BUG: _get_fonts not processing properly CIDFonts and annotations (#2194)

Closes #2192

pypdf/_page.py

@@ -2332,7 +2332,9 @@ def _get_fonts(self) -> Tuple[Set[str], Set[str]]:
  """
  obj = self.get_object()
  assert isinstance(obj, DictionaryObject)
- fonts, embedded = _get_fonts_walk(cast(DictionaryObject, obj[PG.RESOURCES]))
+ fonts: Set[str] = set()
+ embedded: Set[str] = set()
+ fonts, embedded = _get_fonts_walk(cast(DictionaryObject, obj), fonts, embedded)
  unembedded = fonts - embedded
  return embedded, unembedded
 
@@ -2560,8 +2562,8 @@ def __str__(self) -> str:
 
 def _get_fonts_walk(
  obj: DictionaryObject,
- fnt: Optional[Set[str]] = None,
- emb: Optional[Set[str]] = None,
+ fnt: Set[str],
+ emb: Set[str],
 ) -> Tuple[Set[str], Set[str]]:
  """
  Get the set of all fonts and all embedded fonts.
@@ -2581,22 +2583,77 @@ def _get_fonts_walk(
 
  We create and add to two sets, fnt = fonts used and emb = fonts embedded.
  """
- if fnt is None:
- fnt = set()
- if emb is None:
- emb = set()
- if not hasattr(obj, "keys"):
- return set(), set()
  fontkeys = ("/FontFile", "/FontFile2", "/FontFile3")
- if "/BaseFont" in obj:
- fnt.add(cast(str, obj["/BaseFont"]))
- if "/FontName" in obj and [x for x in fontkeys if x in obj]:
- # the list comprehension ensures there is FontFile
- emb.add(cast(str, obj["/FontName"]))
-
- for key in obj:
- _get_fonts_walk(cast(DictionaryObject, obj[key]), fnt, emb)
 
+ def process_font(f: DictionaryObject) -> None:
+ nonlocal fnt, emb
+ f = cast(DictionaryObject, f.get_object()) # to be sure
+ if "/BaseFont" in f:
+ fnt.add(cast(str, f["/BaseFont"]))
+
+ if (
+ ("/CharProcs" in f)
+ or (
+ "/FontDescriptor" in f
+ and any(
+ x in cast(DictionaryObject, f["/FontDescriptor"]) for x in fontkeys
+ )
+ )
+ or (
+ "/DescendantFonts" in f
+ and "/FontDescriptor"
+ in cast(
+ DictionaryObject,
+ cast(ArrayObject, f["/DescendantFonts"])[0].get_object(),
+ )
+ and any(
+ x
+ in cast(
+ DictionaryObject,
+ cast(
+ DictionaryObject,
+ cast(ArrayObject, f["/DescendantFonts"])[0].get_object(),
+ )["/FontDescriptor"],
+ )
+ for x in fontkeys
+ )
+ )
+ ):
+ # the list comprehension ensures there is FontFile
+ emb.add(cast(str, f["/BaseFont"]))
+
+ if "/DR" in obj and "/Font" in cast(DictionaryObject, obj["/DR"]):
+ for f in cast(DictionaryObject, cast(DictionaryObject, obj["/DR"])["/Font"]):
+ process_font(f)
+ if "/Resources" in obj:
+ if "/Font" in cast(DictionaryObject, obj["/Resources"]):
+ for f in cast(
+ DictionaryObject, cast(DictionaryObject, obj["/Resources"])["/Font"]
+ ).values():
+ process_font(f)
+ if "/XObject" in cast(DictionaryObject, obj["/Resources"]):
+ for x in cast(
+ DictionaryObject, cast(DictionaryObject, obj["/Resources"])["/XObject"]
+ ).values():
+ _get_fonts_walk(cast(DictionaryObject, x.get_object()), fnt, emb)
+ if "/Annots" in obj:
+ for a in cast(ArrayObject, obj["/Annots"]):
+ _get_fonts_walk(cast(DictionaryObject, a.get_object()), fnt, emb)
+ if "/AP" in obj:
+ if (
+ cast(DictionaryObject, cast(DictionaryObject, obj["/AP"])["/N"]).get(
+ "/Type"
+ )
+ == "/XObject"
+ ):
+ _get_fonts_walk(
+ cast(DictionaryObject, cast(DictionaryObject, obj["/AP"])["/N"]),
+ fnt,
+ emb,
+ )
+ else:
+ for a in cast(DictionaryObject, cast(DictionaryObject, obj["/AP"])["/N"]):
+ _get_fonts_walk(cast(DictionaryObject, a), fnt, emb)
  return fnt, emb # return the sets for each page
 
 

tests/test_page.py

@@ -750,6 +750,13 @@ def visitor_td(op, args, cm, tm) -> None:
  set(),
  {"/Helvetica"},
  ),
+ # fonts in annotations
+ (
+ RESOURCE_ROOT / "FormTestFromOo.pdf",
+ None,
+ {"/CAAAAA+LiberationSans", "/EAAAAA+SegoeUI", "/BAAAAA+LiberationSerif"},
+ {"/LiberationSans", "/ZapfDingbats"},
+ ),
  ],
 )
 def test_get_fonts(pdf_path, password, embedded, unembedded):
@@ -763,6 +770,44 @@ def test_get_fonts(pdf_path, password, embedded, unembedded):
  assert (a, b) == (embedded, unembedded)
 
 
+@pytest.mark.enable_socket()
+def test_get_fonts2():
+ url = "https://github.com/py-pdf/pypdf/files/12618104/WS_T.483.8-2016.pdf"
+ name = "WS_T.483.8-2016.pdf"
+ reader = PdfReader(BytesIO(get_data_from_url(url, name=name)))
+ assert reader.pages[1]._get_fonts() == (
+ {
+ "/E-HZ9-PK7483a5-Identity-H",
+ "/SSJ-PK748200005d9-Identity-H",
+ "/QGNGZS+FzBookMaker1DlFont10536872415",
+ "/E-BZ9-PK748344-Identity-H",
+ "/E-FZ9-PK74836f-Identity-H",
+ "/O9-PK748464-Identity-H",
+ "/QGNGZR+FzBookMaker0DlFont00536872414",
+ "/SSJ-PK748200005db-Identity-H",
+ "/F-BZ9-PK7483cb-Identity-H",
+ "/SSJ-PK748200005da-Identity-H",
+ "/H-SS9-PK748200005e0-Identity-H",
+ "/H-HT9-PK748200005e1-Identity-H",
+ },
+ set(),
+ )
+ assert reader.pages[2]._get_fonts() == (
+ {
+ "/E-HZ9-PK7483a5-Identity-H",
+ "/E-FZ9-PK74836f-Identity-H",
+ "/E-BZ9-PK748344-Identity-H",
+ "/QGNGZT+FzBookMaker0DlFont00536872418",
+ "/O9-PK748464-Identity-H",
+ "/F-BZ9-PK7483cb-Identity-H",
+ "/H-SS9-PK748200005e0-Identity-H",
+ "/QGNGZU+FzBookMaker1DlFont10536872420",
+ "/H-HT9-PK748200005e1-Identity-H",
+ },
+ set(),
+ )
+
+
 def test_annotation_getter():
  pdf_path = RESOURCE_ROOT / "commented.pdf"
  reader = PdfReader(pdf_path)