flesh out voting module

bigchaindb/voting.py

@@ -1,25 +1,41 @@
+"""
+Everything to do with creating and checking votes.
+All functions in this module should be referentially transparent, that is,
+they always give the same output for a given input. This makes it easier
+to test.
+"""
 import collections
 
 
-def filter_eligible_votes(votes, block_voters, keyring, check_signature):
+def partition_eligible_votes(votes, eligible_voters, verify_vote_signature):
     """
     Filter votes from unknown nodes or nodes that are not listed on
-    block. Here is our sybill protection.
+    block. This is the primary Sybill protection.
     """
-    eligible_voters = set(keyring) & set(block_voters)
-    eligible_votes = []
+    eligible, ineligible = ([], [])
 
     for vote in votes:
-        pubkey = vote['node_pubkey']
-        voter_eligible = pubkey in eligible_voters
-        sig_legit = sig_is_legit(vote)
-        if voter_eligible and sig_legit:
-            eligible_votes[pubkey].append(vote)
+        voter_eligible = vote['node_pubkey'] in eligible_voters
+        if voter_eligible and verify_vote_signature(vote):
+            eligible.append(vote)
+        else:
+            ineligible.append(vote)
 
-    return eligible_votes
+    return eligible, ineligible
 
 
-def count_votes(eligible_votes, check_schema):
+def count_votes(eligible_votes):
+    """
+    Given a list of eligible votes, (votes from known nodes that are listed
+    as voters), count the votes to produce three quantities:
+
+        Number of votes that say valid
+        Number of votes that say invalid
+        Highest agreement on previous block ID
+
+    Also, detect if there are multiple votes from a single node and return them
+    in a separate "cheat" dictionary.
+    """
     by_voter = collections.defaultdict(list)
     for vote in eligible_votes:
         by_voter[vote['node_pubkey']].append(vote)
@@ -27,9 +43,11 @@ def count_votes(eligible_votes, check_schema):
     n_valid = 0
     n_invalid = 0
     prev_blocks = collections.Counter()
+    cheat = {}
 
     for pubkey, votes in by_voter.items():
-        if len(votes) > 1 or not schema_is_correct(votes[0]):
+        if len(votes) > 1:
+            cheat[pubkey] = votes
             n_invalid += 1
             continue
 
@@ -41,7 +59,41 @@ def count_votes(eligible_votes, check_schema):
             n_invalid += 1
 
     return {
-        'valid': n_valid,
-        'invalid': n_invalid,
-        'prev_block': prev_blocks.most_common()[0]
-    }
+        'n_valid': n_valid,
+        'n_invalid': n_invalid,
+        'n_agree_prev_block': prev_blocks.most_common()[0][1]
+    }, cheat
+
+
+def decide_votes(n_voters, n_valid, n_invalid, n_agree_prev_block):
+    """
+    Decide on votes.
+
+    To return VALID there must be a clear majority that say VALID
+    and also agree on the previous block. This is achieved using the > operator.
+
+    A tie on an even number of votes counts as INVALID so the >= operator is
+    used.
+    """
+
+    # Check insane cases. This is basic, not exhaustive.
+    if n_valid + n_invalid > n_voters or n_agree_prev_block > n_voters:
+        raise ValueError('Arguments not sane: %s' % {
+            'n_voters': n_voters,
+            'n_valid': n_valid,
+            'n_invalid': n_invalid,
+            'n_agree_prev_block': n_agree_prev_block,
+        })
+
+    if n_invalid * 2 >= n_voters:
+        return INVALID
+    if n_valid * 2 > n_voters:
+        if n_agree_prev_block * 2 > n_voters:
+            return VALID
+        return INVALID
+    return UNDECIDED
+
+
+INVALID = 'invalid'
+VALID = TX_VALID = 'valid'
+UNDECIDED = TX_UNDECIDED = 'undecided'

tests/test_voting.py

@@ -0,0 +1,61 @@
+import pytest
+
+from bigchaindb.core import Bigchain
+from bigchaindb.voting import (count_votes, partition_eligible_votes,
+                               decide_votes, INVALID, VALID, UNDECIDED)
+
+
+def test_partition_eligible_votes():
+    nodes = list(map(Bigchain, 'abc'))
+    votes = [n.vote('block', 'a', True) for n in nodes]
+
+    el, inel = partition_eligible_votes(votes, 'abc', lambda _: True)
+
+    assert el == votes
+    assert inel == []
+
+
+def test_count_votes():
+    nodes = list(map(Bigchain, 'abc'))
+    votes = [n.vote('block', 'a', True) for n in nodes]
+
+    assert count_votes(votes) == ({
+        'n_valid': 3,
+        'n_invalid': 0,
+        'n_agree_prev_block': 3
+    }, {})
+
+
+DECISION_TESTS = [dict(
+    zip(['n_voters', 'n_valid', 'n_invalid', 'n_agree_prev_block'], t))
+    for t in [
+         (1,          1,         1,           1),
+         (2,          2,         1,           2),
+         (3,          2,         2,           2),
+         (4,          3,         2,           3),
+         (5,          3,         3,           3),
+         (6,          4,         3,           4),
+         (7,          4,         4,           4),
+         (8,          5,         4,           5),
+    ]
+]
+
+
+@pytest.mark.parametrize('kwargs', DECISION_TESTS)
+def test_decide_votes_valid(kwargs):
+    kwargs = kwargs.copy()
+    kwargs['n_invalid'] = 0
+    assert decide_votes(**kwargs) == VALID
+    kwargs['n_agree_prev_block'] -= 1
+    assert decide_votes(**kwargs) == INVALID
+    kwargs['n_valid'] -= 1
+    assert decide_votes(**kwargs) == UNDECIDED
+
+
+@pytest.mark.parametrize('kwargs', DECISION_TESTS)
+def test_decide_votes_invalid(kwargs):
+    kwargs = kwargs.copy()
+    kwargs['n_valid'] = 0
+    assert decide_votes(**kwargs) == INVALID
+    kwargs['n_invalid'] -= 1
+    assert decide_votes(**kwargs) == UNDECIDED