in descendants(p, "w:r", "w:bookmarkStart", "w:hyperlink"):
if x.tag.endswith("}r"):
span = self.convert_run(x)
if current_anchor is not None:
(dest if len(dest) == 0 else span).set("id", current_anchor)
current_anchor = None
if current_hyperlink is not None:
hl = ancestor(x, "w:hyperlink")
if hl is not None:
self.link_map[hl].append(span)
else:
current_hyperlink = None
dest.append(span)
self.layers[p].append(x)
elif x.tag.endswith("}bookmarkStart"):
anchor = get(x, "w:name")
if anchor and anchor not in self.anchor_map:
self.anchor_map[anchor] = current_anchor = generate_anchor(
anchor, frozenset(self.anchor_map.itervalues())
)
elif x.tag.endswith("}hyperlink"):
current_hyperlink = x
m = re.match(r"heading\s+(\d+)$", style.style_name or "", re.IGNORECASE)
if m is not None:
n = min(6, max(1, int(m.group(1))))
dest.tag = "h%d" % n
if style.direction == "rtl":
dest.set("dir", "rtl")
border_runs = []
common_borders = []
for span in dest:
run = self.object_map[span]
style = self.styles.resolve_run(run)
if not border_runs or border_runs[-1][1].same_border(style):
border_runs.append((span, style))
elif border_runs:
if len(border_runs) > 1:
common_borders.append(border_runs)
border_runs = []
for border_run in common_borders:
spans = []
bs = {}
for span, style in border_run:
style.get_border_css(bs)
style.clear_border_css()
spans.append(span)
if bs:
cls = self.styles.register(bs, "text_border")
wrapper = self.wrap_elems(spans, SPAN())
wrapper.set("class", cls)
return dest
def __call__(self):
doc = self.docx.document
relationships_by_id, relationships_by_type = self.docx.document_relationships
self.read_styles(relationships_by_type)
self.images(relationships_by_id)
self.layers = OrderedDict()
self.framed = [[]]
self.framed_map = {}
self.anchor_map = {}
self.link_map = defaultdict(list)
self.read_page_properties(doc)
for wp, page_properties in self.page_map.iteritems():
self.current_page = page_properties
p = self.convert_p(wp)
self.body.append(p)
notes_header = None
if self.footnotes.has_notes:
dl = DL()
dl.set('class', 'notes')
self.body.append(H1(self.notes_text))
notes_header = self.body[-1]
notes_header.set('class', 'notes-header')
self.body.append(dl)
for anchor, text, note in self.footnotes:
dl.append(
DT('[',
A('←' + text, href='#back_%s' % anchor, title=text),
id=anchor))
dl[-1][0].tail = ']'
dl.append(DD())
in_table = False
for wp in note:
if wp.tag.endswith('}tbl'):
self.tables.register(wp)
in_table = True
continue
if in_table:
if ancestor(wp, 'w:tbl') is not None:
self.tables.add(wp)
else:
in_table = False
p = self.convert_p(wp)
dl[-1].append(p)
self.resolve_links(relationships_by_id)
self.styles.cascade(self.layers)
self.tables.apply_markup(self.object_map)
numbered = []
for html_obj, obj in self.object_map.iteritems():
raw = obj.get('calibre_num_id', None)
if raw is not None:
lvl, num_id = raw.partition(':')[0::2]
try:
lvl = int(lvl)
except (TypeError, ValueError):
lvl = 0
numbered.append((html_obj, num_id, lvl))
self.numbering.apply_markup(numbered, self.body, self.styles,
self.object_map)
self.apply_frames()
if len(self.body) > 0:
self.body.text = '\n\t'
for child in self.body:
child.tail = '\n\t'
self.body[-1].tail = '\n'
self.styles.generate_classes()
for html_obj, obj in self.object_map.iteritems():
style = self.styles.resolve(obj)
if style is not None:
css = style.css
if css:
cls = self.styles.class_name(css)
if cls:
html_obj.set('class', cls)
for html_obj, css in self.framed_map.iteritems():
cls = self.styles.class_name(css)
if cls:
html_obj.set('class', cls)
if notes_header is not None:
for h in self.body.iterchildren('h1', 'h2', 'h3'):
notes_header.tag = h.tag
cls = h.get('class', None)
if cls and cls != 'notes-header':
notes_header.set('class', '%s notes-header' % cls)
break
return self.write()
def convert_p(self, p):
dest = P()
self.object_map[dest] = p
style = self.styles.resolve_paragraph(p)
self.layers[p] = []
self.add_frame(dest, style.frame)
current_anchor = None
current_hyperlink = None
for x in descendants(p, 'w:r', 'w:bookmarkStart', 'w:hyperlink'):
if x.tag.endswith('}r'):
span = self.convert_run(x)
if current_anchor is not None:
(dest if len(dest) == 0 else span).set(
'id', current_anchor)
current_anchor = None
if current_hyperlink is not None:
hl = ancestor(x, 'w:hyperlink')
if hl is not None:
self.link_map[hl].append(span)
else:
current_hyperlink = None
dest.append(span)
self.layers[p].append(x)
elif x.tag.endswith('}bookmarkStart'):
anchor = get(x, 'w:name')
if anchor and anchor not in self.anchor_map:
self.anchor_map[anchor] = current_anchor = generate_anchor(
anchor, frozenset(self.anchor_map.itervalues()))
elif x.tag.endswith('}hyperlink'):
current_hyperlink = x
m = re.match(r'heading\s+(\d+)$', style.style_name or '',
re.IGNORECASE)
if m is not None:
n = min(6, max(1, int(m.group(1))))
dest.tag = 'h%d' % n
if style.direction == 'rtl':
dest.set('dir', 'rtl')
border_runs = []
common_borders = []
for span in dest:
run = self.object_map[span]
style = self.styles.resolve_run(run)
if not border_runs or border_runs[-1][1].same_border(style):
border_runs.append((span, style))
elif border_runs:
if len(border_runs) > 1:
common_borders.append(border_runs)
border_runs = []
for border_run in common_borders:
spans = []
bs = {}
for span, style in border_run:
style.get_border_css(bs)
style.clear_border_css()
spans.append(span)
if bs:
cls = self.styles.register(bs, 'text_border')
wrapper = self.wrap_elems(spans, SPAN())
wrapper.set('class', cls)
return dest
def cleanup_markup(log, root, styles, dest_dir, detect_cover):
# Move <hr>s outside paragraphs, if possible.
for hr in root.xpath('//span/hr'):
p = ancestor(hr, 'p')
descendants = tuple(p.iterdescendants())
if descendants[-1] is hr:
parent = p.getparent()
idx = parent.index(p)
parent.insert(idx+1, hr)
hr.tail = '\n\t'
# Merge consecutive spans that have the same styling
current_run = []
for span in root.xpath('//span'):
if not current_run:
current_run.append(span)
else:
last = current_run[-1]
if mergeable(last, span):
current_run.append(span)
else:
if len(current_run) > 1:
merge_run(current_run)
current_run = [span]
# Remove unnecessary span tags that are the only child of a parent block
# element
class_map = dict(styles.classes.itervalues())
parents = ('p', 'div') + tuple('h%d' % i for i in xrange(1, 7))
for parent in root.xpath('//*[(%s) and count(span)=1]' % ' or '.join('name()="%s"' % t for t in parents)):
if len(parent) == 1 and not parent.text and not parent[0].tail and not parent[0].get('id', None):
# We have a block whose contents are entirely enclosed in a <span>
span = parent[0]
span_class = span.get('class', None)
span_css = class_map.get(span_class, {})
if liftable(span_css):
pclass = parent.get('class', None)
if span_class:
pclass = (pclass + ' ' + span_class) if pclass else span_class
parent.set('class', pclass)
parent.text = span.text
parent.remove(span)
for child in span:
parent.append(child)
# Make spans whose only styling is bold or italic into <b> and <i> tags
for span in root.xpath('//span[@class]'):
css = class_map.get(span.get('class', None), {})
if len(css) == 1:
if css == {'font-style':'italic'}:
span.tag = 'i'
del span.attrib['class']
elif css == {'font-weight':'bold'}:
span.tag = 'b'
del span.attrib['class']
# Get rid of <span>s that have no styling
for span in root.xpath('//span[not(@class) and not(@id)]'):
lift(span)
if detect_cover:
# Check if the first image in the document is possibly a cover
img = root.xpath('//img[@src][1]')
if img:
img = img[0]
path = os.path.join(dest_dir, img.get('src'))
if os.path.exists(path) and before_count(root, img, limit=10) < 5:
from calibre.utils.magick.draw import identify
try:
width, height, fmt = identify(path)
except:
width, height, fmt = 0, 0, None
is_cover = 0.8 <= height/width <= 1.8 and height*width >= 160000
if is_cover:
log.debug('Detected an image that looks like a cover')
img.getparent().remove(img)
return path
