def review_file(self, reader: PdfFileReader,
base_revision: Union[int, HistoricalResolver],
field_mdp_spec: Optional[FieldMDPSpec] = None,
doc_mdp: Optional[MDPPerm] = None) \
-> Union[DiffResult, SuspiciousModification]:
"""
Implementation of :meth:`.DiffPolicy.review_file` that reviews
each intermediate revision between the base revision and the current one
individually.
"""
changed_form_fields = set()
rev_count = reader.xrefs.total_revisions
current_max = ModificationLevel.NONE
if isinstance(base_revision, int):
base_rev_resolver = reader.get_historical_resolver(base_revision)
else:
base_rev_resolver = base_revision
base_revision = base_rev_resolver.revision
# Note: there's a pragmatic reason why we iterate over all revisions
# instead of just asking for all updated objects between the signed
# revision and the most recent one:
#
# The effect of intermediate updates may not be detectable anymore in
# the most recent version, so if we'd consolidate all checks into one,
# we would have no way to tell whether or not the objects created
# (and later forgotten) by these intermediate revisions actually
# constituted legitimate changes.
# (see the test_pades_revinfo tests for examples where this applies)
#
# Until we have a reference counter (which comes with its own
# performance problems that may or may not be worse), I don't really
# see a good way around this issue other than diffing every intermediate
# version separately.
for revision in range(base_revision + 1, rev_count):
try:
diff_result = self.apply(
old=base_rev_resolver,
new=reader.get_historical_resolver(revision),
field_mdp_spec=field_mdp_spec,
doc_mdp=doc_mdp)
except SuspiciousModification as e:
logger.warning(
'Error in diff operation between revision '
f'{base_revision} and {revision}',
exc_info=e)
return e
current_max = max(current_max, diff_result.modification_level)
changed_form_fields |= diff_result.changed_form_fields
return DiffResult(current_max, changed_form_fields)
def test_tagged_path_count():
r = PdfFileReader(BytesIO(MINIMAL_TWO_FIELDS_TAGGED))
r = r.get_historical_resolver(0)
r._load_reverse_xref_cache()
# The path simplifier should eliminate all (pseudo-)duplicates refs except
# these three:
# - one from the AcroForm hierarchy
# - one from the pages tree (through /Annots)
# - one from the structure tree
paths_to = r._indirect_object_access_cache[generic.Reference(7, 0, r)]
assert len(paths_to) == 3