diff --git a/mcc.py b/mcc.py
index 4c498efa3d14508f759924138b79e8d1d634cc9c..11d6cb61bd0614389eae4d17f5477af5ff6d2952 100755
--- a/mcc.py
+++ b/mcc.py
@@ -55,21 +55,20 @@ with open(out_file, "r") as file:
next(file) # skip the header's line
cursor.copy_from(file, table_prefix, sep=",", null="", columns=[v.lower() for v in var_names])
+#### CREATION OF D STAR #####
+
missing_selects = []
for v in var_names:
- if is_var(v) and v[1:] not in var_names:
- missing_selects.append("case when i{}=1 then NULL ELSE {} END".format(v,v))
+ if is_var(v) and ("i{}".format(v) in var_names or "I{}".format(v) in var_names):
+ missing_selects.append("case when i{}=1 then NULL ELSE {} END".format(v,v))
elif is_var(v):
missing_selects.append(v)
-#### CREATION OF D STAR #####
-
missing_data_table = "{}_star".format(table_prefix)
deletion_sql = "DROP TABLE IF EXISTS {};".format(missing_data_table)
cursor.execute(deletion_sql)
missing_selects_text = ",".join(missing_selects)
missing_data_sql = "CREATE TABLE {} AS SELECT block, {}, md5(CAST(({}) AS text)) AS superblock FROM {}".format(missing_data_table, missing_selects_text, missing_selects_text, table_prefix)
-print(missing_data_sql)
cursor.execute(missing_data_sql)
#### CREATION OF BID #####
@@ -158,13 +157,21 @@ block_label_table = "{}_block_label".format(table_prefix)
equigraph_table = "{}_equigraph".format(table_prefix)
matching_table = "{}_matching".format(table_prefix)
-nb_tuples = 8
-
-def tuple_joint_probability(id):
- return 2*id/(nb_tuples * (nb_tuples + 1))
+def tuple_joint_probability(tuple):
+ i = bn.completeInstantiation()
+ n = 0
+ p = ie.evidenceJointImpact(vars,{})
+ for v in i.variablesSequence():
+ if n < len(tuple):
+ i.chgVal(v, tuple[n])
+ else:
+ i.chgVal(v, 1)
+ n += 1
+ return p.get(i)
+ # return 2*id/(nb_tuples * (nb_tuples + 1))
-def tuple_occurence_weight(id, occurence):
- return 2*db_size*tuple_joint_probability(id) - 2*occurence +1
+def tuple_occurence_weight(tuple, occurence):
+ return 2*db_size*tuple_joint_probability(tuple) - 2*occurence +1
def initialize_labels():
@@ -173,15 +180,17 @@ def initialize_labels():
cursor.execute("CREATE TABLE {} (tuple INTEGER, occurence INTEGER, label float, weight float, PRIMARY KEY (tuple, occurence)) ".format(tuple_label_table))
insert_sql = "INSERT INTO {}(tuple, occurence, label, weight) VALUES(%s,%s,%s,%s)".format(tuple_label_table)
- cursor.execute("SELECT tuple FROM {}".format(tuple_table))
+ cursor.execute("SELECT * FROM {}".format(tuple_table))
for row in cursor.fetchall():
inserted_row = []
tuple_id = row[0]
+ tuple = row[1:]
inserted_row.append(tuple_id)
inserted_row.append(1)
inserted_row.append(0)
- inserted_row.append(tuple_occurence_weight(tuple_id, 1))
+ inserted_row.append(tuple_occurence_weight(tuple, 1))
+ # inserted_row.append(tuple_id + 2*db_size)
cursor.execute(insert_sql, inserted_row)
deletion_sql = "DROP TABLE IF EXISTS {};".format(block_label_table)
@@ -202,7 +211,7 @@ def initialize_matching():
# return a block free vertex in the equigraph, if it exists
def pick_free_block():
- cursor.execute("SELECT block FROM {} WHERE max_occurence > 0 AND block NOT IN (SELECT block FROM {})".format(equigraph_table, matching_table))
+ cursor.execute("SELECT block FROM {} WHERE block NOT IN (SELECT block FROM {})".format(block_label_table, matching_table))
res = cursor.fetchone()
return res[0] if res is not None else None
@@ -231,76 +240,88 @@ def augmenting_path_insertion(u, y, prec):
def update_labels(S, T):
cursor.execute("""
- SELECT e.block, e.tuple, e.max_occurence + 1, tl.weight - 2, tb.label, tl.label
- FROM {} AS e, {} AS tl, {} AS tb
- WHERE e.block IN %s AND e.max_occurence < %s AND e.max_occurence != 0 AND tl.tuple = e.tuple
- AND tl.occurence = e.max_occurence AND tb.block = e.block
+ SELECT e.block, e.tuple, e.max_occurence + 1, tl2.weight - 2, tb.label, COALESCE(tl.label, 0)
+ FROM ({} AS e LEFT OUTER JOIN {} AS tl ON tl.occurence = e.max_occurence + 1 AND tl.tuple = e.tuple), {} AS tb, {} as tl2
+ WHERE e.block IN %s
+ AND e.max_occurence < %s AND e.max_occurence != 0
+ AND tb.block = e.block AND tl2.occurence = e.max_occurence AND tl2.tuple = e.tuple
UNION ALL
SELECT e.block, e.tuple, 1, tl.weight, tb.label, tl.label
FROM {} AS e, {} AS tl, {} AS tb
- WHERE e.block IN %s AND e.max_occurence = 0 AND tl.tuple = e.tuple
- AND tl.occurence = 1 AND tb.block = e.block""".format(equigraph_table, tuple_label_table, block_label_table, equigraph_table, tuple_label_table, block_label_table), [tuple(S), db_size, tuple(S)])
+ WHERE e.block IN %s
+ AND e.max_occurence = 0 AND tl.tuple = e.tuple
+ AND tl.occurence = 1 AND tb.block = e.block""".format(equigraph_table, tuple_label_table, block_label_table, tuple_label_table, equigraph_table, tuple_label_table, block_label_table), [tuple(S), db_size, tuple(S)])
+
alpha = sys.float_info.max
edges_to_add = []
for row in cursor.fetchall():
+ if (row[1], row[2]) in T:
+ continue
a = row[4] + row[5] - row[3]
if a < alpha :
alpha = a
edges_to_add = []
if a == alpha:
edges_to_add.append(row)
- # print(alpha, edges_to_add)
+
cursor.execute("UPDATE {} SET label = label - %s WHERE block in %s".format(block_label_table), [alpha, tuple(S)])
- cursor.execute("UPDATE {} SET label = label + %s WHERE (tuple, occurence) in %s".format(tuple_label_table), [alpha, tuple(T)])
+ if len(T) > 0:
+ cursor.execute("UPDATE {} SET label = label + %s WHERE (tuple, occurence) in %s".format(tuple_label_table), [alpha, tuple(T)])
+ cursor.execute("UPDATE {} SET max_occurence = CASE WHEN max_occurence > 0 THEN max_occurence -1 ELSE 0 END WHERE (tuple, max_occurence) in %s AND block NOT IN %s".format(equigraph_table), [tuple(T), tuple(S)])
+
for edge in edges_to_add:
- cursor.execute("INSERT INTO {} (block, label) VALUES (%s, %s)".format(block_label_table), [row[0], row[4]])
- cursor.execute("INSERT INTO {} (tuple, occurence, label, weight) VALUES (%s, %s, %s, %s) ON CONFLICT DO NOTHING".format(tuple_label_table), [row[1], row[2], row[5], row[3]])
- cursor.execute("UPDATE {} SET max_occurence = %s WHERE block = %s AND tuple = %s".format(equigraph_table), [row[2], row[0], row[1]])
+ cursor.execute("INSERT INTO {} (tuple, occurence, label, weight) VALUES (%s, %s, %s, %s) ON CONFLICT DO NOTHING".format(tuple_label_table), [edge[1], edge[2], edge[5], edge[3]])
+ cursor.execute("UPDATE {} SET max_occurence = %s WHERE block = %s AND tuple = %s".format(equigraph_table), [edge[2], edge[0], edge[1]])
+ # input("updated labels with S={}, T={} and alpha={} \n with edges {}".format(S,T, alpha, edges_to_add))
+ return list(map(lambda e: [e[0], (e[1],e[2])] ,edges_to_add))
+
def hungarian_step(u, S, T, S_neigh, prec):
- update_count = 0
- prec_z = u
while True :
remaining_neigh = S_neigh - T
if len(remaining_neigh) > 0:
y = next(iter(remaining_neigh))
- prec[y] = prec_z
+ if len(S) == 1:
+ prec[y] = u
z = get_matching_block(y[0], y[1])
if z is None :
+ print("augmenting path from {} to {} with S={} and T={} \n".format(u, y, S, T))
augmenting_path_insertion(u, y, prec)
- print(update_count)
return
else:
S.add(z)
T.add(y)
prec[z] = y
- S_neigh.update(get_neighbors({z}))
+ for nz in get_neighbors({z}):
+ S_neigh.add(nz)
+ if nz != y and nz not in prec :
+ prec[nz] = z
+ # input("matched y={} by z={} with S={}, T={} and neigh={}".format(y,z,S,T, S_neigh))
else:
- update_count+=1
- # print("update labels")
- # print(v, S, T, S_neigh, prec)
- update_labels(S, T)
- S_neigh = get_neighbors(S)
+ new_edges = update_labels(S, T)
+ for e in new_edges:
+ S_neigh.add(e[1])
+ if e[1] not in prec:
+ prec[e[1]] = e[0]
+
def hungarian_algorithm():
- for i in range(1, db_size):
- v = pick_free_block()
- S = {v}
+ for i in range(0, db_size):
+ u = pick_free_block()
+ S = {u}
T = set()
S_neigh = get_neighbors(S)
prec = dict()
- # print(v, S, T, S_neigh, prec)
- hungarian_step(v, S, T, S_neigh, prec)
- cursor.execute("SELECT * FROM {}".format(matching_table))
- # print(cursor.fetchall())
+ hungarian_step(u, S, T, S_neigh, prec)
initialize_labels()
initialize_equigraph()
initialize_matching()
+input("starting hungarian algorithm")
hungarian_algorithm()
###### Imputation