Unpack for binary data

[copyhold]

Hi,

My data in tarantool stored as binary strings, they are not UTF in any manner. It's just a sequence of bytes.
When trying to select data with this module , unpacker fails with error:

Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
 File "/usr/local/lib/python3.6/dist-packages/tarantool/space.py", line 75, in select
 return self.connection.select(self.space_no, *args, **kwargs)
 File "/usr/local/lib/python3.6/dist-packages/tarantool/connection.py", line 775, in select
 response = self._send_request(request)
 File "/usr/local/lib/python3.6/dist-packages/tarantool/connection.py", line 357, in _send_request
 return self._send_request_wo_reconnect(request)
 File "/usr/local/lib/python3.6/dist-packages/tarantool/connection.py", line 264, in _send_request_wo_reconnect
 response = Response(self, self._read_response())
 File "/usr/local/lib/python3.6/dist-packages/tarantool/response.py", line 62, in __init__
 self._body = unpacker.unpack()
 File "msgpack/_unpacker.pyx", line 519, in msgpack._unpacker.Unpacker.unpack
 File "msgpack/_unpacker.pyx", line 499, in msgpack._unpacker.Unpacker._unpack
msgpack.exceptions.UnpackValueError: 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

Can this be resolved somehow?

[31hkim]

Any success with fixing that? Still have that problem with Tarantool 2.2 and connector 0.6.6 even when I don't select binary data from a space containing it

[Totktonada]

It seems, we cannot just set encoding to None when creating a connection, because at least schema reload does not support this.

However we can temporary set encoding to None:

diff --git a/unit/suites/test_dml.py b/unit/suites/test_dml.py
index 31e821e..1264180 100644
--- a/unit/suites/test_dml.py
+++ b/unit/suites/test_dml.py
@@ -70,6 +70,12 @@ class TestSuite_Request(unittest.TestCase):
 self.con.insert('space_1', [i, i%5, 'tuple_'+str(i)])[0],
 [i, i%5, 'tuple_'+str(i)]
 )
+ saved_encoding = self.con.encoding
+ self.con.encoding = None
+ t = [1001, 1, b'\xff']
+ self.assertEqual(self.con.insert('space_1', t)[0], t)
+ self.con.encoding = saved_encoding
+
 def test_00_03_answer_repr(self):
 repr_str = """- [1, 1, 'tuple_1']"""
 self.assertEqual(repr(self.con.select('space_1', 1)), repr_str)
@@ -122,6 +128,12 @@ class TestSuite_Request(unittest.TestCase):
 [[200, 0, 'tuple_200'], [205, 0, 'tuple_205']]
 )
 
+ saved_encoding = self.con.encoding
+ self.con.encoding = None
+ res = self.con.select('space_1', 1001)
+ self.assertSequenceEqual(res, [[1001, 1, b'\xff']])
+ self.con.encoding = saved_encoding
+
 def test_03_delete(self):
 # Check that delete works fine
 self.assertSequenceEqual(self.con.delete('space_1', 20), [[20, 0, 'tuple_20']])

NB: File another issue for Connection(host, port, encoding=None) support.

[31hkim]

Now I'm getting this:

>> print(result)
- None
- b'Type mismatch: can not convert 'REDACTED BINARY STRING' to varbinary'

And honestly I don't understand what encodings have to do when it comes to raw binary data. Like shouldn't it be just directly converted to bytes Python type?

[Totktonada]

Okay, two different problems are discussed here:

1. A user store binary (non-utf-8 data) in a 'string' field and don't want data being encoded or decoded when working with this field. May be worked around as shown above.
2. A user want to send / receive mp_bin to work with a 'varbinary' field. Everything looks okay on decoding side (mp_bin is decoded to bytes), but a bytes value is encoded as mp_str. Maybe we can change it on Python 3 (RFC patch).

[31hkim]

Ah, sorry for confusion, and thank you. Will wait for a complete fix

[31hkim]

Any success?

[Totktonada]

We're at planning of Q3 tasks and I hope this one will be included here. I'll specifically mark it as important.

[Totktonada]

NB: Propose to msgpack-python to provide helpers like msgpack.as_bin() and msgpack.as_str() to encode a particular structure (dict, list, tuple) member as mp_bin or mp_str disregarding use_bin_type option. It may be requirement of an msgpack-based protocol to have a string here and a binary value there (exactly our case).

Usage example:

import msgpack
msgpack.dumps(dict(foo='string', bar=msgpack.as_bin('binary')))