def parse_page_labels(page_labels: PdfArray, number_pages: int) -> List[str]:
page_numbers = []
# add the final stop position
page_labels.append(number_pages)
for i in range(0, len(page_labels) - 1, 2):
start, options, stop = page_labels[i:i + 3]
stop = int(stop)
start = int(start)
# /S specifies the numbering style for page numbers:
# /D - Arabic numerals (1,2,3...)
# /r - lowercase Roman numerals (i, ii, iii,...)
# /R - uppercase Roman numerals (I, II, III,...)
# /A - uppercase letters (A-Z)
# /a - lowercase letters (a-z)
# /P (optional) - page number prefix
# /St (optional) - the value of the first page number in the range (default: 1)
page_offset = int(options.St or 1)
page_range = range(page_offset, (stop - start) + 1)
option_mapping = {
"/D": str,
"/r": lambda x: to_roman(x).lower(),
"/R": to_roman,
"/a": ascii_lowercase.__getitem__,
"/A": ascii_uppercase.__getitem__,
}
range_numbers = map(option_mapping.get(options.S), page_range)
if options.P is not None and options.P != "()":
range_numbers = map(lambda x: options.P + x, range_numbers)
page_numbers.extend(range_numbers)
return page_numbers
def write_pdf_metadata(document, fileobj, scale, metadata, attachments,
url_fetcher):
"""Append to a seekable file-like object to add PDF metadata."""
fileobj.seek(0)
trailer = PdfReader(fileobj)
pages = trailer.Root.Pages.Kids
bookmarks, links = prepare_metadata(document, scale, pages)
if bookmarks:
bookmark_objects, count = create_bookmarks(bookmarks, pages)
trailer.Root.Outlines = PdfDict(Type=PdfName('Outlines'),
Count=count,
First=bookmark_objects[0],
Last=bookmark_objects[-1])
attachments = metadata.attachments + (attachments or [])
if attachments:
embedded_files = []
for attachment in attachments:
attachment_object = _create_pdf_attachment(attachment, url_fetcher)
if attachment_object is not None:
embedded_files.append(PdfString.encode('attachment'))
embedded_files.append(attachment_object)
if embedded_files:
trailer.Root.Names = PdfDict(EmbeddedFiles=PdfDict(
Names=PdfArray(embedded_files)))
# A single link can be split in multiple regions. We don't want to embedded
# a file multiple times of course, so keep a reference to every embedded
# URL and reuse the object number.
# TODO: If we add support for descriptions this won't always be correct,
# because two links might have the same href, but different titles.
annot_files = {}
for page_links in links:
for link_type, target, rectangle in page_links:
if link_type == 'attachment' and target not in annot_files:
# TODO: use the title attribute as description
annot_files[target] = _create_pdf_attachment((target, None),
url_fetcher)
# TODO: splitting a link into multiple independent rectangular annotations
# works well for pure links, but rather mediocre for other annotations and
# fails completely for transformed (CSS) or complex link shapes (area).
# It would be better to use /AP for all links and coalesce link shapes that
# originate from the same HTML link. This would give a feeling similiar to
# what browsers do with links that span multiple lines.
for page, page_links in zip(pages, links):
annotations = PdfArray()
for link_type, target, rectangle in page_links:
if link_type != 'attachment' or annot_files[target] is None:
annotation = PdfDict(Type=PdfName('Annot'),
Subtype=PdfName('Link'),
Rect=PdfArray(rectangle),
Border=PdfArray((0, 0, 0)))
if link_type == 'internal':
destination = (target[0], PdfName('XYZ'), target[1],
target[2], 0)
annotation.A = PdfDict(Type=PdfName('Action'),
S=PdfName('GoTo'),
D=PdfArray(destination))
else:
annotation.A = PdfDict(Type=PdfName('Action'),
S=PdfName('URI'),
URI=PdfString.encode(
iri_to_uri(target)))
else:
assert annot_files[target] is not None
ap = PdfDict(N=PdfDict(BBox=PdfArray(rectangle),
Subtype=PdfName('Form'),
Type=PdfName('XObject')))
# evince needs /T or fails on an internal assertion. PDF
# doesn't require it.
annotation = PdfDict(Type=PdfName('Annot'),
Subtype=PdfName('FileAttachment'),
T=PdfString.encode(''),
Rect=PdfArray(rectangle),
Border=PdfArray((0, 0, 0)),
FS=annot_files[target],
AP=ap)
annotations.append(annotation)
if annotations:
page.Annots = annotations
trailer.Info.Producer = VERSION_STRING
for attr, key in (('title', 'Title'), ('description', 'Subject'),
('generator', 'Creator')):
value = getattr(metadata, attr)
if value is not None:
setattr(trailer.Info, key, value)
for attr, key in (('authors', 'Author'), ('keywords', 'Keywords')):
value = getattr(metadata, attr)
if value is not None:
setattr(trailer.Info, key, ', '.join(getattr(metadata, attr)))
for attr, key in (('created', 'CreationDate'), ('modified', 'ModDate')):
value = w3c_date_to_pdf(getattr(metadata, attr), attr)
if value is not None:
setattr(trailer.Info, key, value)
for page, document_page in zip(pages, document.pages):
left, top, right, bottom = (float(value) for value in page.MediaBox)
# Convert pixels into points
bleed = {
key: value * 0.75
for key, value in document_page.bleed.items()
}
trim_left = left + bleed['left']
trim_top = top + bleed['top']
trim_right = right - bleed['right']
trim_bottom = bottom - bleed['bottom']
page.TrimBox = PdfArray((trim_left, trim_top, trim_right, trim_bottom))
# Arbitrarly set PDF BleedBox between CSS bleed box (PDF MediaBox) and
# CSS page box (PDF TrimBox), at most 10 points from the TrimBox.
bleed_left = trim_left - min(10, bleed['left'])
bleed_top = trim_top - min(10, bleed['top'])
bleed_right = trim_right + min(10, bleed['right'])
bleed_bottom = trim_bottom + min(10, bleed['bottom'])
page.BleedBox = PdfArray(
(bleed_left, bleed_top, bleed_right, bleed_bottom))
fileobj.seek(0)
PdfWriter().write(fileobj, trailer=trailer)
fileobj.truncate()
def do_apply_ocg(basepage, rmpage, i, uses_base_pdf, ocgprop, annotations):
ocgpage = IndirectPdfDict(Type=PdfName('OCG'), Name='Page ' + str(i + 1))
ocgprop.OCGs.append(ocgpage)
# The Order dict is a Page, followed by Inner
ocgorderinner = PdfArray()
# Add Template OCG layer
# If this uses a basepdf, the template is located
# elsewhere.
# If using a basepdf, assign its stream as a
# 'Background' layer under this page. When the page
# primary OCG is disabled, the background will
# remain, making it easy to disable all annotations.
if uses_base_pdf:
ocgorigdoc = IndirectPdfDict(Type=PdfName('OCG'), Name='Background')
ocgprop.OCGs.append(ocgorigdoc)
ocgorderinner.append(ocgorigdoc)
uncompress.uncompress([basepage.Contents])
stream = basepage.Contents.stream
stream = '/OC /ocgorigdoc BDC\n' \
+ stream \
+ 'EMC\n'
basepage.Contents.stream = stream
compress.compress([basepage.Contents])
if '/Properties' in basepage.Resources:
props = basepage.Resources.Properties
else:
props = PdfDict()
props.ocgorigdoc = ocgorigdoc
basepage.Resources.Properties = props
# If not using a basepdf, assign the rmpage's stream
# as a 'Template' layer under this page. It will be
# affected by disabling the primary Page OCG (which
# by itself is kind of useless for exported
# notebooks).
# Regardless of using a basepdf or not, put the
# rmpage layers into their own OCGs.
# If the template has an XObject, we want to skip
# the first one. This happens when the template
# contains a PNG. Question--what happens when the
# template contains more than one PNG? How do we
# detect all of those?
template_xobj_keys = []
vector_layers = []
uncompress.uncompress([rmpage.Contents])
if uses_base_pdf:
# The entire thing is the page ocg
stream = '/OC /ocgpage BDC\n'
stream += rmpage.Contents.stream
stream += 'EMC\n'
rmpage.Contents.stream = stream
else:
stream = rmpage.Contents.stream
# Mark the template ocg separate from page ocg
template_endpos = 0
page_inatpos = 0
findkey = '1 w 2 J 2 j []0 d\nq\n'
# Finds only the first instance, which should be
# for the template.
findloc = stream.find(findkey)
if findloc < 0:
# May be a vector, which we stick a marker
# in for.
# ?? Why is this a half-point off ??
findkey = '799.500000 85 l\n'
m = re.search(findkey, rmpage.Contents.stream)
if m:
findloc = m.start()
if findloc > 0:
template_endpos = findloc + len(findkey)
# Add vector template OCG
stream = '/OC /ocgtemplate BDC\n'
stream += rmpage.Contents.stream[:template_endpos]
stream += 'EMC\n'
page_inatpos = len(stream)
stream += rmpage.Contents.stream[template_endpos:]
# Save stream
rmpage.Contents.stream = stream
# Add template ocg
ocgtemplate = IndirectPdfDict(Type=PdfName('OCG'), Name='Template')
ocgprop.OCGs.append(ocgtemplate)
ocgorderinner.append(ocgtemplate)
# If a template (which is SVG) has embedded PNG
# images, those appear as XObjects. This will
# mess up the layer order, so we will ignore
# them later.
template_xobj_keys = \
re.findall(r'(\/Im[0-9]+)\s',
stream[:template_endpos])
# Page ocg
stream = rmpage.Contents.stream[:page_inatpos]
stream += '/OC /ocgpage BDC\n'
stream += rmpage.Contents.stream[page_inatpos:]
stream += 'EMC\n'
# Save stream
rmpage.Contents.stream = stream
# Find all other vector layers using the magic
# point (DocumentPageLayer.render_to_painter()).
# ?? Why is this a half-point off ??
while True:
m = re.search('420.500000 69 m\n', rmpage.Contents.stream)
if not m:
break
stream = ''
layerid = 'ocglayer{}'.format(len(vector_layers) + 1)
stream = rmpage.Contents.stream[:m.start()]
if len(vector_layers):
# close previous layer
stream += 'EMC\n'
stream += '/OC /{} BDC\n'.format(layerid)
stream += rmpage.Contents.stream[m.end():]
vector_layers.append(layerid)
rmpage.Contents.stream = stream
# If we added vector layers, have to end the
# first one.
if len(vector_layers):
stream = rmpage.Contents.stream + 'EMC\n'
rmpage.Contents.stream = stream
# Done--recompress the stream.
compress.compress([rmpage.Contents])
# There shouldn't be any Properties there since we
# generated the rmpage ourselves, so don't bother
# checking.
rmpage.Resources.Properties = PdfDict(ocgpage=ocgpage)
if not uses_base_pdf:
rmpage.Resources.Properties.ocgtemplate = ocgtemplate
# Add individual OCG layers (Bitmap)
was_vector = True
for n, key in enumerate(rmpage.Resources.XObject):
if str(key) in template_xobj_keys:
continue
was_vector = False
l = n - len(template_xobj_keys)
# This would indicate a bug in the handling of a
# notebook.
try:
layer = annotations[i][l]
except:
log.error(
'could not associate XObject with layer: (i, l) ({}, {})'.
format(i, l))
log.error(str(annotations))
log.error('document: {} ()').format('uuid', 'self.visible_name')
continue
layername = layer[0]
ocg = IndirectPdfDict(Type=PdfName('OCG'), Name=layername)
ocgprop.OCGs.append(ocg)
ocgorderinner.append(ocg)
rmpage.Resources.XObject[key].OC = ocg
# Add individual OCG layers (Vector)
if was_vector:
for l, layerid in enumerate(vector_layers):
# This would indicate a bug in the handling of a
# notebook.
try:
layer = annotations[i][l]
except:
log.error(
'could not associate layerid with layer: (i, l, layerid) ({}, {}, {})'
.format(i, l, layerid))
log.error('document: {} ()').format('uuid',
'self.visible_name')
log.error(str(annotations))
continue
layername = layer[0]
ocg = IndirectPdfDict(Type=PdfName('OCG'), Name=layername)
ocgprop.OCGs.append(ocg)
ocgorderinner.append(ocg)
rmpage.Resources.Properties[PdfName(layerid)] = \
ocg
# Add order of OCGs to primary document
ocgprop.D.Order.append(ocgpage)
ocgprop.D.Order.append(ocgorderinner)
return ocgorderinner
