def handle_p_content(self, e, current_part):
if e.tag == RUN_TAG:
return self.handle_run(e)
elif e.tag == HYPERLINK_TAG:
internalId = e.attrib.get(ns.r('id'))
if internalId is None:
ref = '#' + e.attrib[ns.w('anchor')]
else:
rels = self.doc.get_rels_for(current_part)
ref = rels[internalId].attrib['Target']
# 'u', 'span' = nuke bogus color and underline
# styling that google docs likes to add to links;
# XXX(alexander): rewrite colour less bluntly;
# this also nukes background color
handle_p = partial(self.handle_p_content, current_part=current_part)
body = whack(('u', 'span').__contains__, flatmap(handle_p, e))
if not body:
log.warn('hyperlink with no body to: %r', ref)
return [mkel('a', {'href': ref}, body)]
elif e.tag == BOOKMARK_END_TAG:
return []
elif e.tag == BOOKMARK_START_TAG:
return [mkel('a', {'name': e.attrib[ns.w('name')]}, [])]
elif e.tag == ns.m('oMath'):
return self.handle_omath(e)
else:
log.warn('Ignoring unknown tag %s', e.tag)
return []
def handle_p_content(self, e, current_part):
if e.tag == RUN_TAG:
return self.handle_run(e)
elif e.tag == HYPERLINK_TAG:
internalId = e.attrib.get(ns.r('id'))
if internalId is None:
ref = '#' + e.attrib[ns.w('anchor')]
else:
rels = self.doc.get_rels_for(current_part)
ref = rels[internalId].attrib['Target']
# 'u', 'span' = nuke bogus color and underline
# styling that google docs likes to add to links;
# XXX(alexander): rewrite colour less bluntly;
# this also nukes background color
handle_p = partial(self.handle_p_content,
current_part=current_part)
body = whack(('u', 'span').__contains__, flatmap(handle_p, e))
if not body:
log.warn('hyperlink with no body to: %r', ref)
return [mkel('a', {'href': ref}, body)]
elif e.tag == BOOKMARK_END_TAG:
return []
elif e.tag == BOOKMARK_START_TAG:
return [mkel('a', {'name': e.attrib[ns.w('name')]}, [])]
elif e.tag == ns.m('oMath'):
return self.handle_omath(e)
else:
log.warn('Ignoring unknown tag %s', e.tag)
return []
def get_num_style(self, numid, level):
numid_xpath = self.A_NUMID_XPATH_TEMPL % numid
abstract_num_id = self.numbering.e.find(numid_xpath).attrib[ns.w('val')]
lvl_xpath = self.LVL_XPATH_TEMPL % (abstract_num_id, level)
lvl, = self.numbering.e.iterfind(lvl_xpath)
numFmt = val(lvl, ns.w('numFmt'))
lvlText = val(lvl, ns.w('lvlText'))
return NumStyle(numFmt=numFmt, lvlText=lvlText)
def val(e, child_tag, attrib=ns.w('val')):
if e is None:
return None
for child in e.iter(child_tag):
return child.attrib.get(attrib)
return None
def make_footnote(self, e):
# pylint: disable=W0622
id = e.attrib[ns.w('id')]
ps = (self.doc.get_footnote if e.tag == FOOTNOTE_REFERENCE_TAG else
self.doc.get_endnote)(id).iterfind(P_TAG)
footnote_part = 'footnotes' # XXX what about endnotes
return mkel('.footnote', {},
[self.handle_p(p, current_part=footnote_part) for p in ps])
def handle_p(self, e, current_part, in_list=False):
attrs = {}
pPr = first_of_tag(e, P_PROPS_TAG)
jc_class = self.JC_TO_CLASS.get(val(pPr, ns.w('jc')))
if jc_class:
attrs = add_class(attrs, jc_class)
tag = style_to_tag(val(pPr, ns.w('pStyle')) or '')
content = iter(e) if pPr is None else pPr.itersiblings()
handle_p = partial(self.handle_p_content, current_part=current_part)
ans = mkel(tag, attrs, flatmap(handle_p, content))
left_indent = val(pPr, ns.w('ind'), ns.w('left')) or 0.0
indent = int(round(float(left_indent) / self.default_indent_twips))
if (not in_list) and indent:
ans = lift_code(ans)
ans = mkel('.block', {'indent': indent}, [ans])
ans = hacky_flatten_block(ans)
return ans
def handle_p(self, e, current_part, in_list=False):
attrs = {}
pPr = first_of_tag(e, P_PROPS_TAG)
jc_class = self.JC_TO_CLASS.get(val(pPr, ns.w('jc')))
if jc_class:
attrs = add_class(attrs, jc_class)
tag = style_to_tag(val(pPr, ns.w('pStyle')) or '')
content = iter(e) if pPr is None else pPr.itersiblings()
handle_p = partial(self.handle_p_content, current_part=current_part)
ans = mkel(tag, attrs, flatmap(handle_p, content))
left_indent = val(pPr, ns.w('ind'), ns.w('left')) or 0.0
indent = int(round(float(left_indent) / self.default_indent_twips))
if (not in_list) and indent:
ans = lift_code(ans)
ans = mkel('.block', {'indent': indent}, [ans])
ans = hacky_flatten_block(ans)
return ans
def make_footnote(self, e):
# pylint: disable=W0622
id = e.attrib[ns.w('id')]
ps = (self.doc.get_footnote if e.tag == FOOTNOTE_REFERENCE_TAG
else self.doc.get_endnote)(id).iterfind(P_TAG)
footnote_part = 'footnotes' # XXX what about endnotes
return mkel('.footnote', {},
[self.handle_p(p, current_part=footnote_part) for p in ps])
def process(self, e, handle_p):
numPr = e.find(P_PROPS_TAG + '/' + ns.w('numPr'))
numid = val(numPr, ns.w('numId'))
if not numid:
return self.flush() + [handle_p(e, in_list=False)]
self.in_list = True # pylint: disable=W0201
level = int(val(numPr, ns.w('ilvl')) or 0)
while level != self.level:
if level > self.level:
self.append_points[-1].append([])
self.append_points.append(self.append_points[-1][-1])
else:
self.append_points.pop()
self.append_points[self.level].append(
(self.list_type(numid, level), handle_p(e, in_list=True)))
return []
def process(self, e, handle_p):
numPr = e.find(P_PROPS_TAG + '/' + ns.w('numPr'))
numid = val(numPr, ns.w('numId'))
if not numid:
return self.flush() + [handle_p(e, in_list=False)]
self.in_list = True # pylint: disable=W0201
level = int(val(numPr, ns.w('ilvl')) or 0)
while level != self.level:
if level > self.level:
self.append_points[-1].append([])
self.append_points.append(self.append_points[-1][-1])
else:
self.append_points.pop()
self.append_points[self.level].append(
(self.list_type(numid, level),
handle_p(e, in_list=True)))
return []
def parse_sectPr(e): # pylint: disable=C0103
assert e.tag == ns.w('sectPr')
d = dict(
page_width=map(ns.w, ['pgSz', 'w']),
left_margin=map(ns.w, ['pgMar', 'left']),
right_margin=map(ns.w, ['pgMar', 'right']),
)
return SectPr(**{k: Twips(val(e, *p)) for (k, p) in d.items()})
def handle_run(self, r):
# XXX(ash): pylint is right about this being too complex
# pylint: disable=R0912
_ = Var('_')
ans = []
rPr = first_of_tag(r, RUN_PROPS_TAG)
content = rPr.itersiblings() if rPr is not None else iter(r)
for e in content:
# pylint: disable=W0622
type = e.attrib.get(ns.w('type'))
if e.tag == TEXT_TAG:
ans.append(e.text)
elif e.tag == TAB_TAG:
# XXX(alexander): this can also work like a '_' or '…' \dotfill
ans.append('\t')
elif e.tag in (FOOTNOTE_REF_TAG, ENDNOTE_REF_TAG):
# XXX(ash): what is going on here
pass
elif e.tag == BREAK_TAG and type in ('page', 'column'):
ans.append(mkel('.pagebreak', {}, []))
elif e.tag == BREAK_TAG or e.tag == CR_TAG:
assert (type is None) or (type == 'textWrapping')
ans.append(mkel('br', {}, []))
# FIXME, tags below untested
elif e.tag == SOFT_HYPHEN_TAG:
ans.append(SOFT_HYPHEN)
elif e.tag == NON_BREAKING_HYPHEN_TAG:
ans.append(NON_BREAKING_HYPHEN)
elif e.tag == ns.w('drawing'):
ans.extend(
flatmap(self.transclude, e.xpath(self.IMAGE_XPATH,
namespaces=ns.dict)))
elif e.tag in (FOOTNOTE_REFERENCE_TAG, ENDNOTE_REFERENCE_TAG):
ans.append(self.make_footnote(e))
else:
# movie,
# rt, ruby, rubyAlign etc. for ruby stuff
# sym, with special handling for wingdings I guess...
log.warn('Unknown tag %r', e.tag)
if rPr is not None and ans != Seq[Seq['.footnote', _:], _:]:
ans = self.apply_rpr(rPr, ans)
return ans
def handle_run(self, r):
# XXX(ash): pylint is right about this being too complex
# pylint: disable=R0912
_ = Var('_')
ans = []
rPr = first_of_tag(r, RUN_PROPS_TAG)
content = rPr.itersiblings() if rPr is not None else iter(r)
for e in content:
# pylint: disable=W0622
type = e.attrib.get(ns.w('type'))
if e.tag == TEXT_TAG:
ans.append(e.text)
elif e.tag == TAB_TAG:
# XXX(alexander): this can also work like a '_' or '…' \dotfill
ans.append('\t')
elif e.tag in (FOOTNOTE_REF_TAG, ENDNOTE_REF_TAG):
# XXX(ash): what is going on here
pass
elif e.tag == BREAK_TAG and type in ('page', 'column'):
ans.append(mkel('.pagebreak', {}, []))
elif e.tag == BREAK_TAG or e.tag == CR_TAG:
assert (type is None) or (type == 'textWrapping')
ans.append(mkel('br', {}, []))
# FIXME, tags below untested
elif e.tag == SOFT_HYPHEN_TAG:
ans.append(SOFT_HYPHEN)
elif e.tag == NON_BREAKING_HYPHEN_TAG:
ans.append(NON_BREAKING_HYPHEN)
elif e.tag == ns.w('drawing'):
ans.extend(
flatmap(self.transclude,
e.xpath(self.IMAGE_XPATH, namespaces=ns.dict)))
elif e.tag in (FOOTNOTE_REFERENCE_TAG, ENDNOTE_REFERENCE_TAG):
ans.append(self.make_footnote(e))
else:
# movie,
# rt, ruby, rubyAlign etc. for ruby stuff
# sym, with special handling for wingdings I guess...
log.warn('Unknown tag %r', e.tag)
if rPr is not None and ans != Seq[Seq['.footnote', _:], _:]:
ans = self.apply_rpr(rPr, ans)
return ans
def parse_table(self, e, current_part):
# XXX(ash): simplify
# pylint: disable=R0914
def cell_bg(tc):
if tc[0].tag == TABLE_COLUMN_PROPERTIES_TAG:
bg = val(tc[0], ns.w('shd'), ns.w('fill'))
if bg:
return add_bg({}, '#' + bg)
return {}
def skip_past(e, child):
if e[0].tag == child:
return e[0].itersiblings()
return e.iterchildren()
def parse_rows(e, has_header_row, has_header_col):
def is_header(i, j):
return i == 0 and has_header_row or j == 0 and has_header_col
return [
mkel('tr', {}, [
mkel(
'th' if is_header(i, j) else 'td', cell_bg(tc),
self.parse_body(skip_past(tc,
TABLE_COLUMN_PROPERTIES_TAG),
current_part=current_part))
for (j, tc) in enumerate(tr.iterfind(TABLE_COLUMN_TAG))
]) for (i, tr) in enumerate(e.iterfind(TABLE_ROW_TAG))
]
tblPr = first_of_tag(e, ns.w('tblPr'))
tbl_stuff = tblPr.itersiblings()
tblGrid = next(tbl_stuff)
# according to the schema this is always true
assert tblGrid.tag == ns.w('tblGrid'), tblGrid.tag
look = tblPr.find(ns.w('tblLook'))
if look is None:
has_header_row = has_header_col = False
else:
# this is actually the canonical check;
# the identical per cell/row props are just for caching
has_header_row, has_header_col = (look.attrib.get(k) == "1"
for k in (ns.w('firstRow'),
ns.w('firstColumn')))
grid_cols = tblGrid.iterchildren(ns.w('gridCol'))
col_widths = [int(gc.attrib[ns.w('w')]) for gc in grid_cols]
col_total = sum(col_widths)
col_pcts = [100. * w / col_total for w in col_widths]
cols = [
mkel('col', add_style({}, 'width', '%s%%' % w), [])
for w in col_pcts
]
rows = parse_rows(e, has_header_row, has_header_col)
table = odt_parser.parse_table_body(cols + rows)
return mkel('table', {}, table)
def parse_table(self, e, current_part):
# XXX(ash): simplify
# pylint: disable=R0914
def cell_bg(tc):
if tc[0].tag == TABLE_COLUMN_PROPERTIES_TAG:
bg = val(tc[0], ns.w('shd'), ns.w('fill'))
if bg:
return add_bg({}, '#' + bg)
return {}
def skip_past(e, child):
if e[0].tag == child:
return e[0].itersiblings()
return e.iterchildren()
def parse_rows(e, has_header_row, has_header_col):
def is_header(i, j):
return i == 0 and has_header_row or j == 0 and has_header_col
return [
mkel('tr', {},
[mkel('th' if is_header(i, j) else 'td', cell_bg(tc),
self.parse_body(
skip_past(tc, TABLE_COLUMN_PROPERTIES_TAG),
current_part=current_part))
for (j, tc) in enumerate(tr.iterfind(TABLE_COLUMN_TAG))])
for (i, tr) in enumerate(e.iterfind(TABLE_ROW_TAG))]
tblPr = first_of_tag(e, ns.w('tblPr'))
tbl_stuff = tblPr.itersiblings()
tblGrid = next(tbl_stuff)
# according to the schema this is always true
assert tblGrid.tag == ns.w('tblGrid'), tblGrid.tag
look = tblPr.find(ns.w('tblLook'))
if look is None:
has_header_row = has_header_col = False
else:
# this is actually the canonical check;
# the identical per cell/row props are just for caching
has_header_row, has_header_col = (
look.attrib.get(k) == "1"
for k in (ns.w('firstRow'), ns.w('firstColumn')))
grid_cols = tblGrid.iterchildren(ns.w('gridCol'))
col_widths = [int(gc.attrib[ns.w('w')]) for gc in grid_cols]
col_total = sum(col_widths)
col_pcts = [100. * w / col_total for w in col_widths]
cols = [mkel('col',
add_style({}, 'width', '%s%%' % w),
[]) for w in col_pcts]
rows = parse_rows(e, has_header_row, has_header_col)
table = odt_parser.parse_table_body(cols + rows)
return mkel('table', {}, table)
def __init__(self, infilename, make_transclusions):
self.doc = doc = docxlite.Document(infilename)
# pylint: disable=W0212
self.numbering = doc.numbering
self.body = doc.document.e.find(ns.w('body'))
self.rels = doc.document.rels
sprops = docxlite.parse_sectPr(self.body[-1])
self.textwidth_emu = (sprops.page_width.emu.real
- sprops.right_margin.emu.real
- sprops.left_margin.emu.real)
if make_transclusions:
self.transclusions = make_transclusions(self.doc.get_images())
else:
self.transclusions = None
self.default_indent_twips = 720
def __init__(self, infilename, make_transclusions):
self.doc = doc = docxlite.Document(infilename)
# pylint: disable=W0212
self.numbering = doc.numbering
self.body = doc.document.e.find(ns.w('body'))
self.rels = doc.document.rels
sprops = docxlite.parse_sectPr(self.body[-1])
self.textwidth_emu = (sprops.page_width.emu.real -
sprops.right_margin.emu.real -
sprops.left_margin.emu.real)
if make_transclusions:
self.transclusions = make_transclusions(self.doc.get_images())
else:
self.transclusions = None
self.default_indent_twips = 720
def apply_rpr(self, rPr, ans):
stys = {x.tag for x in rPr.iterchildren(*self.STYLE_TO_HTML)}
if stys:
for (t, html) in self.STYLE_TO_HTML.iteritems():
if t in stys:
ans = [mkel(html, {}, ans)]
color = val(rPr, ns.w('color'))
if color:
a = add_style({}, 'color', '#' + color)
ans = [mkel('span', a, ans)] # FIXME word colors
# `None` here == turn highlighting off; it's different from no value
highlight = self.HIGHLIGHT_TO_RGB.get(val(rPr, ns.w('highlight')),
False)
if highlight is False: # higher precedence than shade
highlight = val(rPr, ns.w('shd'), ns.w('fill'))
if highlight:
ans = [mkel('span', add_bg({}, '#' + highlight), ans)]
vertalign = val(rPr, ns.w('vertAlign'))
if vertalign and vertalign != 'baseline':
ans = [mkel(vertalign[:3], {}, ans)]
if is_code_font(val(rPr, ns.w('rFonts'), ns.w('ascii'))):
ans = [mkel('code', {}, ans)]
return ans
def apply_rpr(self, rPr, ans):
stys = {x.tag for x in rPr.iterchildren(*self.STYLE_TO_HTML)}
if stys:
for (t, html) in self.STYLE_TO_HTML.iteritems():
if t in stys:
ans = [mkel(html, {}, ans)]
color = val(rPr, ns.w('color'))
if color:
a = add_style({}, 'color', '#' + color)
ans = [mkel('span', a, ans)] # FIXME word colors
# `None` here == turn highlighting off; it's different from no value
highlight = self.HIGHLIGHT_TO_RGB.get(val(rPr, ns.w('highlight')),
False)
if highlight is False: # higher precedence than shade
highlight = val(rPr, ns.w('shd'), ns.w('fill'))
if highlight:
ans = [mkel('span', add_bg({}, '#' + highlight), ans)]
vertalign = val(rPr, ns.w('vertAlign'))
if vertalign and vertalign != 'baseline':
ans = [mkel(vertalign[:3], {}, ans)]
if is_code_font(val(rPr, ns.w('rFonts'), ns.w('ascii'))):
ans = [mkel('code', {}, ans)]
return ans
class Document(object):
LVL_XPATH_TEMPL = ('./w:abstractNum[@w:abstractNumId="%s"]/'
'w:lvl[@w:ilvl="%s"]'.replace('w:', ns.w('')))
A_NUMID_XPATH_TEMPL = ('./w:num[@w:numId="%s"]/w:abstractNumId'
.replace('w:', ns.w('')))
def __init__(self, path_or_file):
self.z = read_zip(path_or_file)
self.document = get_part(self.z, MAGIC_WORD + '/document.xml')
self.numbering = get_part(self.z, MAGIC_WORD + '/numbering.xml')
self.footnotes = get_part(self.z, MAGIC_WORD + '/footnotes.xml')
self.endnotes = get_part(self.z, MAGIC_WORD + '/endnotes.xml')
@staticmethod
def _get_by_id(id, xs): # pylint: disable=W0622
return next(x for x in xs if id == x.attrib[ns.w('id')])
def get_footnote(self, id): # pylint: disable=W0622
return self._get_by_id(id, self.footnotes.e)
def get_endnote(self, id): # pylint: disable=W0622
return self._get_by_id(id, self.endnotes.e)
def get_num_style(self, numid, level):
numid_xpath = self.A_NUMID_XPATH_TEMPL % numid
abstract_num_id = self.numbering.e.find(numid_xpath).attrib[ns.w('val')]
lvl_xpath = self.LVL_XPATH_TEMPL % (abstract_num_id, level)
lvl, = self.numbering.e.iterfind(lvl_xpath)
numFmt = val(lvl, ns.w('numFmt'))
lvlText = val(lvl, ns.w('lvlText'))
return NumStyle(numFmt=numFmt, lvlText=lvlText)
def get_or_add_extn(self, mime_type):
path = '[Content_Types].xml'
e = to_etree(self.z.get(path))
ctns = 'http://schemas.openxmlformats.org/package/2006/content-types'
default = e.xpath('./ct:Default[@ContentType="%s"]/@Extension' % mime_type, # pylint: disable=C0301
namespaces={'ct': ctns})
if len(default) == 1:
return default[0]
extn = guess_extension(mime_type)
default = e.xpath('./ct:Default[@Extension="%s"]' % extn,
namespaces={'ct': ctns})
assert len(default) == 0 # XXX(ash): handle this case more gracefully
tup = ('Default', {'ContentType': mime_type, 'Extension': extn}, [])
e[:0] = [tup2etree(tup, {None: ctns})]
self.z[path] = etree2s(e)
return extn
def add_image(self, f, mime_type):
extn = self.get_or_add_extn(mime_type)
img = f.read()
p = fresh_name(set(self.z.keys()),
MAGIC_WORD + '/media/image%d.' + extn)
self.z[p] = img
# XXX(ash): why do we have to relativize the targets
rel_p = p.split('/', 1)[1]
rid = fresh_name(set(self.document.rels), 'rId%d')
self.document.rels[rid] = tup2etree(
('Relationship', {'Target': rel_p,
'Type': IMAGE_REL_URI,
'Id': rid}, []))
return rid
def get_images(self):
# XXX(ash): what about the rels of the other parts...
images = dict((id, r.attrib['Target'])
for (id, r) in self.document.rels.iteritems()
if r.attrib['Type'] == IMAGE_REL_URI)
includes = dict((id, StringIO(self.z[MAGIC_WORD + '/' + fn]))
for id, fn in images.items())
return includes
def get_rels_for(self, part):
return getattr(self, part).rels
def save(self, f):
replacements = {}
for name in ['document', 'numbering', 'footnotes', 'endnotes']:
part = getattr(self, name)
if part.e is not None:
replacements[part.path] = etree2s(part.e, decl=True)
if part.rels:
replacements[part.rels_path] = rels2s(part.rels)
with ZipFile(f, 'w') as outz:
for filename, contents in self.z.items():
replacement = replacements.get(filename, None)
outz.writestr(
filename, replacement if replacement else contents)
def _get_by_id(id, xs): # pylint: disable=W0622
return next(x for x in xs if id == x.attrib[ns.w('id')])
def cell_bg(tc):
if tc[0].tag == TABLE_COLUMN_PROPERTIES_TAG:
bg = val(tc[0], ns.w('shd'), ns.w('fill'))
if bg:
return add_bg({}, '#' + bg)
return {}
def cell_bg(tc):
if tc[0].tag == TABLE_COLUMN_PROPERTIES_TAG:
bg = val(tc[0], ns.w('shd'), ns.w('fill'))
if bg:
return add_bg({}, '#' + bg)
return {}
class Docx(object):
# get "normal" inline images (i.e. ignnores VML and similar crap)
IMAGE_XPATH = (
'./*[self::wp:inline|self::wp:anchor]'
'[.//a:graphicData'
'[@uri="http://schemas.openxmlformats.org/drawingml/2006/picture"]]')
JC_TO_CLASS = {
'left': 'left',
'right': 'right',
'center': 'center',
'both': 'justify',
}
STYLE_TO_HTML = OrderedDict(
(ns.w(x), x[0]) for x in ['u', 'b', 'bCs', 'i', 'iCs', 'strike'])
HIGHLIGHT_TO_RGB = {
'black': '000000',
'blue': '0000ff',
'cyan': '00ffff',
'darkBlue': '000080',
'darkCyan': '008080',
'darkGray': '808080',
'darkGreen': '008000',
'darkMagenta': '800080',
'darkRed': '800000',
'darkYellow': '808000',
'green': '00ff00',
'lightGray': 'c0c0c0',
'magenta': 'ff00ff',
'red': 'ff0000',
'white': 'ffffff',
'yellow': 'ffff00',
'none': None,
}
def __init__(self, infilename, make_transclusions):
self.doc = doc = docxlite.Document(infilename)
# pylint: disable=W0212
self.numbering = doc.numbering
self.body = doc.document.e.find(ns.w('body'))
self.rels = doc.document.rels
sprops = docxlite.parse_sectPr(self.body[-1])
self.textwidth_emu = (sprops.page_width.emu.real -
sprops.right_margin.emu.real -
sprops.left_margin.emu.real)
if make_transclusions:
self.transclusions = make_transclusions(self.doc.get_images())
else:
self.transclusions = None
self.default_indent_twips = 720
def parse(self):
return (self.parse_body(self.body,
current_part='document'), self.transclusions)
def handle_omath(self, e): # pylint: disable=W0613
return []
def handle_p_content(self, e, current_part):
if e.tag == RUN_TAG:
return self.handle_run(e)
elif e.tag == HYPERLINK_TAG:
internalId = e.attrib.get(ns.r('id'))
if internalId is None:
ref = '#' + e.attrib[ns.w('anchor')]
else:
rels = self.doc.get_rels_for(current_part)
ref = rels[internalId].attrib['Target']
# 'u', 'span' = nuke bogus color and underline
# styling that google docs likes to add to links;
# XXX(alexander): rewrite colour less bluntly;
# this also nukes background color
handle_p = partial(self.handle_p_content,
current_part=current_part)
body = whack(('u', 'span').__contains__, flatmap(handle_p, e))
if not body:
log.warn('hyperlink with no body to: %r', ref)
return [mkel('a', {'href': ref}, body)]
elif e.tag == BOOKMARK_END_TAG:
return []
elif e.tag == BOOKMARK_START_TAG:
return [mkel('a', {'name': e.attrib[ns.w('name')]}, [])]
elif e.tag == ns.m('oMath'):
return self.handle_omath(e)
else:
log.warn('Ignoring unknown tag %s', e.tag)
return []
def transclude(self, pic):
# for id:
# pylint: disable=W0622
if self.transclusions is None:
return []
width_emu = float(val(pic, ns.wp('extent'), 'cx'))
embeds = pic.xpath('.//a:blip/@r:embed', namespaces=ns.dict)
try:
id, = embeds
except ValueError:
log.warn('Expected exactly one r:embed with an image id, got %r',
embeds)
return []
href = self.transclusions.normalize_known_transclusion(id)
return [
make_figure(relwidth=width_emu / self.textwidth_emu,
inline={
'anchor': False,
'inline': True
}[pic.tag.split('}')[1]],
body=[mkel('img', {'src': href}, [])],
src=href,
original_href=id)
]
def make_footnote(self, e):
# pylint: disable=W0622
id = e.attrib[ns.w('id')]
ps = (self.doc.get_footnote if e.tag == FOOTNOTE_REFERENCE_TAG else
self.doc.get_endnote)(id).iterfind(P_TAG)
footnote_part = 'footnotes' # XXX what about endnotes
return mkel('.footnote', {},
[self.handle_p(p, current_part=footnote_part) for p in ps])
def handle_run(self, r):
# XXX(ash): pylint is right about this being too complex
# pylint: disable=R0912
_ = Var('_')
ans = []
rPr = first_of_tag(r, RUN_PROPS_TAG)
content = rPr.itersiblings() if rPr is not None else iter(r)
for e in content:
# pylint: disable=W0622
type = e.attrib.get(ns.w('type'))
if e.tag == TEXT_TAG:
ans.append(e.text)
elif e.tag == TAB_TAG:
# XXX(alexander): this can also work like a '_' or '…' \dotfill
ans.append('\t')
elif e.tag in (FOOTNOTE_REF_TAG, ENDNOTE_REF_TAG):
# XXX(ash): what is going on here
pass
elif e.tag == BREAK_TAG and type in ('page', 'column'):
ans.append(mkel('.pagebreak', {}, []))
elif e.tag == BREAK_TAG or e.tag == CR_TAG:
assert (type is None) or (type == 'textWrapping')
ans.append(mkel('br', {}, []))
# FIXME, tags below untested
elif e.tag == SOFT_HYPHEN_TAG:
ans.append(SOFT_HYPHEN)
elif e.tag == NON_BREAKING_HYPHEN_TAG:
ans.append(NON_BREAKING_HYPHEN)
elif e.tag == ns.w('drawing'):
ans.extend(
flatmap(self.transclude,
e.xpath(self.IMAGE_XPATH, namespaces=ns.dict)))
elif e.tag in (FOOTNOTE_REFERENCE_TAG, ENDNOTE_REFERENCE_TAG):
ans.append(self.make_footnote(e))
else:
# movie,
# rt, ruby, rubyAlign etc. for ruby stuff
# sym, with special handling for wingdings I guess...
log.warn('Unknown tag %r', e.tag)
if rPr is not None and ans != Seq[Seq['.footnote', _:], _:]:
ans = self.apply_rpr(rPr, ans)
return ans
def apply_rpr(self, rPr, ans):
stys = {x.tag for x in rPr.iterchildren(*self.STYLE_TO_HTML)}
if stys:
for (t, html) in self.STYLE_TO_HTML.iteritems():
if t in stys:
ans = [mkel(html, {}, ans)]
color = val(rPr, ns.w('color'))
if color:
a = add_style({}, 'color', '#' + color)
ans = [mkel('span', a, ans)] # FIXME word colors
# `None` here == turn highlighting off; it's different from no value
highlight = self.HIGHLIGHT_TO_RGB.get(val(rPr, ns.w('highlight')),
False)
if highlight is False: # higher precedence than shade
highlight = val(rPr, ns.w('shd'), ns.w('fill'))
if highlight:
ans = [mkel('span', add_bg({}, '#' + highlight), ans)]
vertalign = val(rPr, ns.w('vertAlign'))
if vertalign and vertalign != 'baseline':
ans = [mkel(vertalign[:3], {}, ans)]
if is_code_font(val(rPr, ns.w('rFonts'), ns.w('ascii'))):
ans = [mkel('code', {}, ans)]
return ans
def handle_p(self, e, current_part, in_list=False):
attrs = {}
pPr = first_of_tag(e, P_PROPS_TAG)
jc_class = self.JC_TO_CLASS.get(val(pPr, ns.w('jc')))
if jc_class:
attrs = add_class(attrs, jc_class)
tag = style_to_tag(val(pPr, ns.w('pStyle')) or '')
content = iter(e) if pPr is None else pPr.itersiblings()
handle_p = partial(self.handle_p_content, current_part=current_part)
ans = mkel(tag, attrs, flatmap(handle_p, content))
left_indent = val(pPr, ns.w('ind'), ns.w('left')) or 0.0
indent = int(round(float(left_indent) / self.default_indent_twips))
if (not in_list) and indent:
ans = lift_code(ans)
ans = mkel('.block', {'indent': indent}, [ans])
ans = hacky_flatten_block(ans)
return ans
def parse_body(self, xml, current_part):
builder = ListBuilder(self.doc)
body = []
for e in xml:
if e.tag == P_TAG:
handle_p = partial(self.handle_p, current_part=current_part)
body.extend(builder.process(e, handle_p))
else:
body.extend(builder.flush())
if e.tag == TABLE_TAG:
body.append(self.parse_table(e, current_part))
elif e.tag == SECTION_PROPERTIES_TAG:
pass
else:
log.warn('Unrecognized element: %s', e.tag)
body.extend(builder.flush())
return body
def parse_table(self, e, current_part):
# XXX(ash): simplify
# pylint: disable=R0914
def cell_bg(tc):
if tc[0].tag == TABLE_COLUMN_PROPERTIES_TAG:
bg = val(tc[0], ns.w('shd'), ns.w('fill'))
if bg:
return add_bg({}, '#' + bg)
return {}
def skip_past(e, child):
if e[0].tag == child:
return e[0].itersiblings()
return e.iterchildren()
def parse_rows(e, has_header_row, has_header_col):
def is_header(i, j):
return i == 0 and has_header_row or j == 0 and has_header_col
return [
mkel('tr', {}, [
mkel(
'th' if is_header(i, j) else 'td', cell_bg(tc),
self.parse_body(skip_past(tc,
TABLE_COLUMN_PROPERTIES_TAG),
current_part=current_part))
for (j, tc) in enumerate(tr.iterfind(TABLE_COLUMN_TAG))
]) for (i, tr) in enumerate(e.iterfind(TABLE_ROW_TAG))
]
tblPr = first_of_tag(e, ns.w('tblPr'))
tbl_stuff = tblPr.itersiblings()
tblGrid = next(tbl_stuff)
# according to the schema this is always true
assert tblGrid.tag == ns.w('tblGrid'), tblGrid.tag
look = tblPr.find(ns.w('tblLook'))
if look is None:
has_header_row = has_header_col = False
else:
# this is actually the canonical check;
# the identical per cell/row props are just for caching
has_header_row, has_header_col = (look.attrib.get(k) == "1"
for k in (ns.w('firstRow'),
ns.w('firstColumn')))
grid_cols = tblGrid.iterchildren(ns.w('gridCol'))
col_widths = [int(gc.attrib[ns.w('w')]) for gc in grid_cols]
col_total = sum(col_widths)
col_pcts = [100. * w / col_total for w in col_widths]
cols = [
mkel('col', add_style({}, 'width', '%s%%' % w), [])
for w in col_pcts
]
rows = parse_rows(e, has_header_row, has_header_col)
table = odt_parser.parse_table_body(cols + rows)
return mkel('table', {}, table)
def strip_meta(self, unaugmented_meta, transclusions, asides):
# XXX(ash): :(
from converter.postprocess import postprocess
for i in range(len(self.body)):
raw_body_i = self.parse_body(self.body[:i],
current_part='document')
unaugmented_meta_i = postprocess(raw_body_i,
transclusions,
asides=asides)[0]
if unaugmented_meta_i == unaugmented_meta:
self.body[:i] = []
return
raise Exception('failed to find the end of the metadata')
@staticmethod
def meta_to_docx(meta, intern_image, total_w):
tups = []
meta_copy = meta.raw_items().copy()
to_runs = partial(meta_to_runs,
intern_image=intern_image,
total_w=total_w)
for name in ['Title', 'Subtitle']:
bit = meta_copy.pop(name.lower(), None)
if bit:
pr = mkel(
'w:pPr',
{},
[
# FIXME(ash): currently we don't ensure the styles exist
mkel('w:pStyle', {'w:val': name}, [])
])
tups.append(make_p(pr, *to_runs(bit)))
for k, v in meta_copy.iteritems():
body = (to_runs([mkel('u', {}, [str(k) + ':']), ' ']) + to_runs(v))
tups.append(make_p(*body))
return [tup2etree(tup, nsmap=ns.dict) for tup in tups]
def intern_image(self, image):
assert isinstance(image, literal.Image)
f = StringIO(image.data)
rid = self.doc.add_image(f, image.mimetype)
return rid
def insert_meta(self, meta):
self.body[:0] = self.meta_to_docx(meta, self.intern_image,
self.textwidth_emu)
def save_to(self, f):
self.doc.save(f)