Improved filtering capabilities

Oct 13, 2025 by pynkocean.com 39 views

Introduction

Milvus, a popular open-source vector database, has been gaining traction in the industry due to its high-performance capabilities and scalability. One of the key features of Milvus is its indexing mechanism, which enables efficient filtering and querying of large datasets. However, the current filtering capabilities of Milvus are limited to a specific set of data types and operations. In this article, we propose the introduction of two new indexing mechanisms: BF (Bloom Filter) and n-gram Skip filters. These features will significantly enhance the filtering capabilities of Milvus, enabling users to perform more complex queries and improve the overall performance of their applications.

Problem Statement

The current filtering mechanism in Milvus only supports limited data types and operations, such as min/max filtering. This limitation restricts the types of queries that can be performed on the database, making it less flexible and less powerful compared to other vector databases. For instance, users cannot perform queries like column = X or column IN ('a', 'b', 'c') using the current filtering mechanism.

Solution Overview

To address the limitations of the current filtering mechanism, we propose the introduction of two new indexing mechanisms: BF and n-gram Skip filters.

BF Index

A BF index is a type of indexing mechanism that uses a Bloom Filter to accelerate queries like column = X or column IN ('a', 'b', 'c'). A Bloom Filter is a space-efficient data structure that can be used to test whether an element is a member of a set. In the context of Milvus, a BF index can be used to quickly determine whether a particular value exists in a column, without having to scan the entire column.

The BF index is created as a statistic, which means it does not conflict with other inverted indexes. This is a significant advantage, as it allows users to create multiple indexes on the same column without worrying about conflicts.

n-gram Index

An n-gram index is a type of indexing mechanism that uses a combination of n-grams and a Skip filter to accelerate queries like LIKE '%error%'. An n-gram is a contiguous sequence of n items from a string of text. In the context of Milvus, an n-gram index can be used to quickly determine whether a particular string contains a specific pattern.

The n-gram index uses a Skip filter to accelerate the query process. A Skip filter is a data structure that allows users to quickly determine whether a particular value exists in a column, without having to scan the entire column.

Benefits

The introduction of BF and n-gram Skip filters will bring several benefits to Milvus users.

Improved filtering capabilities: With the introduction of BF and n-gram Skip filters, users will be able to perform more complex queries on their data, such as column = X or column IN ('a', 'b', 'c').
Enhanced performance: The BF and n-gram Skip filters will significantly improve the performance of Milvus, enabling users to query their data more efficiently.
Increased flexibility: The introduction of BF and n-gram Skip filters will make Milvus more flexible and powerful, enabling users to perform a wider range of queries on their data.

Implementation

The implementation of BF and n-gram Skip filters will involve the following steps:

Design and development: The design and development of the BF and n-gram Skip filters will involve creating a new indexing mechanism that can be used to accelerate queries on Milvus.
Testing and validation: The BF and n-gram Skip filters will be tested and validated to ensure that they work correctly and efficiently.
Integration with Milvus: The BF and n-gram Skip filters will be integrated with Milvus, enabling users to create and use these indexes on their data.

Conclusion

The introduction of BF and n-gram Skip filters will significantly enhance the filtering capabilities of Milvus, enabling users to perform more complex queries and improve the overall performance of their applications. The benefits of this feature include improved filtering capabilities, enhanced performance, and increased flexibility. The implementation of this feature will involve designing and developing the BF and n-gram Skip filters, testing and validating them, and integrating them with Milvus.

Future Work

The introduction of BF and n-gram Skip filters is just the first step in enhancing the filtering capabilities of Milvus. Future work will involve exploring other indexing mechanisms, such as prefix trees and suffix trees, to further improve the performance and flexibility of Milvus.

References

[1] Bloom, B. H. (1970). Space/time trade-offs in hash coding with allowable errors. Communications of the ACM, 13(7), 422-426.
[2] Navarro, G. (2001). A guided tour to approximate string matching. ACM Computing Surveys, 33(1), 31-88.

Appendix

The following is a list of open issues related to this feature request:

[Issue 1]: Implement BF index
[Issue 2]: Implement n-gram index
[Issue 3]: Integrate BF and n-gram indexes with Milvus

The following is a list of tasks related to this feature request:

[Task 1]: Design and develop BF index
[Task 2]: Design and develop n-gram index
[Task 3]: Test and validate BF and n-gram indexes
[Task 4]: Integrate BF and n-gram indexes with Milvus
Q&A: BF and n-gram Skip Filters in Milvus =============================================

Introduction

In our previous article, we proposed the introduction of BF (Bloom Filter) and n-gram Skip filters in Milvus to enhance the filtering capabilities of the database. In this article, we will answer some of the frequently asked questions related to this feature request.

Q: What is a BF index and how does it work?

A: A BF index is a type of indexing mechanism that uses a Bloom Filter to accelerate queries like column = X or column IN ('a', 'b', 'c'). A Bloom Filter is a space-efficient data structure that can be used to test whether an element is a member of a set. In the context of Milvus, a BF index can be used to quickly determine whether a particular value exists in a column, without having to scan the entire column.

Q: What is an n-gram index and how does it work?

A: An n-gram index is a type of indexing mechanism that uses a combination of n-grams and a Skip filter to accelerate queries like LIKE '%error%'. An n-gram is a contiguous sequence of n items from a string of text. In the context of Milvus, an n-gram index can be used to quickly determine whether a particular string contains a specific pattern.

Q: How do BF and n-gram indexes improve the performance of Milvus?

A: The introduction of BF and n-gram indexes will significantly improve the performance of Milvus, enabling users to query their data more efficiently. By using a Bloom Filter or a combination of n-grams and a Skip filter, users can quickly determine whether a particular value or pattern exists in a column, without having to scan the entire column.

Q: Are BF and n-gram indexes compatible with other indexing mechanisms in Milvus?

A: Yes, BF and n-gram indexes are compatible with other indexing mechanisms in Milvus. The BF index is created as a statistic, which means it does not conflict with other inverted indexes. This is a significant advantage, as it allows users to create multiple indexes on the same column without worrying about conflicts.

Q: How do I create a BF index in Milvus?

A: To create a BF index in Milvus, you will need to follow these steps:

Design and develop the BF index: The first step is to design and develop the BF index. This will involve creating a new indexing mechanism that can be used to accelerate queries on Milvus.
Test and validate the BF index: The next step is to test and validate the BF index to ensure that it works correctly and efficiently.
Integrate the BF index with Milvus: The final step is to integrate the BF index with Milvus, enabling users to create and use this index on their data.

Q: How do I create an n-gram index in Milvus?

A: To create an n-gram index in Milvus, you will need to follow these steps:

Design and develop the n-gram index: The first step is to design and develop the n-gram index. This will involve creating a new indexing mechanism that can be used to accelerate queries on Milvus.
Test and validate the n-gram index: The next step is to test and validate the n-gram index to ensure that it works correctly and efficiently.
Integrate the n-gram index with Milvus: The final step is to integrate the n-gram index with Milvus, enabling users to create and use this index on their data.

Q: What are the benefits of using BF and n-gram indexes in Milvus?

A: The benefits of using BF and n-gram indexes in Milvus include:

Improved filtering capabilities: With the introduction of BF and n-gram indexes, users will be able to perform more complex queries on their data, such as column = X or column IN ('a', 'b', 'c').
Enhanced performance: The BF and n-gram indexes will significantly improve the performance of Milvus, enabling users to query their data more efficiently.
Increased flexibility: The introduction of BF and n-gram indexes will make Milvus more flexible and powerful, enabling users to perform a wider range of queries on their data.

Conclusion

In this article, we have answered some of the frequently asked questions related to the introduction of BF and n-gram Skip filters in Milvus. We hope that this article has provided you with a better understanding of how these features work and how they can improve the performance and flexibility of Milvus.