Fix SAS7BDAT run-length encoding formula (pandas-dev#47099)

* Fix SAS7BDAT run-length encoding formula

See https://bitbucket.org/jaredhobbs/sas7bdat/src/18cbd14407918c1aa90f136c1d6c5d83f307dba0/sas7bdat.py#lines-110:114

* Undo safety change

* Undo unnecessary bit mask

* Undo syntax change

* Update v1.5.0.rst

* Update v1.5.0.rst

* Add test file

* Fix tests

* Update test_sas7bdat.py

Author: jonashaag

doc/source/whatsnew/v1.5.0.rst

@@ -945,6 +945,7 @@ I/O
 - Bug in :meth:`DataFrame.to_excel` when writing an empty dataframe with :class:`MultiIndex` (:issue:`19543`)
 - Bug in :func:`read_sas` with RLE-compressed SAS7BDAT files that contain 0x40 control bytes (:issue:`31243`)
 - Bug in :func:`read_sas` that scrambled column names (:issue:`31243`)
+- Bug in :func:`read_sas` with RLE-compressed SAS7BDAT files that contain 0x00 control bytes (:issue:`47099`)
 -
 
 Period

pandas/io/sas/sas.pyx

@@ -28,9 +28,7 @@ cdef const uint8_t[:] rle_decompress(int result_length, const uint8_t[:] inbuff)
         ipos += 1
 
         if control_byte == 0x00:
-            if end_of_first_byte != 0:
-                raise ValueError("Unexpected non-zero end_of_first_byte")
-            nbytes = <int>(inbuff[ipos]) + 64
+            nbytes = <int>(inbuff[ipos]) + 64 + end_of_first_byte * 256
             ipos += 1
             for _ in range(nbytes):
                 result[rpos] = inbuff[ipos]

pandas/tests/io/sas/data/0x00controlbyte.sas7bdat.bz2

@@ -390,3 +390,10 @@ def test_0x40_control_byte(datapath):
     fname = datapath("io", "sas", "data", "0x40controlbyte.csv")
     df0 = pd.read_csv(fname, dtype="object")
     tm.assert_frame_equal(df, df0)
+
+
+def test_0x00_control_byte(datapath):
+    # GH 47099
+    fname = datapath("io", "sas", "data", "0x00controlbyte.sas7bdat.bz2")
+    df = next(pd.read_sas(fname, chunksize=11_000))
+    assert df.shape == (11_000, 20)