Handle bytes-typed fontnames (fixes #461 + #842)

Came across this bit of code, which helps to solve some of the mystery
in issues #461 and #842:

https://git.ghostscript.com/?p=mupdf.git;a=blob;f=source/pdf/pdf-font.c;h=6322cedf2c26cfb312c0c0878d7aff97b4c7470e;hb=HEAD#l774

Now, for every char's fontname, we:

- Check whether its a `str` or `byte`
    - If the latter, we check whether it's one of the well-known codes from
      the link above
        - If so, we use that (preserving the part, if present, before
          the `+`)
        - If not, we just cast to str

pdfplumber/page.py

@@ -71,6 +71,27 @@ if TYPE_CHECKING:  # pragma: nocover
     from .display import PageImage
     from .pdf import PDF
 
+# via https://git.ghostscript.com/?p=mupdf.git;a=blob;f=source/pdf/pdf-font.c;h=6322cedf2c26cfb312c0c0878d7aff97b4c7470e;hb=HEAD#l774   # noqa
+
+CP936_FONTNAMES = {
+    b"\xcb\xce\xcc\xe5": "SimSun,Regular",
+    b"\xba\xda\xcc\xe5": "SimHei,Regular",
+    b"\xbf\xac\xcc\xe5_GB2312": "SimKai,Regular",
+    b"\xb7\xc2\xcb\xce_GB2312": "SimFang,Regular",
+    b"\xc1\xa5\xca\xe9": "SimLi,Regular",
+}
+
+
+def fix_fontname_bytes(fontname: bytes) -> str:
+    if b"+" in fontname:
+        split_at = fontname.index(b"+") + 1
+        prefix, suffix = fontname[:split_at], fontname[split_at:]
+    else:
+        prefix, suffix = b"", fontname
+
+    suffix_new = CP936_FONTNAMES.get(suffix, str(suffix)[2:-1])
+    return str(prefix)[2:-1] + suffix_new
+
 
 class Page(Container):
     cached_properties: List[str] = Container.cached_properties + ["_layout"]
@@ -221,6 +242,10 @@ class Page(Container):
             attr["stroking_color"] = gs.scolor
             attr["non_stroking_color"] = gs.ncolor
 
+            # Handle (rare) byte-encoded fontnames
+            if isinstance(attr["fontname"], bytes):
+                attr["fontname"] = fix_fontname_bytes(attr["fontname"])
+
         if "pts" in attr:
             attr["pts"] = list(map(self.point2coord, attr["pts"]))
 

tests/test_issues.py

@@ -190,6 +190,31 @@ class Test(unittest.TestCase):
             chars = (char for char in pdf.chars)
             pdfplumber.utils.extract_text(chars)
 
+    def test_issue_461_and_842(self):
+        """
+        pdfplumber should gracefully handle characters with byte-encoded
+        font names.
+        """
+        before = b"RGJSAP+\xcb\xce\xcc\xe5"
+        after = pdfplumber.page.fix_fontname_bytes(before)
+        assert after == "RGJSAP+SimSun,Regular"
+
+        before = b"\xcb\xce\xcc\xe5"
+        after = pdfplumber.page.fix_fontname_bytes(before)
+        assert after == "SimSun,Regular"
+
+        path = os.path.join(HERE, "pdfs/issue-461-example.pdf")
+        with pdfplumber.open(path) as pdf:
+            page = pdf.pages[0]
+            assert all(isinstance(c["fontname"], str) for c in page.chars)
+            page.dedupe_chars()
+
+        path = os.path.join(HERE, "pdfs/issue-842-example.pdf")
+        with pdfplumber.open(path) as pdf:
+            page = pdf.pages[0]
+            assert all(isinstance(c["fontname"], str) for c in page.chars)
+            page.dedupe_chars()
+
     def test_issue_463(self):
         """
         Extracting annotations should not raise UnicodeDecodeError on utf-16 text