def postprocess(raw_body, transclusions, bibliography=None, asides=False):
citations = set()
# kill comments early, because they can mess things up
# (e.g. by splitting commands or citations)
# FIXME(alexander): comments should probably be out-of-band
if not asides:
raw_body = whack('aside'.__eq__, raw_body, kill_body=True)
# FIXME(alexander): investigate the performance impact of this final tidy
# -- it's needed for tidying up stuff that happened after commandification
# (currently that only affects blockquotes)
raw_parsed_body = tidy(
space_normalize(
list(
unwrap_figures(
captionize(
underlines_to_commands(
parse_cites(coalesce(raw_body),
bib_entries=getattr(
bibliography, 'entries', {}),
collect_cite=citations.add)))))))
unaugmented_head, body = extract_meta(raw_parsed_body, transclusions)
if citations:
if 'bibliography' not in unaugmented_head:
docproblem(MISSING_BIBLIOGRAPHY, sorted(citations)[0])
return unaugmented_head, body
def bad_command(self, head, attrs, body):
assert head in ('LIT', 'CMD')
bad_cmd = attrs['class'][0]
n = docproblem('Unknown command: {}', bad_cmd)
warning = small(red(self.latexify(
u"CONVERSION ERROR: Not a valid command"
u" (only use underlining for commands): “")))
the_cmd = self.latexify(
mkel('u', {}, [bad_cmd + (':' if head == 'CMD' else '')]))
warning_end = small(red(self.latexify(u'”')))
return join(problem_anchor(n, join(warning, the_cmd, warning_end)),
self.latexify(body))
def postprocess(raw_body, transclusions, bibliography=None, asides=False):
citations = set()
# kill comments early, because they can mess things up
# (e.g. by splitting commands or citations)
# FIXME(alexander): comments should probably be out-of-band
if not asides:
raw_body = whack('aside'.__eq__, raw_body, kill_body=True)
# FIXME(alexander): investigate the performance impact of this final tidy
# -- it's needed for tidying up stuff that happened after commandification
# (currently that only affects blockquotes)
raw_parsed_body = tidy(
space_normalize(list(unwrap_figures(captionize(
underlines_to_commands(
parse_cites(coalesce(raw_body),
bib_entries=getattr(bibliography, 'entries', {}),
collect_cite=citations.add)))))))
unaugmented_head, body = extract_meta(raw_parsed_body, transclusions)
if citations:
if 'bibliography' not in unaugmented_head:
docproblem(MISSING_BIBLIOGRAPHY, sorted(citations)[0])
return unaugmented_head, body
def bad_command(self, head, attrs, body):
assert head in ('LIT', 'CMD')
bad_cmd = attrs['class'][0]
n = docproblem('Unknown command: {}', bad_cmd)
warning = small(
red(
self.latexify(u"CONVERSION ERROR: Not a valid command"
u" (only use underlining for commands): “")))
the_cmd = self.latexify(
mkel('u', {}, [bad_cmd + (':' if head == 'CMD' else '')]))
warning_end = small(red(self.latexify(u'”')))
return join(problem_anchor(n, join(warning, the_cmd, warning_end)),
self.latexify(body))
def captionize(body):
CBODY, TAG, PATTRS, FATTRS, FBODY = map(
Var, 'CBODY, TAG, PATTRS, FATTRS, FBODY'.split(', '))
ans = []
for e1, e2 in window(body, 2):
if e2 in (varcmd('caption', CBODY),
('p', PATTRS,
[varcmd('caption', CBODY)])):
#XXX(alexander): the right way would probably be to normalize
# justify/left away before we get here.
e1_is_figure = ((e1 == ('p', PATTRS, [(TAG, FATTRS, FBODY)]) and
PATTRS.val in ({}, {'class': ['justify']},
{'class': ['left']})
or e1 == (TAG, FATTRS, FBODY))
and TAG.val in ('table', 'figure'))
if not e1_is_figure:
if TAG.match and TAG.val in H_TAGS:
docproblem(CAPTION_AFTER_HEADING,
plaintextify(CBODY.val),
plaintextify(FBODY.val))
else:
docproblem(CAPTION_AFTER_NON_FLOAT,
plaintextify(CBODY.val))
continue
if PATTRS.match and PATTRS.val:
log.warn(
'Unexpected attrs in paragraph wrapping the caption: %r',
PATTRS.val)
ans[-1] = (TAG.val, FATTRS.val,
[('figcaption' if TAG.val == 'figure' else 'caption',
{},
CBODY.val)] + captionize(FBODY.val))
elif e2 == (TAG, FATTRS, FBODY):
ans.append(mkel(TAG.val, FATTRS.val, captionize(FBODY.val)))
else:
ans.append(e2)
return ans
def captionize(body):
CBODY, TAG, PATTRS, FATTRS, FBODY = map(
Var, 'CBODY, TAG, PATTRS, FATTRS, FBODY'.split(', '))
ans = []
for e1, e2 in window(body, 2):
if e2 in (varcmd('caption',
CBODY), ('p', PATTRS, [varcmd('caption', CBODY)])):
#XXX(alexander): the right way would probably be to normalize
# justify/left away before we get here.
e1_is_figure = ((e1 == ('p', PATTRS, [(TAG, FATTRS, FBODY)])
and PATTRS.val in ({}, {
'class': ['justify']
}, {
'class': ['left']
}) or e1 == (TAG, FATTRS, FBODY))
and TAG.val in ('table', 'figure'))
if not e1_is_figure:
if TAG.match and TAG.val in H_TAGS:
docproblem(CAPTION_AFTER_HEADING, plaintextify(CBODY.val),
plaintextify(FBODY.val))
else:
docproblem(CAPTION_AFTER_NON_FLOAT,
plaintextify(CBODY.val))
continue
if PATTRS.match and PATTRS.val:
log.warn(
'Unexpected attrs in paragraph wrapping the caption: %r',
PATTRS.val)
ans[-1] = (TAG.val, FATTRS.val,
[('figcaption' if TAG.val == 'figure' else 'caption',
{}, CBODY.val)] + captionize(FBODY.val))
elif e2 == (TAG, FATTRS, FBODY):
ans.append(mkel(TAG.val, FATTRS.val, captionize(FBODY.val)))
else:
ans.append(e2)
return ans
def handle_fragment(fragment, indent, transclusions, h_shift, epub_clean,
bibliography):
# pylint: disable=R0911,R0914,R0912,R0913,R0915
# FIXME(alexander): clean this up a bit, and get rid of pylint muffles
if isinstance(fragment, basestring):
return cgi.escape(fragment)
(tag, attrs, content) = fragment
if tag in ['script', 'style'] and content:
content_str, = content
return NOT_INLINE_TEMPLATE % dict(
indent=indent,
tag=tag,
attrs_str=encode_attrs(attrs, transclusions, epub_clean),
content_str=_indent(
'\n' + maybe_cdatafy(_indent(content_str.strip('\n'), ' ')),
indent))
if tag == 'pre':
return '\n' + highlight.as_html(fragment)
# special case figures and tables
if tag == 'figure':
style = attrs['style'].copy()
width = style.pop('width', '100%')
attrs = dict(attrs.items(), style=style)
# FIXME(alexander): dirty hacks to fixup caption & width
img = content[-1]
assert img[0] == 'img'
img[1].setdefault('style', OrderedDict())['width'] = width
# put figcaption towards end
if content[0][0] == 'figcaption':
content[0], content[-1] = content[-1], content[0]
if style['display'] == 'inline':
ATTRS = Var('ATTRS') # pylint: disable=C0103
assert content[:1] == [('img', ATTRS, [])], \
"figure does not begin with an img"
attrs = add_class(ATTRS.val, 'margin')
# peel of the figure tag for inlined stuff
# as a hack to make epub/html validate
# (figures can't occur in all contexts imgs can)
return handle_fragments([('img', attrs, [])],
bibliography=bibliography,
indent=indent,
transclusions=transclusions,
h_shift=h_shift,
epub_clean=epub_clean)
elif tag == 'table':
colgroups = [el for el in content if el[0] == 'colgroup']
COLS = Var("COLS") # pylint: disable=C0103
assert colgroups == [('colgroup', {}, COLS)], \
"Expected single colgroup in table %s" % content
# FIXME(alexander): this deepcopy is a lazy hack so we can mutate away
# imperatively propagate table cell alignment down
# this is a pretty horrible hack and would blow
# up nastily if there is attribute aliasing,
# but deepcopying should kinda make it work
content = copy.deepcopy(content)
_propagate_alignment(content, COLS.val)
elif tag == 'col':
if not epub_clean:
attrs = attrs.copy()
attrs['width'] = attrs['style']['width']
del attrs['style']
# cull
## return handle_fragments(content, indent)
# FIXME(alexander): might make more sense to filter (or h-ify) these out
# elsewhere, but for now this seems not unreasonable
elif tag == 'title':
tag = 'h1'
attrs = add_class(attrs, 'title')
elif tag == 'subtitle':
tag = 'h2'
attrs = add_class(attrs, 'subtitle')
elif tag in ('CMD', 'LIT'):
bad_command = None
cmd_type, = attrs['class']
# FIXME(alexander): convert tex to html for non-math;
# convert tex math to MML for epub
if cmd_type in ('$', 'tex'):
tex, = content
if cmd_type == '$':
tex = r'\(%s\)' % tex
return '<span class="tex2jax_process">%s</span>' % cgi.escape(tex)
elif CITE_REX.match(cmd_type):
if bibliography:
bibliography.cited.add(content[0])
# post = ('[%s]' % content[1] if len(content) > 1 and content[1]
# else '')
# Post is ignored for the moment
return _format_citation(cmd_type, content[0], bibliography)
else:
docerror.docproblem(
'Citation exists, but bibliography is missing')
else:
bad_command = cmd_type + (':' if content else '')
docerror.docproblem('Unknown command type:%s' % cmd_type)
elif epub_clean:
if tag == 'a' and 'name' in attrs:
assert len(attrs) == 1
attrs = {'id': attrs['name']}
elif tag == 'img':
attrs = {
k: attrs[k]
for k in attrs if k not in ('width', 'height')
}
# FIXME(alexander): support continued-list properly in html, by keeping
# track of numbers of items per list-id and translating it to start
if tag in H_TAGS:
if h_shift:
tag = 'h%d' % min(len(H_TAGS), max(1, int(tag[1]) + h_shift))
# generic [tagname].class tags
if '.' in tag:
if tag == '.pagebreak':
tag = 'div.pagebreak' # for whitespace sanitization
tagname, classname = tag.split('.', 1)
tag = tagname or 'span'
attrs = add_class(attrs, classname)
if tag == 'CMD' and bad_command:
tag = 'span'
attrs = {'class': ['bad-command']}
content = [('u', {}, [bad_command])] + content
elif tag == 'ERR':
tag = 'span'
attrs = {'class': ['err'], 'title': attrs['info'][0]}
content_str = handle_fragments(content,
indent=' ' + indent,
transclusions=transclusions,
h_shift=h_shift,
epub_clean=epub_clean,
bibliography=bibliography)
if tag in VOID_TAGS:
assert not content
template = "<%(tag)s%(attrs_str)s/>"
elif tag in INLINE:
template = "<%(tag)s%(attrs_str)s>%(content_str)s</%(tag)s>"
elif '\n' in content_str:
template = NOT_INLINE_TEMPLATE
else:
template = COMPACT_NOT_INLINE_TEMPLATE
# FIXME(alexander): disgusting hack; fix this properly and
# use a set representation to start with!
classes = attrs.get('class')
if classes:
attrs = attrs.copy()
attrs['class'] = sorted(set(classes))
return template % dict(indent=indent,
tag=tag,
attrs_str=encode_attrs(attrs, transclusions,
epub_clean),
content_str=content_str)
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 docwarns(latex_body, *warnings):
ns = [docproblem(*warning, level='warning') for warning in warnings]
return reduce_right(problem_anchor, ns, latex_body)
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 docwarns(latex_body, *warnings):
ns = [docproblem(*warning, level='warning') for warning in warnings]
return reduce_right(problem_anchor, ns, latex_body)
def handle_fragment(fragment, indent,
transclusions, h_shift, epub_clean, bibliography):
# pylint: disable=R0911,R0914,R0912,R0913,R0915
# FIXME(alexander): clean this up a bit, and get rid of pylint muffles
if isinstance(fragment, basestring):
return cgi.escape(fragment)
(tag, attrs, content) = fragment
if tag in ['script', 'style'] and content:
content_str, = content
return NOT_INLINE_TEMPLATE % dict(
indent=indent,
tag=tag,
attrs_str=encode_attrs(attrs, transclusions, epub_clean),
content_str=_indent(
'\n' + maybe_cdatafy(_indent(content_str.strip('\n'), ' ')),
indent))
if tag == 'pre':
return '\n' + highlight.as_html(fragment)
# special case figures and tables
if tag == 'figure':
style = attrs['style'].copy()
width = style.pop('width', '100%')
attrs = dict(attrs.items(), style=style)
# FIXME(alexander): dirty hacks to fixup caption & width
img = content[-1]
assert img[0] == 'img'
img[1].setdefault('style', OrderedDict())['width'] = width
# put figcaption towards end
if content[0][0] == 'figcaption':
content[0], content[-1] = content[-1], content[0]
if style['display'] == 'inline':
ATTRS = Var('ATTRS') # pylint: disable=C0103
assert content[:1] == [('img', ATTRS, [])], \
"figure does not begin with an img"
attrs = add_class(ATTRS.val, 'margin')
# peel of the figure tag for inlined stuff
# as a hack to make epub/html validate
# (figures can't occur in all contexts imgs can)
return handle_fragments([('img', attrs, [])],
bibliography=bibliography,
indent=indent,
transclusions=transclusions,
h_shift=h_shift,
epub_clean=epub_clean)
elif tag == 'table':
colgroups = [el for el in content if el[0] == 'colgroup']
COLS = Var("COLS") # pylint: disable=C0103
assert colgroups == [('colgroup', {}, COLS)], \
"Expected single colgroup in table %s" % content
# FIXME(alexander): this deepcopy is a lazy hack so we can mutate away
# imperatively propagate table cell alignment down
# this is a pretty horrible hack and would blow
# up nastily if there is attribute aliasing,
# but deepcopying should kinda make it work
content = copy.deepcopy(content)
_propagate_alignment(content, COLS.val)
elif tag == 'col':
if not epub_clean:
attrs = attrs.copy()
attrs['width'] = attrs['style']['width']
del attrs['style']
# cull
## return handle_fragments(content, indent)
# FIXME(alexander): might make more sense to filter (or h-ify) these out
# elsewhere, but for now this seems not unreasonable
elif tag == 'title':
tag = 'h1'
attrs = add_class(attrs, 'title')
elif tag == 'subtitle':
tag = 'h2'
attrs = add_class(attrs, 'subtitle')
elif tag in ('CMD', 'LIT'):
bad_command = None
cmd_type, = attrs['class']
# FIXME(alexander): convert tex to html for non-math;
# convert tex math to MML for epub
if cmd_type in ('$', 'tex'):
tex, = content
if cmd_type == '$':
tex = r'\(%s\)' % tex
return '<span class="tex2jax_process">%s</span>' % cgi.escape(tex)
elif CITE_REX.match(cmd_type):
if bibliography:
bibliography.cited.add(content[0])
# post = ('[%s]' % content[1] if len(content) > 1 and content[1]
# else '')
# Post is ignored for the moment
return _format_citation(cmd_type, content[0], bibliography)
else:
docerror.docproblem(
'Citation exists, but bibliography is missing')
else:
bad_command = cmd_type + (':' if content else '')
docerror.docproblem('Unknown command type:%s' % cmd_type)
elif epub_clean:
if tag == 'a' and 'name' in attrs:
assert len(attrs) == 1
attrs = {'id': attrs['name']}
elif tag == 'img':
attrs = {k: attrs[k] for k in attrs if k not in ('width', 'height')}
# FIXME(alexander): support continued-list properly in html, by keeping
# track of numbers of items per list-id and translating it to start
if tag in H_TAGS:
if h_shift:
tag = 'h%d' % min(len(H_TAGS), max(1, int(tag[1]) + h_shift))
# generic [tagname].class tags
if '.' in tag:
if tag == '.pagebreak':
tag = 'div.pagebreak' # for whitespace sanitization
tagname, classname = tag.split('.', 1)
tag = tagname or 'span'
attrs = add_class(attrs, classname)
if tag == 'CMD' and bad_command:
tag = 'span'
attrs = {'class': ['bad-command']}
content = [('u', {}, [bad_command])] + content
elif tag == 'ERR':
tag = 'span'
attrs = {'class': ['err'], 'title': attrs['info'][0]}
content_str = handle_fragments(content,
indent=' ' + indent,
transclusions=transclusions,
h_shift=h_shift,
epub_clean=epub_clean,
bibliography=bibliography)
if tag in VOID_TAGS:
assert not content
template = "<%(tag)s%(attrs_str)s/>"
elif tag in INLINE:
template = "<%(tag)s%(attrs_str)s>%(content_str)s</%(tag)s>"
elif '\n' in content_str:
template = NOT_INLINE_TEMPLATE
else:
template = COMPACT_NOT_INLINE_TEMPLATE
# FIXME(alexander): disgusting hack; fix this properly and
# use a set representation to start with!
classes = attrs.get('class')
if classes:
attrs = attrs.copy()
attrs['class'] = sorted(set(classes))
return template % dict(
indent=indent,
tag=tag,
attrs_str=encode_attrs(attrs, transclusions, epub_clean),
content_str=content_str)