Merge pull request #60 from jcrobak/indexerror-fix

IndexError fix for zero width

Author: jcrobak

parquet/encoding.py

@@ -147,6 +147,8 @@ def read_bitpacked(file_obj, header, width, debug_logging):
     if debug_logging:
         logger.debug("Reading a bit-packed run with: %s groups, count %s, bytes %s",
                      num_groups, count, byte_count)
+    if width == 0:
+        return [0 for _ in range(count)]
     raw_bytes = array.array(ARRAY_BYTE_STR, file_obj.read(byte_count)).tolist()
     current_byte = 0
     data = raw_bytes[current_byte]

test/test_encoding.py

@@ -117,6 +117,13 @@ def testFromExample(self):
         res = parquet.encoding.read_bitpacked(fo, count, 3, True)
         self.assertEqual(list(range(8)), res)
 
+    def test_width_zero(self):
+        """Test a zero-width item."""
+        fo = io.BytesIO()
+        count = 1 << 1
+        res = parquet.encoding.read_bitpacked(fo, count, 0, True)
+        self.assertEqual([0] * 8, res)
+
 
 class TestBitPackedDeprecated(unittest.TestCase):
     """Test reading the deprecated bit-packed encoded data."""