fix binary DXF issue of XDATA in R2000+ files

README.md

@@ -21,9 +21,9 @@ Quick-Info
 - additional required packages: [pyparsing](https://pypi.org/project/pyparsing/)
 - MIT-License
 - read/write/new support for DXF versions: R12, R2000, R2004, R2007, R2010, R2013 and R2018
-- read/write support for ASCII DXF and Binary DXF, export of Binary DXF is experimental and can produce invalid files.
 - additional read support for DXF versions R13/R14 (upgraded to R2000)
 - additional read support for older DXF versions than R12 (upgraded to R12)
+- read/write support for ASCII DXF and Binary DXF
 - preserves third-party DXF content
 
 Included extensions:

docs/source/index.rst

@@ -21,6 +21,7 @@ Quick-Info
 - read/write/new support for DXF versions: R12, R2000, R2004, R2007, R2010, R2013 and R2018
 - additional read support for DXF versions R13/R14 (upgraded to R2000)
 - additional read support for older DXF versions than R12 (upgraded to R12)
+- read/write support for ASCII DXF and Binary DXF
 - preserves third-party DXF content
 
 Included Extensions

integration_tests/test_open_binary_DXF_files.py

@@ -31,6 +31,20 @@ def test_open_R13_R14(filename, tmpdir):
     assert os.path.exists(converted)
 
 
+def test_open_R2000_with_xdata():
+    filename = os.path.join(BASEDIR, DATADIR, 'bin_dxf_r2000.dxf')
+    doc = ezdxf.readfile(filename)
+    assert 'Model' in doc.layouts, 'Model space not found'
+    assert 'Layout1' in doc.layouts, 'Paper space not found'
+    assert doc.dxfversion == 'AC1015'
+    msp = doc.modelspace()
+    text = msp.query('TEXT').first
+    assert text.dxf.text == 'ezdxf'
+    hyperlink, description, location = text.get_hyperlink()
+    assert hyperlink == 'http://ezdxf.mozman.at'
+    msp.add_line((0, 0), (10, 3))
+
+
 def test_open_R12():
     filename = os.path.join(BASEDIR, DATADIR, 'bin_dxf_r12.dxf')
     doc = ezdxf.readfile(filename)

src/ezdxf/lldxf/tagger.py

@@ -140,19 +140,20 @@ def scan_params():
         return encoding, dxfversion
 
     encoding, dxfversion = scan_params()
-    one_byte_group_code = dxfversion < 'AC1012'
+    r12 = dxfversion <= 'AC1009'
     index = 22
     data_length = len(data)
     unpack = struct.unpack_from
 
     while index < data_length:
         # decode next group code
         code = data[index]
-        if code == 255:  # extended data
-            code = (data[index + 2] << 8) | data[index + 1]
-            index += 3
-        elif one_byte_group_code:
-            index += 1
+        if r12:
+            if code == 255:  # extended data
+                code = (data[index + 2] << 8) | data[index + 1]
+                index += 3
+            else:
+                index += 1
         else:  # 2-byte group code
             code = (data[index + 1] << 8) | code
             index += 2

src/ezdxf/lldxf/tagwriter.py

@@ -66,7 +66,7 @@ def __init__(self, stream: BinaryIO, dxfversion=LATEST_DXF_VERSION, write_handle
         super().__init__(None, dxfversion, write_handles)
         self._stream = stream
         self._encoding = encoding  # output encoding
-        self._group_code_size = 1 if self.dxfversion <= 'AC1009' else 2
+        self._r12 = self.dxfversion <= 'AC1009'
 
     def write_signature(self) -> None:
         self._stream.write(b'AutoCAD Binary DXF\r\n\x1a\x00')
@@ -95,13 +95,16 @@ def write_tag2(self, code: int, value: Any) -> None:
         stream = self._stream
 
         # write group code
-        if code >= 1000:  # extended data
-            stream.write(b'\xff')
-            # always 2-byte group code for extended data
+        if self._r12:
+            # Special group code handling if DXF R12 and older
+            if code >= 1000:  # extended data
+                stream.write(b'\xff')
+                # always 2-byte group code for extended data
+                stream.write(code.to_bytes(2, 'little'))
+            else:
+                stream.write(code.to_bytes(1, 'little'))
+        else:  # for R2000+ do not need a leading 0xff in front of extended data
             stream.write(code.to_bytes(2, 'little'))
-        else:
-            stream.write(code.to_bytes(self._group_code_size, 'little'))
-
         # write tag content
         if code in BYTES:
             stream.write(int(value).to_bytes(1, 'little'))
@@ -127,7 +130,7 @@ def _write_binary_chunks(self, code: int, data: bytes) -> None:
 
         while index < size:
             # write group code
-            if code >= 1000:  # extended data, just 1004?
+            if self._r12 and code >= 1000:  # extended data, just 1004?
                 stream.write(b'\xff')  # extended data marker
             # binary data does not exist in regular R12 entities,
             # only 2-byte group codes required