BUG: Make sure that sas7bdat parsers memory is initialized to 0 (pandas-dev#21616)

Memory for numbers in sas7bdat-parsing was not initialized properly to 0.
For sas7bdat files with numbers smaller than 8 bytes this made the
least significant part of the numbers essentially random.
Fix it by initializing memory correctly.

Author: troels

doc/source/whatsnew/v0.24.0.txt

@@ -734,7 +734,7 @@ I/O
 - :func:`read_html()` no longer ignores all-whitespace ``<tr>`` within ``<thead>`` when considering the ``skiprows`` and ``header`` arguments. Previously, users had to decrease their ``header`` and ``skiprows`` values on such tables to work around the issue. (:issue:`21641`)
 - :func:`read_excel()` will correctly show the deprecation warning for previously deprecated ``sheetname`` (:issue:`17994`)
 - :func:`read_csv()` will correctly parse timezone-aware datetimes (:issue:`22256`)
--
+- :func:`read_sas()` will parse numbers in sas7bdat-files that have width less than 8 bytes correctly. (:issue:`21616`)
 
 Plotting
 ^^^^^^^^

pandas/io/sas/sas7bdat.py

@@ -614,7 +614,7 @@ def read(self, nrows=None):
         ns = (self.column_types == b's').sum()
 
         self._string_chunk = np.empty((ns, nrows), dtype=np.object)
-        self._byte_chunk = np.empty((nd, 8 * nrows), dtype=np.uint8)
+        self._byte_chunk = np.zeros((nd, 8 * nrows), dtype=np.uint8)
 
         self._current_row_in_chunk_index = 0
         p = Parser(self)

pandas/tests/io/sas/data/cars.sas7bdat

@@ -183,6 +183,17 @@ def test_date_time(datapath):
     tm.assert_frame_equal(df, df0)
 
 
+def test_compact_numerical_values(datapath):
+    # Regression test for #21616
+    fname = datapath("io", "sas", "data", "cars.sas7bdat")
+    df = pd.read_sas(fname, encoding='latin-1')
+    # The two columns CYL and WGT in cars.sas7bdat have column
+    # width < 8 and only contains integral values. Test
+    # that pandas doesn't corrupt the less significant bits.
+    tm.assert_series_equal(df['WGT'], df['WGT'].round(), check_exact=True)
+    tm.assert_series_equal(df['CYL'], df['CYL'].round(), check_exact=True)
+
+
 def test_zero_variables(datapath):
     # Check if the SAS file has zero variables (PR #18184)
     fname = datapath("io", "sas", "data", "zero_variables.sas7bdat")