from shutil import rmtree
from librarian import RDFNS, WLNS, NCXNS, OPFNS, XHTMLNS, OutputFile
-from librarian.cover import WLCover
+from librarian.cover import ImageCover as WLCover
from librarian import functions, get_resource
return text.replace(u"\ufeff", u"")\
.replace("---", u"\u2014")\
.replace("--", u"\u2013")\
- .replace(",,", u"\u201E")\
+ .replace(",,", u"“")\
.replace('"', u"\u201D")\
.replace("'", u"\u2019")
if node.tag in ('uwaga', 'extra'):
last_node_part = False
for one_part in main_text:
name = one_part.tag
- if name == 'naglowek_czesc':
- yield part_xml
- last_node_part = True
- main_xml_part[:] = [deepcopy(one_part)]
- elif not last_node_part and name in ("naglowek_rozdzial", "naglowek_akt", "srodtytul"):
- yield part_xml
- main_xml_part[:] = [deepcopy(one_part)]
- else:
+ #if name == 'naglowek_czesc':
+ # yield part_xml
+ # last_node_part = True
+ # main_xml_part[:] = [deepcopy(one_part)]
+ #elif not last_node_part and name in ("naglowek_rozdzial", "naglowek_akt", "srodtytul"):
+ # yield part_xml
+ # main_xml_part[:] = [deepcopy(one_part)]
+ #else:
+ if True:
main_xml_part.append(deepcopy(one_part))
last_node_part = False
yield part_xml
""" transforms one chunk, returns a HTML string, a TOC object and a set of used characters """
toc = TOC()
- for element in chunk_xml[0]:
- if element.tag in ("naglowek_czesc", "naglowek_rozdzial", "naglowek_akt", "srodtytul"):
- toc.add(node_name(element), "part%d.html" % chunk_no)
- elif element.tag in ('naglowek_podrozdzial', 'naglowek_scena'):
- subnumber = toc.add(node_name(element), "part%d.html" % chunk_no, level=1, is_part=False)
- element.set('sub', str(subnumber))
+ #for element in chunk_xml[0]:
+ # if element.tag in ("naglowek_czesc", "naglowek_rozdzial", "naglowek_akt", "srodtytul"):
+ # toc.add(node_name(element), "part%d.html" % chunk_no)
+ # elif element.tag in ('naglowek_podrozdzial', 'naglowek_scena'):
+ # subnumber = toc.add(node_name(element), "part%d.html" % chunk_no, level=1, is_part=False)
+ # element.set('sub', str(subnumber))
if empty:
if not _empty_html_static:
_empty_html_static.append(open(get_resource('epub/emptyChunk.html')).read())
chars = set()
output_html = _empty_html_static[0]
else:
- find_annotations(annotations, chunk_xml, chunk_no)
- replace_by_verse(chunk_xml)
- html_tree = xslt(chunk_xml, get_resource('epub/xsltScheme.xsl'))
+ if chunk_no == 1:
+ html_tree = xslt(chunk_xml, get_resource('epub/xsltScheme-FoC.xsl'))
+ else:
+ find_annotations(annotations, chunk_xml, chunk_no)
+ replace_by_verse(chunk_xml)
+ html_tree = xslt(chunk_xml, get_resource('epub/xsltScheme.xsl'))
chars = used_chars(html_tree.getroot())
output_html = etree.tostring(html_tree, method="html", pretty_print=True)
return output_html, toc, chars
# every input file will have a TOC entry,
# pointing to starting chunk
- toc = TOC(wldoc.book_info.title, "part%d.html" % chunk_counter)
+
+ # hack for FoC:
+ if wldoc.book_info.author is not None:
+ toc_title = "%s, %s" % (wldoc.book_info.author.readable(), wldoc.book_info.title)
+ note = wldoc.edoc.find('//dzielo_nadrzedne')
+ if note is not None:
+ toc_title += " (%s)" % note.text
+ else:
+ toc_title = wldoc.book_info.title
+ toc = TOC(toc_title, "part%d.html" % chunk_counter)
chars = set()
if first:
# write book title page
zip.writestr('OPS/title.html',
etree.tostring(html_tree, method="html", pretty_print=True))
# add a title page TOC entry
- toc.add(u"Strona tytułowa", "title.html")
+ toc.add(u"Title page", "title.html")
+ toc.add(u"Dear readers!", "part1.html")
elif wldoc.book_info.parts:
# write title page for every parent
if sample is not None and sample <= 0:
'<rootfiles><rootfile full-path="OPS/content.opf" ' \
'media-type="application/oebps-package+xml" />' \
'</rootfiles></container>')
- zip.write(get_resource('res/wl-logo-small.png'), os.path.join('OPS', 'logo_wolnelektury.png'))
- zip.write(get_resource('res/jedenprocent.png'), os.path.join('OPS', 'jedenprocent.png'))
+ #zip.write(get_resource('res/wl-logo-small.png'), os.path.join('OPS', 'logo_wolnelektury.png'))
+ #zip.write(get_resource('res/jedenprocent.png'), os.path.join('OPS', 'jedenprocent.png'))
+ zip.write('logo.png', os.path.join('OPS', 'logo.png'))
if not style:
style = get_resource('epub/style.css')
zip.write(style, os.path.join('OPS', 'style.css'))
cover_file = StringIO()
c = cover(document.book_info)
+ import Image
+ c.im = Image.open('cover.jpg')
+ c.ext = lambda: 'jpg'
c.save(cover_file)
c_name = 'cover.%s' % c.ext()
zip.writestr(os.path.join('OPS', c_name), cover_file.getvalue())
'<item id="html_toc" href="toc.html" media-type="application/xhtml+xml" />'))
spine.append(etree.fromstring(
'<itemref idref="html_toc" />'))
- guide.append(etree.fromstring('<reference href="toc.html" type="toc" title="Spis treści"/>'))
+ guide.append(etree.fromstring('<reference href="toc.html" type="toc" title="Table of Contents"/>'))
toc, chunk_counter, chars, sample = transform_file(document, sample=sample)
zip.writestr('OPS/annotations.html', etree.tostring(
html_tree, method="html", pretty_print=True))
- toc.add("Strona redakcyjna", "last.html")
+ toc.add("Editorial page", "last.html")
manifest.append(etree.fromstring(
'<item id="last" href="last.html" media-type="application/xhtml+xml" />'))
spine.append(etree.fromstring(
# write TOC
if html_toc:
- toc.add(u"Spis treści", "toc.html", index=1)
+ toc.add(u"Table of Contents", "toc.html", index=1)
zip.writestr('OPS/toc.html', toc.html().encode('utf-8'))
toc.write_to_xml(nav_map)
zip.writestr('OPS/toc.ncx', etree.tostring(toc_file, pretty_print=True))