def __str__(self): # __str__ is used for serialization if self.value == self.original_value: # If unchanged, return the raw original value without decoding/encoding. return PdfString.from_bytes(self.raw_original_value) else: # If the value changed, encode it from Unicode according to the encoding # of the font that is active at the location of this token. return PdfString.from_bytes(fromUnicode(self.value, self.font, fontcache, options))
def update_annotation(annotation, options):
import re
from pdfrw.objects import PdfString
# Contents holds a plain-text representation of the annotation
# content, such as for accessibility. All annotation types may
# have a Contents. NM holds the "annotation name" which also
# could have redactable text, I suppose. Markup annotations have
# "T" fields that hold a title / text label. Subj holds a
# comment subject. CA, RC, and AC are used in widget annotations.
for string_field in ("Contents", "NM", "T", "Subj", "CA", "RC", "AC"):
if getattr(annotation, string_field):
value = getattr(annotation, string_field).to_unicode()
for pattern, function in options.content_filters:
value = pattern.sub(function, value)
setattr(annotation, string_field, PdfString.from_unicode(value))
# A rich-text stream. Not implemented. Bail so that we don't
# accidentally leak something that should be redacted.
if annotation.RC:
raise ValueError(
"Annotation rich-text streams (Annot/RC) are not supported.")
# An action, usually used for links.
if annotation.A:
update_annotation_action(annotation, annotation.A, options)
if annotation.PA:
update_annotation_action(annotation, annotation.PA, options)
# If set, another annotation.
if annotation.Popup:
update_annotation(annotation.Popup, options)
def update_annotation_action(annotation, action, options):
from pdfrw.objects import PdfString
if action.URI and options.link_filters:
value = action.URI.to_unicode()
for func in options.link_filters:
value = func(value, annotation)
if value is None:
# Remove annotation by supressing the action.
action.URI = None
else:
action.URI = PdfString.from_unicode(value)
if action.Next:
# May be an Action or array of Actions to execute next.
next_action = action.Next
if isinstance(action.Next, dict):
next_action = [action.Next]
for a in next_action:
update_annotation_action(annotation, a, options)
def update_metadata(trailer, options):
# Update the PDF's Document Information Dictionary, which contains keys like
# Title, Author, Subject, Keywords, Creator, Producer, CreationDate, and ModDate
# (the latter two containing Date values, the rest strings).
import codecs
from pdfrw.objects import PdfString, PdfName
# Create the metadata dict if it doesn't exist, since the caller may be adding fields.
if not trailer.Info:
trailer.Info = PdfDict()
# Get a list of all metadata fields that exist in the PDF plus any fields
# that there are metadata filters for (since they may insert field values).
keys = set(str(k)[1:] for k in trailer.Info.keys()) \
| set(k for k in options.metadata_filters.keys() if k not in ("DEFAULT", "ALL"))
# Update each metadata field.
for key in keys:
# Get the functions to apply to this field.
functions = options.metadata_filters.get(key)
if functions is None:
# If nothing is defined for this field, use the DEFAULT functions.
functions = options.metadata_filters.get("DEFAULT", [])
# Append the ALL functions.
functions += options.metadata_filters.get("ALL", [])
# Run the functions on any existing values.
value = trailer.Info[PdfName(key)]
for f in functions:
# Before passing to the function, convert from a PdfString to a Python string.
if isinstance(value, PdfString):
# decode from PDF's "(...)" syntax.
value = value.decode()
# Filter the value.
value = f(value)
# Convert Python data type to PdfString.
if isinstance(value, str) or (sys.version_info < (3, )
and isinstance(value, unicode)):
# Convert string to a PdfString instance.
value = PdfString.from_unicode(value)
elif isinstance(value, datetime):
# Convert datetime into a PDF "D" string format.
value = value.strftime("%Y%m%d%H%M%S%z")
if len(value) == 19:
# If TZ info was included, add an apostrophe between the hour/minutes offsets.
value = value[:17] + "'" + value[17:]
value = PdfString("(D:%s)" % value)
elif value is None:
# delete the metadata value
pass
else:
raise ValueError(
"Invalid type of value returned by metadata_filter function. %s was returned by %s."
% (repr(value), f.__name__ or "anonymous function"))
# Replace value.
trailer.Info[PdfName(key)] = value
def _build_font():
with open(FONT_FILENAME, "rb") as f:
embedded_font_stream = f.read()
embedded_font = PdfDict()
embedded_font.indirect = True
embedded_font.Filter = [PdfName.FlateDecode]
embedded_font.stream = zlib.compress(embedded_font_stream, 9).decode(
"latin-1")
embedded_font.Length1 = len(embedded_font_stream)
font_descriptor = PdfDict()
font_descriptor.indirect = True
font_descriptor.Ascent = 1000
font_descriptor.CapHeight = 1000
font_descriptor.Descent = -1
font_descriptor.Flags = 5 # FixedPitch + Symbolic
font_descriptor.FontBBox = PdfArray([0, 0, 1000, 500])
font_descriptor.FontFile2 = embedded_font
font_descriptor.FontName = PdfName.GlyphLessFont
font_descriptor.ItalicAngle = 0
font_descriptor.StemV = 80
font_descriptor.Type = PdfName.FontDescriptor
# Map everything to glyph 1
cid_to_gid_map_stream = b"\0\1" * (1 << 16)
cid_to_gid_map = PdfDict()
cid_to_gid_map.indirect = True
cid_to_gid_map.Filter = [PdfName.FlateDecode]
cid_to_gid_map.stream = zlib.compress(
cid_to_gid_map_stream, 9).decode("latin-1")
cid_to_gid_map.Length1 = len(cid_to_gid_map_stream)
cid_system_info = PdfDict()
cid_system_info.Ordering = PdfString.from_unicode("Identity")
cid_system_info.Registry = PdfString.from_unicode("Adobe")
cid_system_info.Supplement = 0
cid_font = PdfDict()
cid_font.indirect = True
cid_font.CIDToGIDMap = cid_to_gid_map
cid_font.BaseFont = PdfName.GlyphLessFont
cid_font.CIDSystemInfo = cid_system_info
cid_font.FontDescriptor = font_descriptor
cid_font.Subtype = PdfName.CIDFontType2
cid_font.Type = PdfName.Font
cid_font.DW = 500
with open(UNICODE_CMAP_FILENAME, "rb") as f:
unicode_cmap_stream = f.read()
unicode_cmap = PdfDict()
unicode_cmap.indirect = True
unicode_cmap.Filter = [PdfName.FlateDecode]
unicode_cmap.stream = zlib.compress(unicode_cmap_stream, 9).decode(
"latin-1")
font = PdfDict()
font.indirect = True
font.BaseFont = PdfName.GlyphLessFont
font.DescendantFonts = PdfArray([cid_font])
font.Encoding = PdfName("Identity-H")
font.Subtype = PdfName.Type0
font.ToUnicode = unicode_cmap
font.Type = PdfName.Font
return font
def make_page(page, pdf_page, psem):
# Prepare everything in parallel
@asyncio.coroutine
def get_pdf_thumbnail(psem):
if page.thumbnail is None:
return None
return (yield from page.thumbnail.pdf_thumbnail(psem))
@asyncio.coroutine
def get_pdf_background(psem):
if page.background is None:
return None
return (yield from page.background.pdf_image(psem))
@asyncio.coroutine
def get_pdf_mask(foreground, psem):
if foreground.color is not None:
return None
return (yield from foreground.pdf_mask(psem))
pdf_thumbnail, pdf_background, pdf_foregrounds, pdf_masks = (
yield from asyncio.gather(
get_pdf_thumbnail(psem), get_pdf_background(psem),
asyncio.gather(
*[fg.pdf_image(psem) for fg in page.foreground]),
asyncio.gather(
*[get_pdf_mask(fg, psem) for fg in page.foreground])))
pdf_page.MediaBox = PdfArray(
[0, 0, PdfNumber(page.width),
PdfNumber(page.height)])
pdf_page.Group = pdf_group
pdf_resources = PdfDict()
pdf_colorspace = PdfDict()
pdf_colorspace.DefaultRGB = default_rgb_colorspace
pdf_resources.ColorSpace = pdf_colorspace
pdf_xobject = PdfDict()
if pdf_thumbnail is not None:
pdf_page.Thumb = pdf_thumbnail
im_index = 0
# Save graphics state and scale unity rectangle to page size
matrix = TransformationMatrix()
matrix.scale(page.width, page.height)
before_graphics = ("q\n" + "%s cm\n" % matrix.to_pdf())
after_graphics = "\nQ\n"
contents = ""
graphics = ""
current_color = None
if page.color != self._factory.WHITE:
if current_color != page.color:
current_color = page.color
graphics += page.color.to_pdf() + " rg "
graphics += ("0 0 1 1 re " + "f\n")
if pdf_background is not None:
pdf_xobject[PdfName("Im%d" % im_index)] = pdf_background
graphics += "/Im%d Do\n" % im_index
im_index += 1
for foreground, pdf_foreground, pdf_mask in zip(
page.foreground, pdf_foregrounds, pdf_masks):
if pdf_mask is not None:
pdf_xobject[PdfName("Im%d" % im_index)] = pdf_mask
im_index += 1
pdf_xobject[PdfName("Im%d" % im_index)] = pdf_foreground
if (foreground.color is not None
and current_color != foreground.color):
current_color = foreground.color
graphics += foreground.color.to_pdf() + " rg "
graphics += "/Im%d Do\n" % im_index
im_index += 1
if graphics:
contents += (before_graphics + graphics.rstrip(" \n") +
after_graphics)
current_color = None
before_text = ("BT\n" + "/F1 1 Tf 3 Tr\n")
after_text = "\nET\n"
text = ""
pdf_annots = []
for t in page.text:
if t.text:
matrix = TransformationMatrix()
# Glyph size is 0.5 x 1
matrix.scale(2 / len(t.text), 1)
matrix.translate(-0.5, -0.5)
if t.direction == "ltr":
pass
elif t.direction == "rtl":
matrix.translate(0, -1)
elif t.direction == "ttb":
matrix.rotate(90)
matrix.rotate(-t.rotation)
matrix.translate(0.5, 0.5)
matrix.scale(t.width, t.height)
matrix.translate(t.x, t.y)
text += "%s Tm %s Tj\n" % (
matrix.to_pdf(), PdfString().from_bytes(
t.text.encode("utf-16-be"), bytes_encoding="hex"))
if t.external_link is not None or t.internal_link is not None:
pdf_annot = PdfDict()
pdf_annots.append(pdf_annot)
pdf_annot.Type = PdfName.Annot
pdf_annot.Subtype = PdfName.Link
pdf_annot.Border = [0, 0, 0]
pdf_annot.Rect = [
PdfNumber(t.x),
PdfNumber(t.y),
PdfNumber(t.x + t.width),
PdfNumber(t.y + t.height)
]
if t.external_link is not None:
pdf_a = PdfDict()
pdf_annot.A = pdf_a
pdf_a.Type = PdfName.Action
pdf_a.S = PdfName.URI
pdf_a.URI = t.external_link.decode("latin-1")
if t.internal_link is not None:
pdf_target_page = pdf_pages[t.internal_link[0]]
target_x, target_y = t.internal_link[1]
pdf_annot.Dest = [
pdf_target_page, PdfName.XYZ,
PdfNumber(target_x),
PdfNumber(target_y), 0
]
text = text.rstrip(" \n")
if text:
pdf_resources.Font = pdf_font_mapping
contents += (before_text + text + after_text)
contents = contents.rstrip(" \n")
if contents:
pdf_contents = PdfDict()
pdf_contents.indirect = True
pdf_page.Contents = pdf_contents
if COMPRESS_PAGE_CONTENTS:
pdf_contents.Filter = [PdfName.FlateDecode]
pdf_contents.stream = zlib.compress(
contents.encode("latin-1"), 9).decode("latin-1")
else:
pdf_contents.stream = contents
if pdf_annots:
pdf_page.Annots = pdf_annots
if pdf_xobject:
pdf_resources.XObject = pdf_xobject
if pdf_resources:
pdf_page.Resources = pdf_resources
# Report progress
nonlocal finished_pages
finished_pages += 1
if progress_cb:
progress_cb(finished_pages / len(self._pages))
def write_async(self, outfile, process_semaphore, progress_cb=None):
pdf_writer = PdfWriter(version="1.5")
pdf_group = PdfDict()
pdf_group.indirect = True
pdf_group.CS = PdfName.DeviceRGB
pdf_group.I = PdfBool(True)
pdf_group.S = PdfName.Transparency
pdf_font_mapping = PdfDict()
pdf_font_mapping.indirect = True
pdf_font_mapping.F1 = self._build_font()
for _ in self._pages:
pdf_page = PdfDict()
pdf_page.Type = PdfName.Page
pdf_writer.addpage(pdf_page)
# pdfrw makes a internal copy of the pages
# use the copy so that references to pages in links are correct
pdf_pages = list(pdf_writer.pagearray)
srgb_colorspace = PdfDict()
srgb_colorspace.indirect = True
srgb_colorspace.N = 3 # Number of components (red, green, blue)
with open(SRGB_ICC_FILENAME, "rb") as f:
srgb_colorspace_stream = f.read()
srgb_colorspace.Filter = [PdfName.FlateDecode]
srgb_colorspace.stream = zlib.compress(srgb_colorspace_stream,
9).decode("latin-1")
srgb_colorspace.Length1 = len(srgb_colorspace_stream)
default_rgb_colorspace = PdfArray([PdfName.ICCBased, srgb_colorspace])
default_rgb_colorspace.indirect = True
# Handle all pages in parallel
@asyncio.coroutine
def make_page(page, pdf_page, psem):
# Prepare everything in parallel
@asyncio.coroutine
def get_pdf_thumbnail(psem):
if page.thumbnail is None:
return None
return (yield from page.thumbnail.pdf_thumbnail(psem))
@asyncio.coroutine
def get_pdf_background(psem):
if page.background is None:
return None
return (yield from page.background.pdf_image(psem))
@asyncio.coroutine
def get_pdf_mask(foreground, psem):
if foreground.color is not None:
return None
return (yield from foreground.pdf_mask(psem))
pdf_thumbnail, pdf_background, pdf_foregrounds, pdf_masks = (
yield from asyncio.gather(
get_pdf_thumbnail(psem), get_pdf_background(psem),
asyncio.gather(
*[fg.pdf_image(psem) for fg in page.foreground]),
asyncio.gather(
*[get_pdf_mask(fg, psem) for fg in page.foreground])))
pdf_page.MediaBox = PdfArray(
[0, 0, PdfNumber(page.width),
PdfNumber(page.height)])
pdf_page.Group = pdf_group
pdf_resources = PdfDict()
pdf_colorspace = PdfDict()
pdf_colorspace.DefaultRGB = default_rgb_colorspace
pdf_resources.ColorSpace = pdf_colorspace
pdf_xobject = PdfDict()
if pdf_thumbnail is not None:
pdf_page.Thumb = pdf_thumbnail
im_index = 0
# Save graphics state and scale unity rectangle to page size
matrix = TransformationMatrix()
matrix.scale(page.width, page.height)
before_graphics = ("q\n" + "%s cm\n" % matrix.to_pdf())
after_graphics = "\nQ\n"
contents = ""
graphics = ""
current_color = None
if page.color != self._factory.WHITE:
if current_color != page.color:
current_color = page.color
graphics += page.color.to_pdf() + " rg "
graphics += ("0 0 1 1 re " + "f\n")
if pdf_background is not None:
pdf_xobject[PdfName("Im%d" % im_index)] = pdf_background
graphics += "/Im%d Do\n" % im_index
im_index += 1
for foreground, pdf_foreground, pdf_mask in zip(
page.foreground, pdf_foregrounds, pdf_masks):
if pdf_mask is not None:
pdf_xobject[PdfName("Im%d" % im_index)] = pdf_mask
im_index += 1
pdf_xobject[PdfName("Im%d" % im_index)] = pdf_foreground
if (foreground.color is not None
and current_color != foreground.color):
current_color = foreground.color
graphics += foreground.color.to_pdf() + " rg "
graphics += "/Im%d Do\n" % im_index
im_index += 1
if graphics:
contents += (before_graphics + graphics.rstrip(" \n") +
after_graphics)
current_color = None
before_text = ("BT\n" + "/F1 1 Tf 3 Tr\n")
after_text = "\nET\n"
text = ""
pdf_annots = []
for t in page.text:
if t.text:
matrix = TransformationMatrix()
# Glyph size is 0.5 x 1
matrix.scale(2 / len(t.text), 1)
matrix.translate(-0.5, -0.5)
if t.direction == "ltr":
pass
elif t.direction == "rtl":
matrix.translate(0, -1)
elif t.direction == "ttb":
matrix.rotate(90)
matrix.rotate(-t.rotation)
matrix.translate(0.5, 0.5)
matrix.scale(t.width, t.height)
matrix.translate(t.x, t.y)
text += "%s Tm %s Tj\n" % (
matrix.to_pdf(), PdfString().from_bytes(
t.text.encode("utf-16-be"), bytes_encoding="hex"))
if t.external_link is not None or t.internal_link is not None:
pdf_annot = PdfDict()
pdf_annots.append(pdf_annot)
pdf_annot.Type = PdfName.Annot
pdf_annot.Subtype = PdfName.Link
pdf_annot.Border = [0, 0, 0]
pdf_annot.Rect = [
PdfNumber(t.x),
PdfNumber(t.y),
PdfNumber(t.x + t.width),
PdfNumber(t.y + t.height)
]
if t.external_link is not None:
pdf_a = PdfDict()
pdf_annot.A = pdf_a
pdf_a.Type = PdfName.Action
pdf_a.S = PdfName.URI
pdf_a.URI = t.external_link.decode("latin-1")
if t.internal_link is not None:
pdf_target_page = pdf_pages[t.internal_link[0]]
target_x, target_y = t.internal_link[1]
pdf_annot.Dest = [
pdf_target_page, PdfName.XYZ,
PdfNumber(target_x),
PdfNumber(target_y), 0
]
text = text.rstrip(" \n")
if text:
pdf_resources.Font = pdf_font_mapping
contents += (before_text + text + after_text)
contents = contents.rstrip(" \n")
if contents:
pdf_contents = PdfDict()
pdf_contents.indirect = True
pdf_page.Contents = pdf_contents
if COMPRESS_PAGE_CONTENTS:
pdf_contents.Filter = [PdfName.FlateDecode]
pdf_contents.stream = zlib.compress(
contents.encode("latin-1"), 9).decode("latin-1")
else:
pdf_contents.stream = contents
if pdf_annots:
pdf_page.Annots = pdf_annots
if pdf_xobject:
pdf_resources.XObject = pdf_xobject
if pdf_resources:
pdf_page.Resources = pdf_resources
# Report progress
nonlocal finished_pages
finished_pages += 1
if progress_cb:
progress_cb(finished_pages / len(self._pages))
finished_pages = 0
yield from asyncio.gather(*[
make_page(page, pdf_page, process_semaphore)
for page, pdf_page in zip(self._pages, pdf_pages)
])
trailer = pdf_writer.trailer
document_id = PdfString().from_bytes(os.urandom(16))
trailer.ID = [document_id, document_id]
mark_info = PdfDict()
mark_info.Marked = PdfBool(True)
trailer.Root.MarkInfo = mark_info
struct_tree_root = PdfDict()
struct_tree_root.Type = PdfName.StructTreeRoot
trailer.Root.StructTreeRoot = struct_tree_root
metadata = PdfDict()
metadata.indirect = True
metadata.Type = PdfName.Metadata
metadata.Subtype = PdfName.XML
xmp = XMPMeta()
xmp.set_property(XMP_NS_PDFA_ID, "part", "2")
xmp.set_property(XMP_NS_PDFA_ID, "conformance", "A")
metadata_stream = xmp.serialize_to_str().encode("utf-8")
metadata.Filter = [PdfName.FlateDecode]
metadata.stream = zlib.compress(metadata_stream, 9).decode("latin-1")
metadata.Length1 = len(metadata_stream)
trailer.Root.Metadata = metadata
with TemporaryDirectory(prefix="djpdf-") as temp_dir:
pdf_writer.write(path.join(temp_dir, "temp.pdf"))
cmd = [
QPDF_CMD, "--stream-data=preserve",
"--object-streams=preserve", "--normalize-content=n",
"--newline-before-endstream"
]
if LINEARIZE_PDF:
cmd.extend(["--linearize"])
cmd.extend([
path.abspath(path.join(temp_dir, "temp.pdf")),
path.abspath(outfile)
])
yield from run_command_async(cmd, process_semaphore)
def _build_font():
with open(FONT_FILENAME, "rb") as f:
embedded_font_stream = f.read()
embedded_font = PdfDict()
embedded_font.indirect = True
embedded_font.Filter = [PdfName.FlateDecode]
embedded_font.stream = zlib.compress(embedded_font_stream,
9).decode("latin-1")
embedded_font.Length1 = len(embedded_font_stream)
font_descriptor = PdfDict()
font_descriptor.indirect = True
font_descriptor.Ascent = 1000
font_descriptor.CapHeight = 1000
font_descriptor.Descent = -1
font_descriptor.Flags = 5 # FixedPitch + Symbolic
font_descriptor.FontBBox = PdfArray([0, 0, 1000, 500])
font_descriptor.FontFile2 = embedded_font
font_descriptor.FontName = PdfName.GlyphLessFont
font_descriptor.ItalicAngle = 0
font_descriptor.StemV = 80
font_descriptor.Type = PdfName.FontDescriptor
# Map everything to glyph 1
cid_to_gid_map_stream = b"\0\1" * (1 << 16)
cid_to_gid_map = PdfDict()
cid_to_gid_map.indirect = True
cid_to_gid_map.Filter = [PdfName.FlateDecode]
cid_to_gid_map.stream = zlib.compress(cid_to_gid_map_stream,
9).decode("latin-1")
cid_to_gid_map.Length1 = len(cid_to_gid_map_stream)
cid_system_info = PdfDict()
cid_system_info.Ordering = PdfString.from_unicode("Identity")
cid_system_info.Registry = PdfString.from_unicode("Adobe")
cid_system_info.Supplement = 0
cid_font = PdfDict()
cid_font.indirect = True
cid_font.CIDToGIDMap = cid_to_gid_map
cid_font.BaseFont = PdfName.GlyphLessFont
cid_font.CIDSystemInfo = cid_system_info
cid_font.FontDescriptor = font_descriptor
cid_font.Subtype = PdfName.CIDFontType2
cid_font.Type = PdfName.Font
cid_font.DW = 500
with open(UNICODE_CMAP_FILENAME, "rb") as f:
unicode_cmap_stream = f.read()
unicode_cmap = PdfDict()
unicode_cmap.indirect = True
unicode_cmap.Filter = [PdfName.FlateDecode]
unicode_cmap.stream = zlib.compress(unicode_cmap_stream,
9).decode("latin-1")
font = PdfDict()
font.indirect = True
font.BaseFont = PdfName.GlyphLessFont
font.DescendantFonts = PdfArray([cid_font])
font.Encoding = PdfName("Identity-H")
font.Subtype = PdfName.Type0
font.ToUnicode = unicode_cmap
font.Type = PdfName.Font
return font