I am using PyPDF4 to access to the "full" pdf-structures of a pdf and recursively store its values in a dictionary. The algorithm is supposed to work page-wise.
A pdf object for common data structure, IndirectObject, needs to be cast to dictionary via the getObject() method. Once the casting is performed then it is possible that some of its values are IndirectObject s themselves. They may occurs in lists and dictionaries and standalone:
{..., '/Kids': [IndirectObject(32, 0), ...], '/Count': 4}{'/ColorSpace': {'/CS0': IndirectObject(19023, 0), ...}{..., '/Parent': IndirectObject(26, 0), '/Resources': ...}
The output of the program is incorrect, still some IndirectObject are presents. I tried several different implementations but I cannot see what I am doing wrong. How can I fix it?
For debugging purposes to each such substitution the key is modified to have the following pattern: /OriginalKey_NumID.
The output refers to a pdf that can be found here.
PyPDF4 should be back-compatible with PyPDF2, just change the import part with PyPDF2. Now, PyPDF2 seems to be in a reborn phase and some changes were made, see history.
from PyPDF4 import PdfFileReader # should be also compatible with PyPDF2
from PyPDF4.pdf import IndirectObject
def inspect_page(c_dict):
dg = {}
def __inspect_page(c_dict):
d = {}
for k, v in c_dict.items():
if isinstance(v, IndirectObject):
#d[f'{k}_{v.idnum}'] = __inspect_page(v.getObject()) # lead to oo-recursion
d[f'{k}_{v.idnum}'] = v.getObject()
elif isinstance(v, list):
loc_tmp = []
for vv in v:
if isinstance(vv, IndirectObject):
local_d_tmp = {f'{k}_{vv.idnum}': __inspect_page(vv.getObject())}
loc_tmp.append(local_d_tmp)
elif isinstance(vv, dict):
loc_tmp.append(__inspect_page(vv))
else:
loc_tmp.append(vv)
d[k] = loc_tmp
elif isinstance(v, dict):
v.update(__inspect_page(v))
d[k] = v
else:
d[k] = v
dg.update(d)
return d
__inspect_page(c_dict)
return dg
#process pdf
# sample src: https://www.adobe.com/support/products/enterprise/knowledgecenter/media/c4611_sample_explain.pdf
path = # path to pdf
objects_pages = {}
rpdf = PdfFileReader(open(path, 'rb'))
for i in range(rpdf.getNumPages()):
page = rpdf.getPage(i)
objects_pages[i] = inspect_page(page)
if i == 0:
break
# check output
page_num = 0
print(f'content of page {page_num}')
for k, v in objects_pages[page_num].items():
print(k, v)
Output (IndirectObject are still presents, both in list and as dictionary values)
content of page 0
/StructParents 0
/Rotate 0
/CS0_19023 ['/Separation', '/Black', '/DeviceCMYK', IndirectObject(19029, 0)]
/ColorSpace {'/CS0': IndirectObject(19023, 0), '/CS0_19023': ['/Separation', '/Black', '/DeviceCMYK', IndirectObject(19029, 0)]}
/T1_0_19022 {'/Subtype': '/Type1', '/FontDescriptor': IndirectObject(19024, 0), '/LastChar': 255, '/Widths': [...], '/BaseFont': '/Times-Roman', '/FirstChar': 0, '/Encoding': '/MacRomanEncoding', '/Type': '/Font'}
/T1_1_19027 {'/Subtype': '/Type1', '/FontDescriptor': IndirectObject(19028, 0), '/LastChar': 255, '/Widths': [...], '/BaseFont': '/Times-Italic', '/FirstChar': 0, '/Encoding': '/WinAnsiEncoding', '/Type': '/Font'}
/Font {'/T1_0': IndirectObject(19022, 0), '/T1_1': IndirectObject(19027, 0), '/T1_0_19022': {'/Subtype': '/Type1', '/FontDescriptor': IndirectObject(19024, 0), '/LastChar': 255, '/Widths': [...], '/BaseFont': '/Times-Roman', '/FirstChar': 0, '/Encoding': '/MacRomanEncoding', '/Type': '/Font'}, '/T1_1_19027': {'/Subtype': '/Type1', '/FontDescriptor': IndirectObject(19028, 0), '/LastChar': 255, '/Widths': [...], '/BaseFont': '/Times-Italic', '/FirstChar': 0, '/Encoding': '/WinAnsiEncoding', '/Type': '/Font'}}
/GS0_19025 {'/OPM': 1, '/op': <PyPDF4.generic.BooleanObject object at 0x7f0264627280>, '/Type': '/ExtGState', '/SA': <PyPDF4.generic.BooleanObject object at 0x7f0264627310>, '/SM': 0.02}
/ExtGState {'/GS0': IndirectObject(19025, 0), '/GS0_19025': {'/OPM': 1, '/op': <PyPDF4.generic.BooleanObject object at 0x7f0264627280>, '/Type': '/ExtGState', '/SA': <PyPDF4.generic.BooleanObject object at 0x7f0264627310>, '/SM': 0.02}}
/ProcSet ['/PDF', '/Text']
/Resources {'/ColorSpace': {'/CS0': IndirectObject(19023, 0), '/CS0_19023': ['/Separation', '/Black', '/DeviceCMYK', IndirectObject(19029, 0)]}, '/Font': {'/T1_0': IndirectObject(19022, 0), '/T1_1': IndirectObject(19027, 0), '/T1_0_19022': {'/Subtype': '/Type1', '/FontDescriptor': IndirectObject(19024, 0), '/LastChar': 255, '/Widths': [...], '/BaseFont': '/Times-Roman', '/FirstChar': 0, '/Encoding': '/MacRomanEncoding', '/Type': '/Font'}, '/T1_1_19027': {'/Subtype': '/Type1', '/FontDescriptor': IndirectObject(19028, 0), '/LastChar': 255, '/Widths': [...], '/BaseFont': '/Times-Italic', '/FirstChar': 0, '/Encoding': '/WinAnsiEncoding', '/Type': '/Font'}}, '/ProcSet': ['/PDF', '/Text'], '/ExtGState': {'/GS0': IndirectObject(19025, 0), '/GS0_19025': {'/OPM': 1, '/op': <PyPDF4.generic.BooleanObject object at 0x7f0264627280>, '/Type': '/ExtGState', '/SA': <PyPDF4.generic.BooleanObject object at 0x7f0264627310>, '/SM': 0.02}}}
/Type /Page
/CropBox [0, 0, 441, 666.36]
/Parent_18919 {'/Parent': IndirectObject(18918, 0), '/Count': 7, '/Type': '/Pages', '/Kids': [IndirectObject(19021, 0), IndirectObject(1, 0), IndirectObject(10, 0), IndirectObject(13, 0), IndirectObject(15, 0), IndirectObject(19, 0), IndirectObject(23, 0)]}
/Contents_19026 {'/Filter': '/FlateDecode'}
/BleedBox [0, 0, 441, 666.36]
/MediaBox [0, 0, 441, 666.36]
/Thumb_4489 {'/Filter': '/FlateDecode', '/BitsPerComponent': 8, '/ColorSpace': IndirectObject(2517, 0), '/Width': 55, '/Height': 83}
/TrimBox [0, 0, 441, 666.36]
EDIT:
IndirectObjects are uniquely identified by their idnum but a are not unique inside the PDF's structure. To avoid oo-recursion a set to keep track of each idnum should be created to avoid duplicates.
