def unparse_literal(lit, roundtrip=True, plain=False): # pylint: disable=R0911
"""Return a string representation of `lit`.
- `roundtrip` affects how literals with context-dependent values are
hanlded, e.g. when ``roundtrip=False`` then
``Date('today') -> "2014-01-01"`` (instead of ``"today"``).
- `plain` controls if rich text content is converted to plaintext (e.g.
for pdf or epub metadata)
"""
# FIXME(alexander): try to break cyclic imports
import converter.html_writer
if lit is None:
return '' # XXX(alexander)
if isinstance(lit, basestring):
return lit
if isinstance(lit, bool):
return ('no', 'yes')[lit]
if not roundtrip and hasattr(lit, 'to_value'):
return lit.to_value()
if hasattr(lit, 'to_string'):
return lit.to_string()
if isinstance(lit, list): # Rich-text
if plain:
return plaintextify(lit)
return converter.html_writer.write_body(lit)
assert False, "Unknown literal type %r" % (lit,)
def unparse_literal(lit, roundtrip=True, plain=False): # pylint: disable=R0911
"""Return a string representation of `lit`.
- `roundtrip` affects how literals with context-dependent values are
hanlded, e.g. when ``roundtrip=False`` then
``Date('today') -> "2014-01-01"`` (instead of ``"today"``).
- `plain` controls if rich text content is converted to plaintext (e.g.
for pdf or epub metadata)
"""
# FIXME(alexander): try to break cyclic imports
import converter.html_writer
if lit is None:
return '' # XXX(alexander)
if isinstance(lit, basestring):
return lit
if isinstance(lit, bool):
return ('no', 'yes')[lit]
if not roundtrip and hasattr(lit, 'to_value'):
return lit.to_value()
if hasattr(lit, 'to_string'):
return lit.to_string()
if isinstance(lit, list): # Rich-text
if plain:
return plaintextify(lit)
return converter.html_writer.write_body(lit)
assert False, "Unknown literal type %r" % (lit, )
def tocify_heading(e, gensym):
"""Transform a heading into `('h*', {'id':ID}, [STRING])`.
This assumes `h*` already has an id or is followed by an anchor.
"""
# pylint: disable=C0103
h, a, b = e
assert h in H_TAGS
a, b = lift_anchor_id(a, b, gensym, kill_anchor=True)
return (h, {'id': a['id']}, [plaintextify(b)])
def tocify_heading(e, gensym):
"""Transform a heading into `('h*', {'id':ID}, [STRING])`.
This assumes `h*` already has an id or is followed by an anchor.
"""
# pylint: disable=C0103
h, a, b = e
assert h in H_TAGS
a, b = lift_anchor_id(a, b, gensym, kill_anchor=True)
return (h, {'id': a['id']}, [plaintextify(b)])
def check_supplied(): # pylint: disable=R0912
def try_to_reify(v, parse):
try:
return parse(v)
except (KeyboardInterrupt, SystemExit):
raise
except Exception as ex:
# pylint: disable=W0631
log.info('Meta conversion error on %s, %s', k, ex)
error('Not a valid %s format (expected %s)' % (
right_type, TYPE_EXAMPLES[right_type]),
k,
supplied=meta[k])
for k in meta:
canonical_meta[k] = meta[k] # default
if k not in self._info:
maybe_meants = spellsuggest.spell_suggest(
k, self._info.keys())
suggestion = (" (did you mean '%s'?)" % maybe_meants[0]
if maybe_meants else '')
if k not in ('title', 'subtitle'):
error("Unexpected field '%s'%s" % (k, suggestion),
k, meta[k])
else:
error("This document type does not have a %s" % k,
k, meta[k])
continue
potential_types = PY_TYPE_TO_TYPESETR_TYPES[type(meta[k])]
right_type = self._info[k]['type']
if right_type in potential_types:
if right_type == 'bibliography':
parsed[k] = try_to_reify(meta[k], Bibliography)
else:
if 'rich-text' in potential_types:
if not isinstance(meta[k], basestring):
meta[k] = postprocess.plaintextify(meta[k])
potential_types = ('text',)
if potential_types == ('text',):
parsed[k] = try_to_reify(
meta[k],
# pylint: disable=W0640
lambda v: parse_literal(v, right_type))
else:
error("Expected meta field '%s:' to be"
" of type '%s', not '%s'" % (
k, right_type, potential_types[0]),
k, supplied=meta[k])
def latexify(self, ast): # pylint: disable=E0102,R0914,R0915,R0911,R0912
if isinstance(ast, list):
return re.sub('\n\n$', '\n',
join(*map(self.latexify, ast)))
else:
node = ast
if isinstance(node, basestring):
return quote(node)
else:
assert isinstance(node, tuple)
h, a, b = node
if h == 'div': # canonicalize pseudo-elements
h = a['class'].pop()
assert not a['class']
del a['class']
if h[:-1] == 'h':
if self.am_inside('list') or self.am_inside('table'):
return docwarn(
self.latexify(b),
'Cannot have sections inside lists or tables: %r' %
postprocess.plaintextify(b))
else:
with self.inside('section'):
if a:
log.warn('heading w/ attr %r', a)
labels, b = extract_labels(b)
return self.section(h, b, labels)
elif h == 'p':
ans = nl(self.latexify(b))
if self.am_inside('.footnote') and self.am_inside('table'):
return docwarn(ans,
'Multi-paragraph footnotes in tables are'
' unsupported')
return nl(ans)
elif h == 'span':
return self.latexify(b) # XXX
elif h in ('ol', 'ul'):
ol = partial(self.enumerate_,
start=a.get('start'),
series=a.get('id'),
resume=a.get('data-continue-list'))
with self.inside('list'):
return nl(
freshline({
'ol': ol,
'ul': itemize}[h](
self.latexify(b))))
elif h == 'li':
labels, b = extract_labels(b)
labelling = (join(*(map(mklabel, labels) + [' ']))
if labels else '')
return join(freshline(cmd('item')),
labelling, self.latexify(b))
elif h == 'table':
nested_table = self.am_inside('table')
with self.inside('table'):
# pylint: disable=C0103
CLASS_TO_SPEC = {'left': 'P', 'center': 'C',
'right': 'R', 'justify': 'N'}
b = b[:]
tablecaption = None
if b[0][0] == 'caption':
with self.inside('caption'):
tablecaption = self.latexify(b[0][2])
del b[0]
colgroup = [el for el in b if el[0] == 'colgroup']
rows = [el for el in b if el[0] == 'tr']
assert len(colgroup) == 1, \
"Expected single colgroup in table %s" % b
cols = colgroup[0][2]
colspecs = []
for col_h, col_a, col_b in cols:
if col_h != 'col':
break
assert not col_b
coltype = 'P'
for cls in CLASS_TO_SPEC:
if cls in col_a.get('class', []):
coltype = CLASS_TO_SPEC[cls]
coltype = "%s{%s}" % (coltype, textwidth_percent(
col_a['style']['width']))
colspecs.append(coltype)
rows = "\\tabularnewline\n".join(
map(self.latexify, rows))
if nested_table and tablecaption:
docproblem(
"Tables within tables can't have captions;"
" outputing caption as normal text",
level='warning')
ans = join(nl(table(colspecs, rows)), tablecaption)
else:
ans = table(colspecs, rows, tablecaption)
if self.post_float_yuck and not self.am_inside('table'):
ans = join(ans, *self.post_float_yuck)
del self.post_float_yuck[:]
return ans
elif h == 'col': # FIXME
assert False, "Unexpected col"
elif h == 'tr':
return " & ".join(map(self.latexify, b))
elif h == 'td':
if 'headcol' in a.get('class', []):
return colh(self.latexify(b))
return self.latexify(b)
elif h == 'th':
if 'headcol' in a.get('class', []):
return rowh(colh(self.latexify(b)))
return rowh(self.latexify(b))
elif h == 'figure':
b = b[:]
if b[0][0] == 'figcaption':
with self.inside('caption'):
figcaption = self.latexify(b[0][2])
del b[0]
else:
figcaption = None
assert len(b) == 1 and b[0][0] == 'img'
img = b[0][1]['src']
inline = False
warns = []
if a['style']['display'] == 'inline':
if self.am_inside('table'):
warns.append([
'Margin figures not supported in tables, '
'inserting into table cell'])
else:
inline = True
if inline:
if figcaption:
warns.append(
['Ignoring figcaption for inline figure:'
' "%s"', figcaption])
ans = marginfigure(img=img)
else:
fakecaption = figcaption and self.am_inside('table')
if fakecaption:
warns.append([
"Figures in tables can't have captions; "
"outputing caption as normal text"])
# inside blockquotes more complicated figure
# environments don't seem to work reliably
rawincludegraphics = self.am_inside('blockquote')
ans = figure(img=img,
classes=a.get('class', []),
width=a['style']['width'],
figcaption=figcaption,
fakecaption=fakecaption,
rawincludegraphics=rawincludegraphics)
if self.post_float_yuck and not self.am_inside('table'):
ans = join(ans, *self.post_float_yuck)
del self.post_float_yuck[:]
return ans if not warns else docwarns(ans, *warns)
elif h == 'img':
assert False, 'unexpected image'
elif h == 'a':
if 'name' in a:
# we can't do that blindly, because we want to
# generate labels for things like lists and headings
# this is only a fallback for anchors outside of
# 'labelled' envs
return cmd('hypertarget', [],
[a['name'].lstrip('#'), ''])
elif 'href' in a:
if a['href'].startswith('#'):
return cmd('hyperref',
[latexify_href(a['href'][1:])],
[self.latexify(b)])
##
# XXX(alexander): handle bare urls specially, because
# we want more relaxed linebreaking rules for them.
# Note that we're not using \url directly, because
# it's not robust and also can't cope with certain
# arguments, such as unbalanced '{'/'}'s. Also, even
# with fairly aggressive hyphenization params, this is
# in in itself not enough to resolve all overfull hbox
# issues with urls, although it's not 100% clear to me
# why.
elif b and a['href'] in (b[0], url_fix(b[0])):
# XXX(alexander): use url_fixed version here?
return urldef(a['href'], self.urldefs)
else:
ans = cmd('href', [], [latexify_href(a['href']),
self.latexify(b)])
if b[0].startswith('http'):
ans = docwarn(
ans,
'Suspicious link with body/href'
' mismatch: %r != %r' % (
a['href'].encode('utf-8'), b[0]))
return ans
else:
assert False, 'Malformed link: %s' % ((h, a, b),)
elif h == 'aside':
return cmd('comment', [], [self.latexify(b)])
elif h in ('b', 'i', 'u', 's'):
assert not a, 'unexpected <%s %r' % (h, a)
return self.handle_emphasis(h, b)
elif h == 'code':
#FIXME: write something more specialized
return cmd('texttt', [], [self.latexify(b)])
elif h == 'sup':
return cmd('textsuperscript', [], [self.latexify(b)])
elif h == 'sub':
return cmd('textsubscript', [], [self.latexify(b)])
elif h == '.footnote':
with self.inside('.footnote'):
if self.am_inside('caption'):
self.post_float_yuck.append(cmd('footnotetext',
[],
[self.latexify(b)]))
return cmd(r'protect\footnotemark', [], [])
else:
return cmd('footnote', [], [self.latexify(b)])
elif h == '.pagebreak':
return nl(cmd('clearpage', [], [self.latexify(b)]))
elif h == 'br':
assert a == {}
assert b == []
return nl(cmd('newline'))
elif h == 'blockquote':
with self.inside('blockquote'):
return blockquote(self.latexify(b))
elif (h == 'footer' and b == [Seq['cite', :]]
and self.am_inside('blockquote')):
return nl(cmd('attrib', [], [self.latexify(b[0][2])]))
elif node == ('CMD', {'class': ['$']}, b):
return join('$', b[0], '$')
elif node == ('CMD', {'class': [Var('CITE', CITE_REX.match)]},
b):
return self.munge_cite(node, b)
elif node == ('CMD', {'class': ['tex']}, b):
return b[0]
elif h in ('CMD', 'LIT'):
return self.bad_command(*node)
elif h == 'pre':
return highlight.as_latex(node)
elif h == 'wbr':
return '{}'
else:
#FIXME(alexander): set 1 as error-code?
log.error('Unexpected tag: %s %r %r', h, a, b)
return join("")
def parse_body(xml, context, normalize_transclusion):
# pylint: disable=R0912,R0915,R0914
for e in xml:
text = (e.text or '')
tail = (e.tail or '')
# some style properties should be promoted to tags, e.g. underlining
# and bolding
tags_from_style = []
stys_dealt_with = []
if e.tag in (S_TAG, TAB_TAG):
yield ' \t'[e.tag == TAB_TAG] * int(e.attrib.get(
ns.text('c'), '1'))
if tail:
yield tail
continue
if e.tag == LINEBREAK_TAG:
yield mkel('br', {}, [])
continue
sty = context.stys.get(
e.get(STYLE_NAME_ATTR) or e.get(TABLE_STYLE_NAME_ATTR))
# handle page breaks
if sty and sty.par_break:
assert e.tag in (H_TAG, P_TAG), \
"Unexpected page-break in %r" % e.tag
yield mkel('.pagebreak', {}, [])
stys_dealt_with.append('par_break')
# Handle lists specially
if e.tag == LIST_TAG:
new_context = context.bump_list_level(sty)
stys_dealt_with.append('sub_list_styles')
else:
new_context = context
body = list(parse_body(e, new_context, normalize_transclusion))
assert type(body) is list and not body or type(body[0]) is not list
attrs = {}
if text:
body = [text] + body
if sty and sty.type.endswith('title'):
head = sty.type
body = [plaintextify(body)]
sty = None
elif e.tag == H_TAG:
# skip empty headings; NB: this *must* happen
# after we extracted eventual page-breaks, which are the only
# useful information empty headings can contain
if blank(body):
continue
head = sty.type
# FIXME(alexander): keep track of the headings breadcrumbs in
# context for two reasons
#
# 1. to associate errors with specific headings
# 2. to warn about bad structure e.g. h1 followed by h4,
# rather than h2
elif e.tag == LIST_TAG:
head = new_context.list_type
assert head in ('ol', 'ul')
list_start = new_context.list_start
if list_start is not None:
assert head == 'ol'
attrs['start'] = str(list_start)
id_ = e.attrib.get(ns.xml('id')) # pylint: disable=E1101
if id_ is not None:
attrs['id'] = id_
continues = e.attrib.get(ns.text('continue-list'))
if continues is not None:
# make this a data attrib, so we can stuff it
# into the html, which doesn't have direct support
attrs['data-continue-list'] = continues
elif e.tag == LIST_ITEM_TAG:
head = 'li'
elif e.tag == ANNOTATION_TAG:
head = 'aside'
elif e.tag in (CREATOR_TAG, NOTE_CITATION_TAG, BOOKMARK_END_TAG):
#FIXME: extract content
if text:
log.warning('Hey, someone actually specified a %s: %s', e.tag,
text)
if tail:
yield tail
continue
elif e.tag == NOTE_TAG:
# other valid option is 'endnote'
assert e.attrib[ns.text('note-class')] == 'footnote'
# skip ahead and exit early; we only represent the note-body
assert len(e) == 2 and e[1].tag == NOTE_BODY_TAG
assert len(body) == 1
yield body[0]
if tail:
yield tail
continue
elif e.tag == NOTE_BODY_TAG:
head = '.footnote'
# FIXME(alexander): sucky hack to strip the bogus whitespace
# google docs enters at the beginning of a footnote for some
# reason. I should really write a more generic whitespace
# stripping mechanism in the postprocess module that can recognize
# consecutive whitespace even if seperated-by/wrapped-in inline
# tags.
_, B1, B2, = map(Var, '_, B1, B2'.split(', '))
SPACED_STR = Var(
'SPACED_STR', lambda s:
(isinstance(s, basestring) and re.match(r'\s+', s)))
if body == Seq[('p', _, Seq[SPACED_STR, B2:]), B1:]:
body[0][2][0] = SPACED_STR.val.lstrip()
# FIXME(alexander): add anchors for all paras
elif e.tag == P_TAG:
margin = sty.margin_left or sty.text_indent if sty else None
indent_level = in_indents(margin) if margin else 0
if indent_level:
head = '.block'
attrs['indent'] = indent_level
else:
head = 'p'
#FIXME styled links etc. gdocs might not use that...
#... but we should be able to handle non-span bolding etc.
elif e.tag == SPAN_TAG:
# XXX: order can matter; we need
# <b><u>command</u><b>
# not
# <u><b>command</b><u>
#
# but more generally the minimal coalescing of abutting partially
# overlapping styles is something that needs to be thought about
# properly at some point.
for attr, on_values, html_tags in [
('underline', [True], ['u']), ('font_weight', ['bold'], ['b']),
('font_style', ['italic'], ['i']),
('line_through', [True], ['s']),
('text_position', ['sub', 'super'], ['sub', 'sup'])
]:
value = getattr(sty, attr, None)
if value:
if value not in on_values:
log.error("Bad value for %s: %s in %s", attr, value,
e.tag)
continue
tags_from_style.append(html_tags[on_values.index(value)])
stys_dealt_with.append(attr)
if is_code_font(sty.font_family):
tags_from_style.append('code')
stys_dealt_with.append('font_family')
head = 'span'
elif e.tag == A_TAG:
assert e.attrib[ns.xlink('type')] == 'simple'
head = 'a'
attrs = dict(href=e.attrib[HREF_ATTR])
# FIXME the in 'span' check is a bit too general, should use
# something else to markup textcolor
body = tidy(whack(lambda x: x in ('span', 'u'), body))
elif e.tag == BOOKMARK_START_TAG:
head = 'a'
attrs = dict(name=e.attrib[TEXT_NAME_ATTR])
assert (blank(text) and blank(tail)
and next(e.itersiblings()).tag == BOOKMARK_END_TAG)
elif e.tag == TABLE_TAG:
head = 'table'
body = parse_table_body(body)
elif e.tag == TABLE_ROW_TAG:
head = 'tr'
elif e.tag == TABLE_CELL_TAG:
head = 'td'
#FIXME repetition via table:number-columns-repeated
#FIXME handle column-groups
elif e.tag == TABLE_COLUMN_TAG:
head = 'col'
sty = context.stys.get(e.attrib.get(ns.table('style-name')))
if sty and sty.width is not None:
# XXX this isn't really the column width
# since google moronically saves this even
# if set column width is turned off thank you google!
attrs = dict(style=OrderedDict(width=sty.width))
stys_dealt_with.append('width')
elif e.tag == FRAME_TAG:
# XXX: try to find caption
# FIXME(alexander): keep figures/tables with captions in context,
# so that we can produce a lot/loi; add an id for all of them
inline = e.attrib[ns.text('anchor-type')] == 'as-char'
width = (
e.attrib.get(ns.svg('width')) # pylint: disable=E1101
or e.attrib[ns.style('rel-width')])
# FIXME(alexander): should handle all these, in theory:
# <http://www.w3.org/TR/SVG11/struct.html#SVGElementWidthAttribute>
# ("em" | "ex" | "px" | "in" | "cm" | "mm" | "pt" | "pc" )
assert width.endswith('cm'), \
'Expected figure width in cm, got %s' % width
relwidth = float(width[:-2]) / context.stys.textwidth
head, attrs, body = make_figure(
relwidth=relwidth,
inline=inline,
# FIXME(alexander): the body[0][1] to access the image
# will blow up on leading whitespace in the body
body=list(x for x in body
if not (isinstance(x, basestring) and blank(x))),
src=body[0][1]['src'],
original_href=e.find(ns.draw('image')).get(ns.xlink('href')))
elif e.tag == IMAGE_TAG:
head = 'img'
attrs = dict(src=normalize_transclusion(e.attrib[HREF_ATTR]))
else:
log.warning('Ignoring tag %s', e.tag)
continue
# FIXME raise RuntimeError('Unexpected tag: %s' % e.tag)
sty_tagged = reduce(lambda parsed, tag: [mkel(tag, {}, parsed)],
tags_from_style, tidy(body))
if sty:
if sty.text_align:
stys_dealt_with.append('text_align')
attrs = add_class(attrs, sty.text_align)
if sty.background_color:
stys_dealt_with.append('background_color')
iadd_style(attrs, 'background-color', sty.background_color)
if sty.color:
stys_dealt_with.append('color')
iadd_style(attrs, 'color', sty.color)
if e.tag == LIST_TAG:
if new_context.list_style_type:
attrs = add_class(attrs, new_context.list_style_type)
# FIXME additional tidy
parsed = mkel(head, attrs, sty_tagged)
if head == 'span' and 'style' in attrs:
B = Var('B')
if parsed == ('span', attrs, [('code', {}, B)]):
parsed = mkel('code', {}, [('span', attrs, B.val)])
leftover_styles = sty and set(
sty.active_props()) - set(stys_dealt_with)
if leftover_styles:
log.warn('Ignoring style elements: %r in %r "%s"',
([(k, getattr(sty, k)) for k in leftover_styles]), head,
plaintextify(body))
preprocess.maybe_anchorize_id(head, attrs, sty_tagged)
yield parsed
if tail:
yield tail
def latexify(self, ast): # pylint: disable=E0102,R0914,R0915,R0911,R0912
if isinstance(ast, list):
return re.sub('\n\n$', '\n', join(*map(self.latexify, ast)))
else:
node = ast
if isinstance(node, basestring):
return quote(node)
else:
assert isinstance(node, tuple)
h, a, b = node
if h == 'div': # canonicalize pseudo-elements
h = a['class'].pop()
assert not a['class']
del a['class']
if h[:-1] == 'h':
if self.am_inside('list') or self.am_inside('table'):
return docwarn(
self.latexify(b),
'Cannot have sections inside lists or tables: %r' %
postprocess.plaintextify(b))
else:
with self.inside('section'):
if a:
log.warn('heading w/ attr %r', a)
labels, b = extract_labels(b)
return self.section(h, b, labels)
elif h == 'p':
ans = nl(self.latexify(b))
if self.am_inside('.footnote') and self.am_inside('table'):
return docwarn(
ans, 'Multi-paragraph footnotes in tables are'
' unsupported')
return nl(ans)
elif h == 'span':
return self.latexify(b) # XXX
elif h in ('ol', 'ul'):
ol = partial(self.enumerate_,
start=a.get('start'),
series=a.get('id'),
resume=a.get('data-continue-list'))
with self.inside('list'):
return nl(
freshline({
'ol': ol,
'ul': itemize
}[h](self.latexify(b))))
elif h == 'li':
labels, b = extract_labels(b)
labelling = (join(*(map(mklabel, labels) +
[' '])) if labels else '')
return join(freshline(cmd('item')), labelling,
self.latexify(b))
elif h == 'table':
nested_table = self.am_inside('table')
with self.inside('table'):
# pylint: disable=C0103
CLASS_TO_SPEC = {
'left': 'P',
'center': 'C',
'right': 'R',
'justify': 'N'
}
b = b[:]
tablecaption = None
if b[0][0] == 'caption':
with self.inside('caption'):
tablecaption = self.latexify(b[0][2])
del b[0]
colgroup = [el for el in b if el[0] == 'colgroup']
rows = [el for el in b if el[0] == 'tr']
assert len(colgroup) == 1, \
"Expected single colgroup in table %s" % b
cols = colgroup[0][2]
colspecs = []
for col_h, col_a, col_b in cols:
if col_h != 'col':
break
assert not col_b
coltype = 'P'
for cls in CLASS_TO_SPEC:
if cls in col_a.get('class', []):
coltype = CLASS_TO_SPEC[cls]
coltype = "%s{%s}" % (coltype,
textwidth_percent(
col_a['style']['width']))
colspecs.append(coltype)
rows = "\\tabularnewline\n".join(
map(self.latexify, rows))
if nested_table and tablecaption:
docproblem(
"Tables within tables can't have captions;"
" outputing caption as normal text",
level='warning')
ans = join(nl(table(colspecs, rows)), tablecaption)
else:
ans = table(colspecs, rows, tablecaption)
if self.post_float_yuck and not self.am_inside('table'):
ans = join(ans, *self.post_float_yuck)
del self.post_float_yuck[:]
return ans
elif h == 'col': # FIXME
assert False, "Unexpected col"
elif h == 'tr':
return " & ".join(map(self.latexify, b))
elif h == 'td':
if 'headcol' in a.get('class', []):
return colh(self.latexify(b))
return self.latexify(b)
elif h == 'th':
if 'headcol' in a.get('class', []):
return rowh(colh(self.latexify(b)))
return rowh(self.latexify(b))
elif h == 'figure':
b = b[:]
if b[0][0] == 'figcaption':
with self.inside('caption'):
figcaption = self.latexify(b[0][2])
del b[0]
else:
figcaption = None
assert len(b) == 1 and b[0][0] == 'img'
img = b[0][1]['src']
inline = False
warns = []
if a['style']['display'] == 'inline':
if self.am_inside('table'):
warns.append([
'Margin figures not supported in tables, '
'inserting into table cell'
])
else:
inline = True
if inline:
if figcaption:
warns.append([
'Ignoring figcaption for inline figure:'
' "%s"', figcaption
])
ans = marginfigure(img=img)
else:
fakecaption = figcaption and self.am_inside('table')
if fakecaption:
warns.append([
"Figures in tables can't have captions; "
"outputing caption as normal text"
])
# inside blockquotes more complicated figure
# environments don't seem to work reliably
rawincludegraphics = self.am_inside('blockquote')
ans = figure(img=img,
classes=a.get('class', []),
width=a['style']['width'],
figcaption=figcaption,
fakecaption=fakecaption,
rawincludegraphics=rawincludegraphics)
if self.post_float_yuck and not self.am_inside('table'):
ans = join(ans, *self.post_float_yuck)
del self.post_float_yuck[:]
return ans if not warns else docwarns(ans, *warns)
elif h == 'img':
assert False, 'unexpected image'
elif h == 'a':
if 'name' in a:
# we can't do that blindly, because we want to
# generate labels for things like lists and headings
# this is only a fallback for anchors outside of
# 'labelled' envs
return cmd('hypertarget', [],
[a['name'].lstrip('#'), ''])
elif 'href' in a:
if a['href'].startswith('#'):
return cmd('hyperref',
[latexify_href(a['href'][1:])],
[self.latexify(b)])
##
# XXX(alexander): handle bare urls specially, because
# we want more relaxed linebreaking rules for them.
# Note that we're not using \url directly, because
# it's not robust and also can't cope with certain
# arguments, such as unbalanced '{'/'}'s. Also, even
# with fairly aggressive hyphenization params, this is
# in in itself not enough to resolve all overfull hbox
# issues with urls, although it's not 100% clear to me
# why.
elif b and a['href'] in (b[0], url_fix(b[0])):
# XXX(alexander): use url_fixed version here?
return urldef(a['href'], self.urldefs)
else:
ans = cmd(
'href', [],
[latexify_href(a['href']),
self.latexify(b)])
if b[0].startswith('http'):
ans = docwarn(
ans, 'Suspicious link with body/href'
' mismatch: %r != %r' %
(a['href'].encode('utf-8'), b[0]))
return ans
else:
assert False, 'Malformed link: %s' % ((h, a, b), )
elif h == 'aside':
return cmd('comment', [], [self.latexify(b)])
elif h in ('b', 'i', 'u', 's'):
assert not a, 'unexpected <%s %r' % (h, a)
return self.handle_emphasis(h, b)
elif h == 'code':
#FIXME: write something more specialized
return cmd('texttt', [], [self.latexify(b)])
elif h == 'sup':
return cmd('textsuperscript', [], [self.latexify(b)])
elif h == 'sub':
return cmd('textsubscript', [], [self.latexify(b)])
elif h == '.footnote':
with self.inside('.footnote'):
if self.am_inside('caption'):
self.post_float_yuck.append(
cmd('footnotetext', [], [self.latexify(b)]))
return cmd(r'protect\footnotemark', [], [])
else:
return cmd('footnote', [], [self.latexify(b)])
elif h == '.pagebreak':
return nl(cmd('clearpage', [], [self.latexify(b)]))
elif h == 'br':
assert a == {}
assert b == []
return nl(cmd('newline'))
elif h == 'blockquote':
with self.inside('blockquote'):
return blockquote(self.latexify(b))
elif (h == 'footer' and b == [Seq['cite', :]]
and self.am_inside('blockquote')):
return nl(cmd('attrib', [], [self.latexify(b[0][2])]))
elif node == ('CMD', {'class': ['$']}, b):
return join('$', b[0], '$')
elif node == ('CMD', {
'class': [Var('CITE', CITE_REX.match)]
}, b):
return self.munge_cite(node, b)
elif node == ('CMD', {'class': ['tex']}, b):
return b[0]
elif h in ('CMD', 'LIT'):
return self.bad_command(*node)
elif h == 'pre':
return highlight.as_latex(node)
elif h == 'wbr':
return '{}'
else:
#FIXME(alexander): set 1 as error-code?
log.error('Unexpected tag: %s %r %r', h, a, b)
return join("")
def parse_body(xml, context, normalize_transclusion):
# pylint: disable=R0912,R0915,R0914
for e in xml:
text = (e.text or '')
tail = (e.tail or '')
# some style properties should be promoted to tags, e.g. underlining
# and bolding
tags_from_style = []
stys_dealt_with = []
if e.tag in (S_TAG, TAB_TAG):
yield ' \t'[e.tag == TAB_TAG] * int(e.attrib.get(ns.text('c'), '1'))
if tail:
yield tail
continue
if e.tag == LINEBREAK_TAG:
yield mkel('br', {}, [])
continue
sty = context.stys.get(e.get(STYLE_NAME_ATTR) or
e.get(TABLE_STYLE_NAME_ATTR))
# handle page breaks
if sty and sty.par_break:
assert e.tag in (H_TAG, P_TAG), \
"Unexpected page-break in %r" % e.tag
yield mkel('.pagebreak', {}, [])
stys_dealt_with.append('par_break')
# Handle lists specially
if e.tag == LIST_TAG:
new_context = context.bump_list_level(sty)
stys_dealt_with.append('sub_list_styles')
else:
new_context = context
body = list(parse_body(e, new_context, normalize_transclusion))
assert type(body) is list and not body or type(body[0]) is not list
attrs = {}
if text:
body = [text] + body
if sty and sty.type.endswith('title'):
head = sty.type
body = [plaintextify(body)]
sty = None
elif e.tag == H_TAG:
# skip empty headings; NB: this *must* happen
# after we extracted eventual page-breaks, which are the only
# useful information empty headings can contain
if blank(body):
continue
head = sty.type
# FIXME(alexander): keep track of the headings breadcrumbs in
# context for two reasons
#
# 1. to associate errors with specific headings
# 2. to warn about bad structure e.g. h1 followed by h4,
# rather than h2
elif e.tag == LIST_TAG:
head = new_context.list_type
assert head in ('ol', 'ul')
list_start = new_context.list_start
if list_start is not None:
assert head == 'ol'
attrs['start'] = str(list_start)
id_ = e.attrib.get(ns.xml('id')) # pylint: disable=E1101
if id_ is not None:
attrs['id'] = id_
continues = e.attrib.get(ns.text('continue-list'))
if continues is not None:
# make this a data attrib, so we can stuff it
# into the html, which doesn't have direct support
attrs['data-continue-list'] = continues
elif e.tag == LIST_ITEM_TAG:
head = 'li'
elif e.tag == ANNOTATION_TAG:
head = 'aside'
elif e.tag in (CREATOR_TAG, NOTE_CITATION_TAG, BOOKMARK_END_TAG):
#FIXME: extract content
if text:
log.warning('Hey, someone actually specified a %s: %s',
e.tag, text)
if tail:
yield tail
continue
elif e.tag == NOTE_TAG:
# other valid option is 'endnote'
assert e.attrib[ns.text('note-class')] == 'footnote'
# skip ahead and exit early; we only represent the note-body
assert len(e) == 2 and e[1].tag == NOTE_BODY_TAG
assert len(body) == 1
yield body[0]
if tail:
yield tail
continue
elif e.tag == NOTE_BODY_TAG:
head = '.footnote'
# FIXME(alexander): sucky hack to strip the bogus whitespace
# google docs enters at the beginning of a footnote for some
# reason. I should really write a more generic whitespace
# stripping mechanism in the postprocess module that can recognize
# consecutive whitespace even if seperated-by/wrapped-in inline
# tags.
_, B1, B2, = map(Var, '_, B1, B2'.split(', '))
SPACED_STR = Var('SPACED_STR', lambda s: (isinstance(s, basestring)
and re.match(r'\s+', s)))
if body == Seq[('p', _, Seq[SPACED_STR, B2:]), B1:]:
body[0][2][0] = SPACED_STR.val.lstrip()
# FIXME(alexander): add anchors for all paras
elif e.tag == P_TAG:
margin = sty.margin_left or sty.text_indent if sty else None
indent_level = in_indents(margin) if margin else 0
if indent_level:
head = '.block'
attrs['indent'] = indent_level
else:
head = 'p'
#FIXME styled links etc. gdocs might not use that...
#... but we should be able to handle non-span bolding etc.
elif e.tag == SPAN_TAG:
# XXX: order can matter; we need
# <b><u>command</u><b>
# not
# <u><b>command</b><u>
#
# but more generally the minimal coalescing of abutting partially
# overlapping styles is something that needs to be thought about
# properly at some point.
for attr, on_values, html_tags in [
('underline', [True], ['u']),
('font_weight', ['bold'], ['b']),
('font_style', ['italic'], ['i']),
('line_through', [True], ['s']),
('text_position',
['sub', 'super'],
['sub', 'sup'])
]:
value = getattr(sty, attr, None)
if value:
if value not in on_values:
log.error("Bad value for %s: %s in %s",
attr, value, e.tag)
continue
tags_from_style.append(html_tags[on_values.index(value)])
stys_dealt_with.append(attr)
if is_code_font(sty.font_family):
tags_from_style.append('code')
stys_dealt_with.append('font_family')
head = 'span'
elif e.tag == A_TAG:
assert e.attrib[ns.xlink('type')] == 'simple'
head = 'a'
attrs = dict(href=e.attrib[HREF_ATTR])
# FIXME the in 'span' check is a bit too general, should use
# something else to markup textcolor
body = tidy(whack(lambda x: x in ('span', 'u'), body))
elif e.tag == BOOKMARK_START_TAG:
head = 'a'
attrs = dict(name=e.attrib[TEXT_NAME_ATTR])
assert (blank(text) and blank(tail) and
next(e.itersiblings()).tag == BOOKMARK_END_TAG)
elif e.tag == TABLE_TAG:
head = 'table'
body = parse_table_body(body)
elif e.tag == TABLE_ROW_TAG:
head = 'tr'
elif e.tag == TABLE_CELL_TAG:
head = 'td'
#FIXME repetition via table:number-columns-repeated
#FIXME handle column-groups
elif e.tag == TABLE_COLUMN_TAG:
head = 'col'
sty = context.stys.get(e.attrib.get(ns.table('style-name')))
if sty and sty.width is not None:
# XXX this isn't really the column width
# since google moronically saves this even
# if set column width is turned off thank you google!
attrs = dict(style=OrderedDict(width=sty.width))
stys_dealt_with.append('width')
elif e.tag == FRAME_TAG:
# XXX: try to find caption
# FIXME(alexander): keep figures/tables with captions in context,
# so that we can produce a lot/loi; add an id for all of them
inline = e.attrib[ns.text('anchor-type')] == 'as-char'
width = (e.attrib.get(ns.svg('width')) # pylint: disable=E1101
or e.attrib[ns.style('rel-width')])
# FIXME(alexander): should handle all these, in theory:
# <http://www.w3.org/TR/SVG11/struct.html#SVGElementWidthAttribute>
# ("em" | "ex" | "px" | "in" | "cm" | "mm" | "pt" | "pc" )
assert width.endswith('cm'), \
'Expected figure width in cm, got %s' % width
relwidth = float(width[:-2]) / context.stys.textwidth
head, attrs, body = make_figure(
relwidth=relwidth, inline=inline,
# FIXME(alexander): the body[0][1] to access the image
# will blow up on leading whitespace in the body
body=list(x for x in body
if not (isinstance(x, basestring) and blank(x))),
src=body[0][1]['src'],
original_href=e.find(ns.draw('image')).get(ns.xlink('href')))
elif e.tag == IMAGE_TAG:
head = 'img'
attrs = dict(src=normalize_transclusion(e.attrib[HREF_ATTR]))
else:
log.warning('Ignoring tag %s', e.tag)
continue
# FIXME raise RuntimeError('Unexpected tag: %s' % e.tag)
sty_tagged = reduce(lambda parsed, tag: [mkel(tag, {}, parsed)],
tags_from_style, tidy(body))
if sty:
if sty.text_align:
stys_dealt_with.append('text_align')
attrs = add_class(attrs, sty.text_align)
if sty.background_color:
stys_dealt_with.append('background_color')
iadd_style(attrs, 'background-color', sty.background_color)
if sty.color:
stys_dealt_with.append('color')
iadd_style(attrs, 'color', sty.color)
if e.tag == LIST_TAG:
if new_context.list_style_type:
attrs = add_class(attrs, new_context.list_style_type)
# FIXME additional tidy
parsed = mkel(head, attrs, sty_tagged)
if head == 'span' and 'style' in attrs:
B = Var('B')
if parsed == ('span', attrs, [('code', {}, B)]):
parsed = mkel('code', {}, [('span', attrs, B.val)])
leftover_styles = sty and set(sty.active_props()) - set(stys_dealt_with)
if leftover_styles:
log.warn('Ignoring style elements: %r in %r "%s"', (
[(k, getattr(sty, k)) for k in leftover_styles]), head,
plaintextify(body))
preprocess.maybe_anchorize_id(head, attrs, sty_tagged)
yield parsed
if tail:
yield tail