2012-05-19 16:50:38 +04:00
|
|
|
|
# coding: utf8
|
|
|
|
|
r"""
|
|
|
|
|
weasyprint.pdf
|
|
|
|
|
--------------
|
|
|
|
|
|
|
|
|
|
Post-process the PDF files created by cairo and add metadata such as
|
|
|
|
|
hyperlinks and bookmarks.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Rather than trying to parse any valid PDF, we make some assumptions
|
|
|
|
|
that hold for cairo in order to simplify the code:
|
|
|
|
|
|
|
|
|
|
* All newlines are '\n', not '\r' or '\r\n'
|
|
|
|
|
* Except for number 0 (which is always free) there is no "free" object.
|
|
|
|
|
* Most white space separators are made of a single 0x20 space.
|
|
|
|
|
* Indirect dictionary objects do not contain '>>' at the start of a line
|
|
|
|
|
except to mark the end of the object, followed by 'endobj'.
|
|
|
|
|
(In other words, '>>' markers for sub-dictionaries are indented.)
|
|
|
|
|
* The Page Tree is flat: all kids of the root page node are page objects,
|
|
|
|
|
not page tree nodes.
|
|
|
|
|
|
|
|
|
|
However the code uses a lot of assert statements so that if an assumptions
|
|
|
|
|
is not true anymore, the code should (hopefully) fail with an exception
|
|
|
|
|
rather than silently behave incorrectly.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
:copyright: Copyright 2011-2012 Simon Sapin and contributors, see AUTHORS.
|
|
|
|
|
:license: BSD, see LICENSE for details.
|
|
|
|
|
|
|
|
|
|
"""
|
|
|
|
|
|
|
|
|
|
from __future__ import division, unicode_literals
|
2012-05-20 15:34:42 +04:00
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
import os
|
|
|
|
|
import re
|
2012-05-20 15:34:42 +04:00
|
|
|
|
import string
|
2012-05-19 16:50:38 +04:00
|
|
|
|
import itertools
|
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
from . import VERSION_STRING
|
|
|
|
|
from .compat import xrange, iteritems
|
2012-05-20 17:55:57 +04:00
|
|
|
|
from .utils import iri_to_uri
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
|
|
|
|
|
2012-05-20 15:34:42 +04:00
|
|
|
|
class PDFFormatter(string.Formatter):
|
|
|
|
|
"""Like str.format except:
|
|
|
|
|
|
|
|
|
|
* Results are byte strings
|
|
|
|
|
* The new !P conversion flags encodes a PDF string.
|
|
|
|
|
(UTF-16 BE with a BOM, then backslash-escape parentheses.)
|
|
|
|
|
|
|
|
|
|
Except for fields marked !P, everything should be ASCII-only.
|
|
|
|
|
|
|
|
|
|
"""
|
|
|
|
|
def convert_field(self, value, conversion):
|
|
|
|
|
if conversion == 'P':
|
|
|
|
|
# Make a round-trip back through Unicode for the .translate()
|
|
|
|
|
# method. (bytes.translate only maps to single bytes.)
|
|
|
|
|
# Use latin1 to map all byte values.
|
|
|
|
|
return '({0})'.format(
|
|
|
|
|
('\ufeff' + value).encode('utf-16-be').decode('latin1')
|
|
|
|
|
.translate({40: r'\(', 41: r'\)', 92: r'\\'}))
|
|
|
|
|
else:
|
|
|
|
|
return super(PDFFormatter, self).convert_field(value, conversion)
|
|
|
|
|
|
|
|
|
|
def vformat(self, format_string, args, kwargs):
|
|
|
|
|
result = super(PDFFormatter, self).vformat(format_string, args, kwargs)
|
|
|
|
|
return result.encode('latin1')
|
|
|
|
|
|
|
|
|
|
pdf_format = PDFFormatter().format
|
|
|
|
|
|
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
class PDFDictionary(object):
|
|
|
|
|
def __init__(self, object_number, byte_string):
|
|
|
|
|
self.object_number = object_number
|
|
|
|
|
self.byte_string = byte_string
|
|
|
|
|
|
|
|
|
|
def __repr__(self):
|
|
|
|
|
return self.__class__.__name__ + repr(
|
|
|
|
|
(self.object_number, self.byte_string))
|
|
|
|
|
|
2012-05-20 15:00:39 +04:00
|
|
|
|
_re_cache = {}
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
2012-05-20 15:00:39 +04:00
|
|
|
|
def _get_value(self, key, value_re):
|
|
|
|
|
regex = self._re_cache.get((key, value_re))
|
2012-05-19 16:50:38 +04:00
|
|
|
|
if not regex:
|
2012-05-20 15:34:42 +04:00
|
|
|
|
regex = re.compile(pdf_format('/{0} {1}', key, value_re))
|
2012-05-20 15:00:39 +04:00
|
|
|
|
self._re_cache[key, value_re] = regex
|
|
|
|
|
return regex.search(self.byte_string).group(1)
|
2012-05-20 14:46:33 +04:00
|
|
|
|
|
2012-05-20 15:00:39 +04:00
|
|
|
|
def get_type(self):
|
|
|
|
|
"""
|
|
|
|
|
:returns: the value for the /Type key.
|
2012-05-20 14:46:33 +04:00
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
"""
|
2012-05-20 15:00:39 +04:00
|
|
|
|
# No end delimiter, + defaults to greedy
|
|
|
|
|
return self._get_value('Type', '/(\w+)').decode('ascii')
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
def get_indirect_dict(self, key, pdf_file):
|
|
|
|
|
"""Read the value for `key` and follow the reference, assuming
|
|
|
|
|
it is an indirect dictionary object.
|
|
|
|
|
|
|
|
|
|
:return: a new PDFDictionary instance.
|
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
"""
|
2012-05-20 15:00:39 +04:00
|
|
|
|
object_number = int(self._get_value(key, '(\d+) 0 R'))
|
2012-05-20 14:46:33 +04:00
|
|
|
|
return type(self)(object_number, pdf_file.read_object(object_number))
|
|
|
|
|
|
|
|
|
|
def get_indirect_dict_array(self, key, pdf_file):
|
|
|
|
|
"""Read the value for `key` and follow the references, assuming
|
|
|
|
|
it is an array of indirect dictionary objects.
|
|
|
|
|
|
|
|
|
|
:return: a list of new PDFDictionary instance.
|
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
"""
|
2012-05-20 15:00:39 +04:00
|
|
|
|
parts = self._get_value(key, '\[([^\]]+)\]').split(b' 0 R')
|
2012-05-20 14:46:33 +04:00
|
|
|
|
# The array looks like this: ' <a> 0 R <b> 0 R <c> 0 R '
|
|
|
|
|
# so `parts` ends up like this [' <a>', ' <b>', ' <c>', ' ']
|
|
|
|
|
# With the trailing white space in the list.
|
2012-05-19 16:50:38 +04:00
|
|
|
|
trail = parts.pop()
|
|
|
|
|
assert not trail.strip()
|
2012-05-20 14:46:33 +04:00
|
|
|
|
class_ = type(self)
|
|
|
|
|
read = pdf_file.read_object
|
|
|
|
|
return [class_(n, read(n)) for n in map(int, parts)]
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class PDFFile(object):
|
|
|
|
|
"""
|
|
|
|
|
:param fileobj:
|
|
|
|
|
A seekable binary file-like object for a PDF generated by cairo.
|
2012-05-20 14:46:33 +04:00
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
"""
|
2012-05-20 15:00:39 +04:00
|
|
|
|
trailer_re = re.compile(
|
|
|
|
|
b'\ntrailer\n(.+)\nstartxref\n(\d+)\n%%EOF\n$', re.DOTALL)
|
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
def __init__(self, fileobj):
|
2012-05-20 14:46:33 +04:00
|
|
|
|
# cairo’s trailer only has Size, Root and Info.
|
|
|
|
|
# The trailer + startxref + EOF is typically under 100 bytes
|
2012-05-19 16:50:38 +04:00
|
|
|
|
fileobj.seek(-200, os.SEEK_END)
|
2012-05-20 15:00:39 +04:00
|
|
|
|
trailer, startxref = self.trailer_re.search(fileobj.read()).groups()
|
2012-05-19 16:50:38 +04:00
|
|
|
|
trailer = PDFDictionary(None, trailer)
|
2012-05-20 14:46:33 +04:00
|
|
|
|
startxref = int(startxref)
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
fileobj.seek(startxref)
|
2012-05-19 16:50:38 +04:00
|
|
|
|
line = next(fileobj)
|
|
|
|
|
assert line == b'xref\n'
|
|
|
|
|
|
|
|
|
|
line = next(fileobj)
|
|
|
|
|
first_object, total_objects = line.split()
|
|
|
|
|
assert first_object == b'0'
|
|
|
|
|
total_objects = int(total_objects)
|
|
|
|
|
|
|
|
|
|
line = next(fileobj)
|
|
|
|
|
assert line == b'0000000000 65535 f \n'
|
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
objects_offsets = [None]
|
2012-05-19 16:50:38 +04:00
|
|
|
|
for object_number in xrange(1, total_objects):
|
|
|
|
|
line = next(fileobj)
|
|
|
|
|
assert line[10:] == b' 00000 n \n'
|
2012-05-20 14:46:33 +04:00
|
|
|
|
objects_offsets.append(int(line[:10]))
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
|
|
|
|
self.fileobj = fileobj
|
|
|
|
|
#: Maps object number -> bytes from the start of the file
|
2012-05-20 14:46:33 +04:00
|
|
|
|
self.objects_offsets = objects_offsets
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
|
|
|
|
info = trailer.get_indirect_dict('Info', self)
|
|
|
|
|
catalog = trailer.get_indirect_dict('Root', self)
|
|
|
|
|
page_tree = catalog.get_indirect_dict('Pages', self)
|
|
|
|
|
pages = page_tree.get_indirect_dict_array('Kids', self)
|
|
|
|
|
# Check that the tree is flat
|
|
|
|
|
assert all(p.get_type() == 'Page' for p in pages)
|
|
|
|
|
|
|
|
|
|
self.startxref = startxref
|
|
|
|
|
self.info = info
|
|
|
|
|
self.catalog = catalog
|
|
|
|
|
self.page_tree = page_tree
|
|
|
|
|
self.pages = pages
|
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
self.finished = False
|
|
|
|
|
self.overwritten_objects_offsets = {}
|
|
|
|
|
self.new_objects_offsets = []
|
|
|
|
|
|
|
|
|
|
def read_object(self, object_number):
|
|
|
|
|
"""
|
|
|
|
|
:param object_number:
|
|
|
|
|
An integer N so that 1 <= N < len(self.objects_offsets)
|
|
|
|
|
:returns:
|
|
|
|
|
The object content as a byte string.
|
|
|
|
|
|
|
|
|
|
"""
|
|
|
|
|
fileobj = self.fileobj
|
|
|
|
|
fileobj.seek(self.objects_offsets[object_number])
|
|
|
|
|
line = next(fileobj)
|
|
|
|
|
assert line.endswith(b' 0 obj\n')
|
|
|
|
|
assert int(line[:-7]) == object_number # len(b' 0 obj\n') == 7
|
|
|
|
|
object_lines = []
|
|
|
|
|
for line in fileobj:
|
|
|
|
|
object_lines.append(line)
|
|
|
|
|
if line == b'>>\n':
|
|
|
|
|
assert next(fileobj) == b'endobj\n'
|
|
|
|
|
return b''.join(object_lines)
|
|
|
|
|
|
|
|
|
|
def overwrite_object(self, object_number, byte_string):
|
|
|
|
|
"""Write the new content for an existing object at the end of the file.
|
|
|
|
|
|
|
|
|
|
:param object_number:
|
|
|
|
|
An integer N so that 1 <= N < len(self.objects_offsets)
|
|
|
|
|
:param byte_string:
|
|
|
|
|
The new object content as a byte string.
|
|
|
|
|
|
|
|
|
|
"""
|
|
|
|
|
self.overwritten_objects_offsets[object_number] = (
|
|
|
|
|
self._write_object(object_number, byte_string))
|
|
|
|
|
|
2012-05-20 17:55:57 +04:00
|
|
|
|
def extend_dict(self, dictionary, new_content):
|
|
|
|
|
"""Overwrite a dictionary object after adding content inside
|
|
|
|
|
the << >> delimiters.
|
|
|
|
|
|
|
|
|
|
"""
|
|
|
|
|
assert dictionary.byte_string.endswith(b'>>\n')
|
|
|
|
|
self.overwrite_object(
|
|
|
|
|
dictionary.object_number,
|
|
|
|
|
dictionary.byte_string[:-3] + new_content + b'\n>>\n')
|
|
|
|
|
|
|
|
|
|
def next_object_number(self):
|
|
|
|
|
"""Return the object number that would be used by write_new_object().
|
|
|
|
|
"""
|
|
|
|
|
return len(self.objects_offsets) + len(self.new_objects_offsets)
|
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
def write_new_object(self, byte_string):
|
|
|
|
|
"""Write a new object at the end of the file.
|
|
|
|
|
|
|
|
|
|
:param byte_string:
|
|
|
|
|
The object content as a byte string.
|
|
|
|
|
:return:
|
|
|
|
|
The new object number.
|
|
|
|
|
|
|
|
|
|
"""
|
2012-05-20 17:55:57 +04:00
|
|
|
|
object_number = self.next_object_number()
|
|
|
|
|
self.new_objects_offsets.append(
|
2012-05-20 14:46:33 +04:00
|
|
|
|
self._write_object(object_number, byte_string))
|
|
|
|
|
return object_number
|
|
|
|
|
|
|
|
|
|
def finish(self):
|
|
|
|
|
"""
|
|
|
|
|
Write the cross-reference table and the trailer for the new and
|
|
|
|
|
overwritten objects. This makes `fileobj` a valid (updated) PDF file.
|
|
|
|
|
|
|
|
|
|
"""
|
|
|
|
|
new_startxref, write = self._start_writing()
|
|
|
|
|
self.finished = True
|
|
|
|
|
write(b'xref\n')
|
|
|
|
|
|
|
|
|
|
# Don’t bother sorting or finding contiguous numbers,
|
|
|
|
|
# just write a new sub-section for each overwritten object.
|
|
|
|
|
for object_number, offset in iteritems(
|
|
|
|
|
self.overwritten_objects_offsets):
|
2012-05-20 15:34:42 +04:00
|
|
|
|
write(pdf_format(
|
|
|
|
|
'{0} 1\n{1:010} 00000 n \n', object_number, offset))
|
2012-05-20 14:46:33 +04:00
|
|
|
|
|
|
|
|
|
if self.new_objects_offsets:
|
|
|
|
|
first_new_object = len(self.objects_offsets)
|
2012-05-20 15:34:42 +04:00
|
|
|
|
write(pdf_format(
|
|
|
|
|
'{0} {1}\n', first_new_object, len(self.new_objects_offsets)))
|
2012-05-20 14:46:33 +04:00
|
|
|
|
for object_number, offset in enumerate(
|
|
|
|
|
self.new_objects_offsets, start=first_new_object):
|
2012-05-20 15:34:42 +04:00
|
|
|
|
write(pdf_format('{0:010} 00000 n \n', offset))
|
|
|
|
|
|
|
|
|
|
write(pdf_format(
|
|
|
|
|
'trailer\n<< '
|
|
|
|
|
'/Size {size} /Root {root} 0 R /Info {info} 0 R /Prev {prev}'
|
|
|
|
|
' >>\nstartxref\n{startxref}\n%%EOF\n',
|
2012-05-20 17:55:57 +04:00
|
|
|
|
size=self.next_object_number(),
|
2012-05-20 15:34:42 +04:00
|
|
|
|
root=self.catalog.object_number,
|
|
|
|
|
info=self.info.object_number,
|
|
|
|
|
prev=self.startxref,
|
|
|
|
|
startxref=new_startxref))
|
2012-05-20 14:46:33 +04:00
|
|
|
|
|
|
|
|
|
def _write_object(self, object_number, byte_string):
|
|
|
|
|
offset, write = self._start_writing()
|
2012-05-20 15:34:42 +04:00
|
|
|
|
write(pdf_format('{0} 0 obj\n', object_number))
|
2012-05-20 14:46:33 +04:00
|
|
|
|
write(byte_string)
|
|
|
|
|
write(b'\nendobj\n')
|
|
|
|
|
return offset
|
|
|
|
|
|
|
|
|
|
def _start_writing(self):
|
|
|
|
|
assert not self.finished
|
|
|
|
|
fileobj = self.fileobj
|
|
|
|
|
fileobj.seek(0, os.SEEK_END)
|
|
|
|
|
return fileobj.tell(), fileobj.write
|
|
|
|
|
|
|
|
|
|
|
2012-05-20 17:55:57 +04:00
|
|
|
|
def add_pdf_metadata(fileobj, links, destinations, bookmarks):
|
2012-05-20 14:46:33 +04:00
|
|
|
|
pdf = PDFFile(fileobj)
|
2012-05-20 15:34:42 +04:00
|
|
|
|
pdf.overwrite_object(pdf.info.object_number, pdf_format(
|
|
|
|
|
'<< /Producer {producer!P} >>',
|
|
|
|
|
producer=VERSION_STRING))
|
2012-05-20 17:55:57 +04:00
|
|
|
|
|
|
|
|
|
root, bookmarks = bookmarks
|
|
|
|
|
if bookmarks:
|
|
|
|
|
bookmark_root = pdf.next_object_number()
|
|
|
|
|
pdf.write_new_object(pdf_format(
|
|
|
|
|
'<< /Type /Outlines /Count {0} /First {1} 0 R /Last {2} 0 R\n>>',
|
|
|
|
|
root['Count'],
|
|
|
|
|
root['First'] + bookmark_root,
|
|
|
|
|
root['Last'] + bookmark_root))
|
|
|
|
|
pdf.extend_dict(pdf.catalog, pdf_format(
|
|
|
|
|
'/Outlines {0} 0 R', bookmark_root))
|
|
|
|
|
for bookmark in bookmarks:
|
|
|
|
|
content = [pdf_format('<< /Title {0!P}\n', bookmark['label'])]
|
|
|
|
|
if bookmark['Count']:
|
|
|
|
|
content.append(pdf_format('/Count {0}\n', bookmark['Count']))
|
|
|
|
|
for key in ['Parent', 'Prev', 'Next', 'First', 'Last']:
|
|
|
|
|
if bookmark[key]:
|
|
|
|
|
content.append(pdf_format(
|
|
|
|
|
'/{0} {1} 0 R\n', key, bookmark[key] + bookmark_root))
|
|
|
|
|
content.append(pdf_format(
|
|
|
|
|
'/A << /Type /Action /S /GoTo '
|
|
|
|
|
'/D [{0} /XYZ {1:f} {2:f} 0] >>\n>>',
|
|
|
|
|
*bookmark['destination']))
|
|
|
|
|
pdf.write_new_object(b''.join(content))
|
|
|
|
|
|
|
|
|
|
for page, page_links in zip(pdf.pages, links):
|
|
|
|
|
annotations = []
|
|
|
|
|
for uri, x1, y1, x2, y2 in page_links:
|
|
|
|
|
content = [pdf_format(
|
|
|
|
|
'<< /Type /Annot /Subtype /Link '
|
|
|
|
|
'/Rect [{0:f} {1:f} {2:f} {3:f}] /Border [0 0 0]\n',
|
|
|
|
|
x1, y1, x2, y2)]
|
|
|
|
|
if uri and uri.startswith('#') and uri[1:] in destinations:
|
|
|
|
|
content.append(pdf_format(
|
|
|
|
|
'/A << /Type /Action /S /GoTo '
|
|
|
|
|
'/D [{0} /XYZ {1:f} {2:f} 0] >>\n',
|
|
|
|
|
*destinations[uri[1:]]))
|
|
|
|
|
elif uri:
|
|
|
|
|
content.append(pdf_format(
|
|
|
|
|
'/A << /Type /Action /S /URI /URI ({0}) >>\n',
|
|
|
|
|
iri_to_uri(uri)))
|
|
|
|
|
content.append(b'>>')
|
|
|
|
|
annotations.append(pdf.write_new_object(b''.join(content)))
|
|
|
|
|
|
|
|
|
|
if annotations:
|
|
|
|
|
pdf.extend_dict(page, pdf_format(
|
|
|
|
|
'/Annots [{0}]', ' '.join(
|
|
|
|
|
'{0} 0 R'.format(n) for n in annotations)))
|
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
pdf.finish()
|
|
|
|
|
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
|
|
|
|
def test():
|
|
|
|
|
import cairo
|
|
|
|
|
import io
|
|
|
|
|
fileobj = io.BytesIO()
|
|
|
|
|
surface = cairo.PDFSurface(fileobj, 100, 100)
|
2012-05-20 14:46:33 +04:00
|
|
|
|
# for i in xrange(20):
|
|
|
|
|
# surface.show_page()
|
2012-05-19 16:50:38 +04:00
|
|
|
|
surface.finish()
|
2012-05-20 14:46:33 +04:00
|
|
|
|
add_pdf_metadata(fileobj)
|
2012-05-20 15:34:42 +04:00
|
|
|
|
print(fileobj.getvalue().decode('latin1'))
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
2012-05-20 14:46:33 +04:00
|
|
|
|
# pdf = PDFFile(fileobj)
|
|
|
|
|
# print(pdf.page_tree)
|
|
|
|
|
# print(len(pdf.pages))
|
2012-05-19 16:50:38 +04:00
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
|
test()
|