core.py: Enable correct exporting of char data

As the data type for saving the char in pyulog was np.int8, the topics
that had fields with characters weren't able to get encoded (with error:
pyulog struct.error: char format requires a bytes object of length 1).

Therefore, this enables manual conversion of np.int8 object into bytes()
with single character, so that struct.pack command works as expected.

Notable example of where this is needed is the "transponder_report"
topic (included in PX4 v1.14)'s callsign[] character array. Before this
fix, the "write_ulog()" function would fail as the struct.pack couldn't
correctly encode the characters stored in int8 type

Co-authored-by: Junwoo Hwang <[email protected]>

pyulog/core.py

@@ -432,6 +432,12 @@ def _make_data_items(self):
                     field_type = field.type_str
                     field_encoding = self._UNPACK_TYPES[field_type][0]
                     field_data = data_set.data[field_name][i_sample]
+
+                    # For char type, convert np.int8 into single bytes() object
+                    # so that struct.pack can handle it
+                    if field_encoding == 'c':
+                        field_data = bytes(chr(field_data), 'utf-8')
+
                     data.extend(struct.pack('<' + field_encoding, field_data))
 
                 header = struct.pack('<HB', len(data), self.MSG_TYPE_DATA)