Bad data if incomplete tuple

[Kostja Osipov]

I have the following namespace configuration:

namespace[0].enabled = 1
namespace[0].index[0].type = "HASH"
namespace[0].index[0].unique = 1
namespace[0].index[0].key_field[0].fieldno = 0
namespace[0].index[0].key_field[0].type = "STR"
namespace[0].index[1].type = "TREE"
namespace[0].index[1].unique = 0
namespace[0].index[1].key_field[0].fieldno = 1
namespace[0].index[1].key_field[0].type = "STR"

Since this namespace has 2 keys, the minimal cardinality for an acceptable tuple
must be 2. However, Tarantool/Box accepts incomplete tuples:

tarantool> insert into t0 values ('Britney')
Insert OK, 1 row affected

What is worse, these incomplete tuples always match when the search is made on the second key:

tarantool> select * from t0 where k1='Anything'
Found 1 tuple:
['Britney']

tarantool> insert into t0 values ('Stephanie')
Insert OK, 1 row affected
tarantool> select * from t0 where k1='Anything'
Found 2 tuples:
['Britney']
['Stephanie']

Suggested fix:
- do not accept incomplete tuples. Should we still have a namespace option that allows incomplete tuples, they should not match the search pattern on the second key.

[Kostja Osipov (kostja)]

Whether or not the second namespace is unique does not matter, the result is the same.

[Kostja Osipov (kostja)]

The returned result is even more obscure if a disjunct is used:

tarantool> select * from t0 where k1='Anything' or k1='More tuples, please'
Found 2 tuples:
['Stephanie']
['Stephanie']

[Kostja Osipov (kostja)]

If the index is unique, any attempt to insert a tuple with non-zero field in the position of the corrupted index will lead to ERR_CODE_INDEX_VIOLATION, which convinces me that this is a bug.
Additionally, the bug does not occur when there is at least one correct tuple in the namespace already.