April 2022 Summaries

Kostas Kyrimis, a member of Memgraph's Core team, is highlighted in a blog post for his contributions and positive demeanor after completing one year with the company. Known for his enthusiasm and community involvement as a C++ engineer, Kostas is described as a cheerful presence who uplifts his colleagues daily. The blog invites readers to learn more about his reasons for joining Memgraph, his current projects, and his views on remote work through a video interview. Additionally, the Core team is seeking more passionate C++ engineers to tackle complex and engaging challenges.

Antonio Filipovic's blog post discusses the development and application of Temporal Graph Neural Networks (TGNs) using PyTorch at Memgraph, specifically focusing on creating a simple recommendation engine based on an Amazon dataset. The article explains how TGNs, which are designed to handle continuous-time dynamic graphs, enhance traditional graph neural networks by allowing them to process temporal data, making them suitable for real-time applications like recommendation systems. The post provides a detailed tutorial on setting up Memgraph using Docker and GQLAlchemy to analyze Amazon user-item reviews, emphasizing the importance of node embeddings and message passing for link prediction tasks. It highlights the process of training and evaluating TGNs, detailing the steps to configure parameters, import queries, and visualize results to measure the precision of the model. The author illustrates how TGNs can be utilized to predict user preferences, underscoring the potential of TGNs to revolutionize dynamic, real-time data analysis in graph-based applications.

The Graph Data Zagreb meetup, a community for graph enthusiasts in Croatia, is hosting its third edition on May 4 at WESPA Spaces in Zagreb with a focus on education. Katarina Šupe, a Developer Relations Engineer from Memgraph, will present a talk titled "Game of Graphs - Graph analytics on a GoT dataset," where she will use graph algorithms to analyze a Game of Thrones dataset, revealing insights such as which characters were responsible for the most deaths and how various characters are connected. The event encourages participation and feedback from attendees, inviting them to suggest future topics and join the community on Discord.

Ante Pusic's blog post explores the impact of disruptions in the global submarine internet cable network using dynamic betweenness centrality, particularly when connections are lost due to unforeseen events like shark attacks. By employing Memgraph's graph analytics capabilities, including online graph algorithms like Node2Vec and PageRank, the analysis focuses on updating previous outputs rather than recalculating everything, thus allowing a rapid response to network changes. By utilizing Docker, GQLAlchemy, and Jupyter Notebook, the blog details setting up Memgraph and analyzing the submarine cable network graph, with nodes representing landing points and edges representing cables. The iCentral algorithm is discussed as a new advancement, enabling efficient recalculations of betweenness centrality scores by focusing only on affected areas of the graph, thereby improving performance and scalability. The post illustrates this with an example where a cable cut between Tuas, Singapore, and Jeddah, Saudi Arabia, prompts updates in centrality scores, highlighting the dynamic nature of internet traffic flow. The blog concludes by emphasizing the importance of online algorithms for scaling graph analytics and invites community engagement through Memgraph's open-source projects and Discord channel.

Memgraph's MAGE 1.2 release brings significant enhancements to its graph analytics library, introducing both dynamic and static graph algorithms and marking its foray into graph machine learning. This update features new dynamic algorithms—Temporal Graph Networks, Dynamic Betweenness Centrality, and Dynamic Katz Centrality—designed to improve computation efficiency by focusing on updated graph sections rather than recalculating entire graphs. Temporal Graph Networks, powered by PyTorch, enable deep learning on dynamic graphs, offering capabilities like handling node and edge updates for predictive tasks. Additionally, the library expands its static algorithm offerings with Louvain community detection, maximum flow calculation, and static Katz centrality. The release also improves utility functions for importing and exporting graphs using JSON, with plans for broader format support in future updates. Looking ahead, Memgraph plans to enhance scalability with GPU algorithms, introduce custom functions via a C API, and continue integrating machine learning into graph analytics.

Benchmarking databases like Memgraph and Neo4j can be a complex and time-consuming task, especially when trying to set up a "production-like" scenario. Benchgraph, a tool used for executing benchmarks, can be integrated into a CI/CD environment to simplify running benchmarks on these databases with user-specific hardware, workloads, and conditions. To set up a benchmark, users need at least Python 3.7, Docker, and basic Python knowledge. They must create a Python script that inherits the workload class and define methods for dataset and index generation, as well as the queries to be benchmarked. Benchmarks are executed using a command line interface, and results can be exported and compared using an HTML file for a visual representation. The tool allows for configuration through various arguments to simulate different conditions and can benchmark different database vendors by adjusting queries and indexes accordingly. Users are encouraged to share their results and engage with the community for support and further exploration.

At the second Graph Data Zagreb meetup held on April 6, 2022, Goran Cvijanović presented on using graph databases to detect vulnerabilities in GitHub repositories, focusing on mitigating the rise of dependency chain attacks prevalent in NPM and PyPI package repositories. The talk highlighted how visualizing software structures as graphs can help identify components exposed to CVE security issues, despite the complexity added by dependencies on other repositories and libraries. Goran introduced Code Graph, an online service leveraging Python, JavaScript, and Memgraph for analyzing public GitHub repositories for cybersecurity vulnerabilities. The event also provided attendees with networking opportunities and encouraged further engagement through Memgraph's Discord server and the Graph Data Zagreb community.

The blog post by Katarina Supe provides a comprehensive guide on using the GQLAlchemy query builder to efficiently create, modify, retrieve, and manage Cypher queries within the Memgraph database. It begins with instructions on setting up the necessary environment, including the installation of Docker and GQLAlchemy, due to Memgraph's compatibility requirements with Linux. The post details the creation of graph schemas using Python classes to ensure object types are correctly returned, followed by practical examples of creating nodes and relationships such as users and movies, and connecting them with relationships like "RATED." Additionally, it covers filtering techniques to refine query results, as well as methods to delete nodes and relationships or remove properties from the database. The blog also explores using the query builder to determine the most significant movie in the network based on user ratings. Throughout, readers are encouraged to experiment with Memgraph through Memgraph Lab and seek assistance or explore use cases in the Memgraph Playground or Discord community.

GQLAlchemy 1.2 is a new release of an Object Graph Mapper (OGM) aimed at simplifying the interaction with graph databases, particularly Memgraph, using Python instead of the Cypher query language. This update introduces a table data importer module for importing data from sources like Amazon S3, Azure Blob, and local storage, and an instance runner module for managing Memgraph instances directly through Python. The release also includes numerous bug fixes and improvements such as enhanced label inheritance, new query builder methods, and options for creating label indexes and batch save methods for nodes and relationships. GQLAlchemy supports importing various file formats via the PyArrow package, and users are encouraged to contribute their own importers. Future developments are underway, and users are invited to join the community on GitHub and Discord for suggestions and discussions.

Managing Memgraph Docker instances in Python can streamline the development of graph-based applications by using the instance_runner module, which allows users to start, stop, connect to, and monitor Memgraph instances directly from Python scripts. The process begins with importing the necessary components and creating a Memgraph instance using the MemgraphInstanceDocker class, which utilizes the memgraph/memgraph Docker image. The instance is configured with default settings such as the latest Docker image tag, a wildcard host address, and the default port 7687. Users can customize configurations with additional flags, and once the instance is running, it can be queried using GQLAlchemy. The instance's status can be checked with the is_running() method, and it can be stopped using the stop() method. The guide encourages further engagement through its community on Discord for additional questions or support.

Bruno Sacaric's blog post provides a comprehensive tutorial on migrating table data to a graph database using GQLAlchemy, specifically focusing on importing data into the Memgraph graph database. The guide covers the prerequisites needed, such as installing Jupyter, Docker, and GQLAlchemy, and details the steps for setting up Memgraph through Docker. It explains how to connect to Memgraph using GQLAlchemy and execute Cypher queries, which are similar to SQL commands for graph databases. The tutorial includes creating mock CSV table data and translating it into a graph form using a data configuration object. It demonstrates importing data from both local file systems and online storage providers like Amazon S3 and Azure Blob, and concludes with querying and visualizing the imported data using Memgraph Lab, encouraging users to explore further with graph algorithms and community engagement on Discord.