July 2024 Summaries

Code coverage is a metric that shows how much of a software system's source code has been executed during testing. It helps improve the quality and reliability of software by identifying untested parts of an application, reducing the risk of undetected bugs. There are various types of code coverage, including statement, branch, function, path, and patch coverage. While high code coverage is beneficial, it's unusual to achieve 100% for an entire codebase. Aiming for strong coverage (80-90%) is generally considered good practice. To improve code coverage, write testable code, identify gaps in testing, automate tracking and reporting, follow SOLID principles, and pay attention to cyclomatic complexity.

In a recent blog post, our CEO Dan Rogers discussed strategies for reducing the impact of painful software releases by implementing progressive release methods, automating monitoring and rollbacks, targeting audience segments, and using runtime configuration management. By decoupling deployment from release, organizations can safely test new features in production with reduced risk. Progressive releases involve incrementally increasing user exposure to new features, allowing issues to be caught before affecting a significant portion of the user base. Proactive monitoring and automatic rollbacks help identify software issues early on and enable quick remediation. Targeted releases and customized experiences allow for different feature offerings based on various factors such as geographical regions, subscription tiers, device versions, personalization preferences, and more. Runtime configuration management enables dynamic updates to application configurations, allowing for control over configuration values during runtime. By aligning people, processes, and tools, organizations can mitigate the risks associated with software releases.

This tutorial guides you through building a Next.js Pokédex using the PokeAPI, with an interactive game hidden behind a feature flag controlled by LaunchDarkly. The project uses Next.js, TypeScript, Vercel, and LaunchDarkly's JavaScript Client SDK for feature-flagging. The frontend design is inspired by Emmanuel Tabarez's Pokédex CSS grid on Codepen. The game tests players' knowledge of Pokémon types in quiz mode or displays random Pokémon images in random mode.

The increasing pace of innovation and software delivery brings both excitement and risks. As the world becomes more reliant on digital solutions, it's crucial for the industry to minimize inherent risks associated with software delivery. Bugs are inevitable, but their impact can be minimized through proper internal processes and tools. LaunchDarkly suggests progressive rollouts, automated monitoring and instant rollbacks, runtime configuration management, and targeted segments as ways to prevent disruptions in business operations. Advanced release management and software stability are not just technical necessities but also business imperatives for C-suite executives and technology leaders.

A recent survey of 500 DevOps and IT professionals by LaunchDarkly revealed that customers using their feature management platform experienced significant benefits compared to non-customers. Key findings include: 1. Deployment frequency: LaunchDarkly customers deployed code 84% more frequently than non-customers, indicating a higher developer innovation velocity. 2. Software reliability and customer satisfaction: Customers had superior mean time to recover (MTTR) from incidents, fewer reliability issues overall, and less negative user feedback related to unreliable software. 3. Release overhead: LaunchDarkly customers required far less overhead than non-customers to execute software releases, particularly when it came to labor costs. They spent significantly less time on non-innovative work such as testing code in pre-production environments and fixing bugs. These results demonstrate the benefits of using LaunchDarkly's feature management and experimentation platform, including high developer innovation velocity, faster incident recovery times, greater software reliability, higher customer satisfaction, and low overhead when releasing and managing software.

This tutorial guides users through setting up, running, and analyzing an A/B experiment in LaunchDarkly using the Galaxy Marketplace example application. It covers cloning the app, installing dependencies, connecting to a LaunchDarkly account, creating feature flags, establishing hypotheses, tracking events, creating experiments, reviewing data, drawing conclusions, implementing changes, and monitoring post-launch results. The tutorial emphasizes the importance of statistical significance, performance comparison, contextual factors, and communication with team members during the experimentation process.

Continuous Integration (CI), Continuous Delivery (CD), and Continuous Deployment (CD) are practices that automate the process of getting code from development to production, each playing a unique role in the software development lifecycle. CI involves frequently integrating code changes into a shared repository with automated testing, while CD extends CI by automating the release process to a staging environment and retaining manual approval for final deployment. Continuous Deployment fully automates releases without any manual intervention. These practices improve productivity, reduce integration issues, and accelerate software delivery.

Continuous Integration and Continuous Delivery (CI/CD) are crucial practices for delivering high-quality software at scale. They streamline development processes, reduce risks, and accelerate time to market. CI/CD best practices include committing code frequently, optimizing pipeline stages, building code artifacts once, automating tests, keeping builds fast and simple, using shared pipelines (DRY), taking a security-first approach, creating test environments on demand, monitoring and measuring your pipeline, involving the whole team in CI/CD implementation, implementing progressive delivery strategies, choosing CI/CD tools with care, and creating a culture of continuous improvement. Adopting these practices helps organizations deliver software more efficiently and reliably.

In this tutorial, we learn how to build a sentiment analysis app using Hugging Face Spaces and LaunchDarkly AI model feature flags. We set up the environment with Python, Streamlit, Transformers, PyTorch, and LDClient libraries. We create an interactive web application within the larger Hugging Face Space for this tutorial. The application uses a sentiment analysis model to determine the sentiment expressed in a text as positive, negative, or neutral. LaunchDarkly's AI Model Flag feature allows us to customize which model to use and its parameters dynamically. This setup enables users to switch models on the fly and experiment with different configurations effortlessly.