[Bug]: load from file,using vector to search failed

[DarthVaderYTX]

Description

`import os.path

import zvec
from zvec import VectorSchema, VectorQuery, FieldSchema

Define collection schema

schema = zvec.CollectionSchema(
name="example",
fields=[
FieldSchema("user_id", data_type=zvec.DataType.STRING, nullable=True),
FieldSchema("name", data_type=zvec.DataType.STRING, nullable=True),
],
vectors=[
zvec.VectorSchema("embedding", zvec.DataType.VECTOR_FP32, 4),
zvec.VectorSchema("embedding2", zvec.DataType.VECTOR_FP32, 4)
],
)

Create collection

path = "./zvec_example"
if os.path.exists(path):
collection = zvec.open(path=path,option=zvec.CollectionOption(read_only=False, enable_mmap=True))
else:
collection = zvec.create_and_open(path=path, schema=schema,option=zvec.CollectionOption(read_only=False, enable_mmap=True))
print(collection.stats)

Insert documents

status = collection.upsert([
zvec.Doc(id="doc_1", vectors={"embedding": [0.1, 0.2, 0.3, 0.4], "embedding2": [0.2, 0.3, 0.4, 0.1]},fields={"user_id":"123","name":"jack"}),
zvec.Doc(id="doc_2", vectors={"embedding": [0.2, 0.3, 0.4, 0.1], "embedding2": [0.1, 0.2, 0.3, 0.4]},fields={"user_id":"456","name":"pony"}),
])
print(status)

Search by vector similarity

results = collection.query(
zvec.VectorQuery("embedding", vector=[0.4, 0.3, 0.3, 0.1]),
topk=10,
include_vector=True,
output_fields=["user_id","name"],
)

Results: list of {'id': str, 'score': float, ...}, sorted by relevance

print(results)
print(len(results))

d = collection.query(include_vector=True)
print(len(d))`

Steps to Reproduce

1. when create the zvec file first, it can search by vector success
2. when load the database from file, search by vector failed

Logs / Stack Trace

Operating System

macOS

Build & Runtime Environment

python 3.10

Additional Context

• I've checked git status — no uncommitted submodule changes
• I built with CMAKE_BUILD_TYPE=Debug
• This occurs with or without COVERAGE=ON
• The issue involves Python ↔ C++ integration (pybind11)

[liututu12]

Hello, when you reopen a collection from a file, you need to first execute the command: del created_coll to release the reference count to the collection object, allowing Python's garbage collection mechanism to reclaim the associated resources at an appropriate time. Otherwise, directly opening the collection might lead to resource conflicts or other issues.
You can try adding the above command; the code reference is as follow：
import os
import shutil
import zvec
from zvec import VectorSchema, VectorQuery, FieldSchema

def test_zvec_issue():
# Define collection schema
schema = zvec.CollectionSchema(
name="example",
fields=[
FieldSchema("user_id", data_type=zvec.DataType.STRING, nullable=True),
FieldSchema("name", data_type=zvec.DataType.STRING, nullable=True),
],
vectors=[
zvec.VectorSchema("embedding", zvec.DataType.VECTOR_FP32, 4),
zvec.VectorSchema("embedding2", zvec.DataType.VECTOR_FP32, 4)
],
)

path = "./zvec_example_test"

# Clean up any existing test data
if os.path.exists(path):
    shutil.rmtree(path)

print("=== Step 1: Creating new zvec collection ===")
collection = zvec.create_and_open(path=path, schema=schema, option=zvec.CollectionOption(read_only=False, enable_mmap=True))
print("Collection stats:", collection.stats)

# Insert documents
status = collection.upsert([
    zvec.Doc(id="doc_1", vectors={"embedding": [0.1, 0.2, 0.3, 0.4], "embedding2": [0.2, 0.3, 0.4, 0.1]}, fields={"user_id": "123", "name": "jack"}),
    zvec.Doc(id="doc_2", vectors={"embedding": [0.2, 0.3, 0.4, 0.1], "embedding2": [0.1, 0.2, 0.3, 0.4]}, fields={"user_id": "456", "name": "pony"}),
])
print("Insert status:", status)

# Search by vector similarity - this should work in step 1
results = collection.query(
    zvec.VectorQuery("embedding", vector=[0.4, 0.3, 0.3, 0.1]),
    topk=10,
    include_vector=True,
    output_fields=["user_id", "name"],
)
print(f"Step 1 search results (should work): {len(results)} results")
print("Results:", results)

# Close the collection
#collection.close()
del collection

print("\n=== Step 2: Loading existing zvec collection ===")
# Load the collection from existing file
collection = zvec.open(path=path, option=zvec.CollectionOption(read_only=False, enable_mmap=True))
print("Loaded collection stats:", collection.stats)

# Search by vector similarity - this is where the issue occurs
try:
    results = collection.query(
        zvec.VectorQuery("embedding", vector=[0.4, 0.3, 0.3, 0.1]),
        topk=10,
        include_vector=True,
        output_fields=["user_id", "name"],
    )
    print(f"Step 2 search results (may fail): {len(results)} results")
    print("Results:", results)
except Exception as e:
    print(f"Step 2 search failed with error: {str(e)}")
    import traceback
    traceback.print_exc()

# Also test querying all documents
try:
    d = collection.query(include_vector=True)
    print(f"All documents query: {len(d)} results")
except Exception as e:
    print(f"All documents query failed: {str(e)}")
    
# Close the collection
#collection.close()
del collection

# Clean up test data
if os.path.exists(path):
    shutil.rmtree(path)

if name == "main":
test_zvec_issue()