From f7ca3c09f6ae974f085afd14a75f4cd4692ab51b Mon Sep 17 00:00:00 2001
From: "zhenshan.cao" <zhenshan.cao@zilliz.com>
Date: Tue, 30 May 2023 17:17:45 +0800
Subject: [PATCH] Fix bug: failed to insert binary vector

Signed-off-by: zhenshan.cao <zhenshan.cao@zilliz.com>
---
 examples/binary_example.py       | 69 ++++++++++++++++++++++++++++++++
 pymilvus/client/entity_helper.py | 10 ++---
 2 files changed, 74 insertions(+), 5 deletions(-)
 create mode 100644 examples/binary_example.py

diff --git a/examples/binary_example.py b/examples/binary_example.py
new file mode 100644
index 000000000..80e5da627
--- /dev/null
+++ b/examples/binary_example.py
@@ -0,0 +1,69 @@
+import time
+import random
+import numpy as np
+from pymilvus import (
+     connections,
+     utility,
+     FieldSchema, CollectionSchema, DataType,
+     Collection,
+ )
+
+
+bin_index_types = ["BIN_FLAT", "BIN_IVF_FLAT"]
+
+default_bin_index_params = [{"nlist": 128}, {"nlist": 128}]
+
+def gen_binary_vectors(num, dim):
+    raw_vectors = []
+    binary_vectors = []
+    for _ in range(num):
+        raw_vector = [random.randint(0, 1) for _ in range(dim)]
+        raw_vectors.append(raw_vector)
+        # packs a binary-valued array into bits in a unit8 array, and bytes array_of_ints
+        binary_vectors.append(bytes(np.packbits(raw_vector, axis=-1).tolist()))
+    return raw_vectors, binary_vectors
+
+
+def binary_vector_search():
+    connections.connect()
+    int64_field = FieldSchema(name="int64", dtype=DataType.INT64, is_primary=True, auto_id=True)
+    dim = 128
+    nb = 3000
+    vector_field_name = "binary_vector"
+    binary_vector = FieldSchema(name=vector_field_name, dtype=DataType.BINARY_VECTOR, dim=dim)
+    schema = CollectionSchema(fields=[int64_field, binary_vector], enable_dynamic_field=True)
+
+    has = utility.has_collection("hello_milvus")
+    if has:
+        hello_milvus = Collection("hello_milvus_bin")
+        hello_milvus.drop()
+    else:
+        hello_milvus = Collection("hello_milvus_bin", schema)
+
+    _, vectors = gen_binary_vectors(nb, dim)
+    rows = [
+        {vector_field_name: vectors[0]},
+        {vector_field_name: vectors[1]},
+        {vector_field_name: vectors[2]},
+        {vector_field_name: vectors[3]},
+        {vector_field_name: vectors[4]},
+        {vector_field_name: vectors[5]},
+    ]
+
+    hello_milvus.insert(rows)
+    hello_milvus.flush()
+    for i, index_type in enumerate(bin_index_types):
+        index_params = default_bin_index_params[i]
+        hello_milvus.create_index(vector_field_name,
+                                  index_params={"index_type": index_type, "params": index_params, "metric_type": "HAMMING"})
+        hello_milvus.load()
+        print("index_type = ", index_type)
+        res = hello_milvus.search(vectors[:1], vector_field_name, {"metric_type": "HAMMING"}, limit=1)
+        print("res = ", res)
+        hello_milvus.release()
+        hello_milvus.drop_index()
+    hello_milvus.drop()
+
+
+if __name__ == "__main__":
+    binary_vector_search()
diff --git a/pymilvus/client/entity_helper.py b/pymilvus/client/entity_helper.py
index cd9508a74..4c1002d62 100644
--- a/pymilvus/client/entity_helper.py
+++ b/pymilvus/client/entity_helper.py
@@ -84,7 +84,7 @@ def pack_field_value_to_field_data(field_value, field_data, field_info):
         field_data.vectors.float_vector.data.extend(field_value)
     elif field_type in (DataType.BINARY_VECTOR,):
         field_data.vectors.dim = len(field_value) * 8
-        field_data.vectors.binary_vector.data.append(b''.join(field_value))
+        field_data.vectors.binary_vector += bytes(field_value)
     elif field_type in (DataType.VARCHAR,):
         field_data.scalars.string_data.data.append(
             convert_to_str_array(field_value, field_info, True))
@@ -204,10 +204,10 @@ def extract_row_data_from_fields_data(fields_data, index, dynamic_output_fields=
                                                           start_pos:end_pos]]
         elif field_data.type == DataType.BINARY_VECTOR:
             dim = field_data.vectors.dim
-            if len(field_data.vectors.binary_vector.data) >= index * (dim / 8):
-                start_pos = index * (dim / 8)
-                end_pos = (index + 1) * (dim / 8)
+            if len(field_data.vectors.binary_vector) >= index * (dim // 8):
+                start_pos = index * (dim // 8)
+                end_pos = (index + 1) * (dim // 8)
                 entity_row_data[field_data.field_name] = [
-                    field_data.vectors.binary_vector.data[start_pos:end_pos]]
+                    field_data.vectors.binary_vector[start_pos:end_pos]]
 
     return entity_row_data