Ignore (rather than crash on) broken references in annotations

Tentative fix for issue #42

Also:
 * use PSLiteral in ANNOT_SUBTYPES rather than hashing the same name strings
 * improve typing of _mkannotation, and placate mypy

pdfannots/__init__.py

@@ -29,11 +29,12 @@
 
 logger = logging.getLogger(__name__)
 
-ANNOT_SUBTYPES = {e.name: e for e in AnnotationType}
+ANNOT_SUBTYPES: typing.Dict[PSLiteral, AnnotationType] = {
+ PSLiteralTable.intern(e.name): e for e in AnnotationType}
 
 
 def _mkannotation(
- pa: typing.Any,
+ pa: typing.Dict[str, typing.Any],
  page: Page
 ) -> typing.Optional[Annotation]:
  """
@@ -44,10 +45,15 @@ def _mkannotation(
  """
 
  subtype = pa.get('Subtype')
- try:
- annot_type = ANNOT_SUBTYPES[subtype.name]
- except (TypeError, KeyError):
- # subtype is missing (None), or is an unknown/unsupported type
+ annot_type = None
+ if isinstance(subtype, PSLiteral):
+ try:
+ annot_type = ANNOT_SUBTYPES[subtype]
+ except KeyError:
+ pass
+
+ if annot_type is None:
+ logger.warning("Unsupported annotation subtype: %r", subtype)
  return None
 
  contents = pa.get('Contents')
@@ -445,7 +451,7 @@ def emit_progress(msg: str) -> None:
  # Construct Annotation objects, and append them to the page.
  for pa in pdftypes.resolve1(pdfpage.annots) if pdfpage.annots else []:
  if isinstance(pa, pdftypes.PDFObjRef):
- annot = _mkannotation(pa.resolve(), page)
+ annot = _mkannotation(pdftypes.dict_value(pa), page)
  if annot is not None:
  page.annots.append(annot)
  else: