when insert vectors with the same id, milvus won't check duplicate IDs and store them all.

[zhudd-hub]

向milvus插入特征向量，指定特征的ID，将该特征重复多次插入，然后创建index，然后search，发现top K结果严重不全
A clear and concise description of what the bug is.

完整的实验代码如下

param = {'collection_name': 'test','dimension': 512}
milvus.create_collection(param)
feats=[[random.random() for _ in range(512)] for _ in range(2)]
for _ in range(4):
    status, ids = milvus.insert(collection_name=collection_name, records=feats, ids=[1,9])
    print(status)
param = {'collection_name': 'test','query_records': feats[0:1],'top_k': 4,'params': {}}
status, results = milvus.search(**param)
print(status)
results

————————————————————
代码输出：

Status(code=0, message='Add vectors successfully!')
Status(code=0, message='Add vectors successfully!')
Status(code=0, message='Add vectors successfully!')
Status(code=0, message='Add vectors successfully!')

[
[
(id:1, distance:0.0),
(id:1, distance:0.0),
(id:1, distance:0.0),
(id:1, distance:0.0)
]
]

期待只输出两个结果，其ID分别为1和9

[BossZou]

@zhudd-hub , the duplicate ID check is not supported currently, so if user want to specify ID during inserting, they should make sure ID is unique.

[keotic]

What about the same id to different partitions on the same collection?

[XuanYang-cn]

@keotic Looks ok to me.