Parse SDK and Mobile Backend Development

10/1/2025

This post details Meta's adoption and successful implementation of Android's Baseline Profiles feature to significantly enhance the performance of its Android applications. It explains the underlying ART performance considerations, the specific challenges Meta faces with large and dynamic codebases, and how Baseline Profiles, by enabling install-time optimizations like AOT compilation and app images, have addressed these issues. The post highlights that Baseline Profiles have improved critical performance metrics by up to 40% across Meta's apps, offering a more controlled and immediate optimization compared to Cloud Profiles.

3/31/2025

This post highlights the ongoing evolution and future of Mobile GraphQL at Meta, emphasizing its role in building better user experiences for Meta's family of apps and simplifying developer lives through innovative APIs. It features insights from the Mobile GraphQL Platform Team on challenges and best practices for building such platforms.

2/25/2025

This post details Meta's significant undertaking to translate its large Android codebase from Java to Kotlin. It highlights the development of open-sourced tools for manipulating Kotlin code and discusses the challenges and strategies involved in this large-scale migration, including avoiding nullability issues and generating idiomatic code for internal frameworks. This directly contributes to the ongoing evolution of Meta's mobile development tools and practices.

10/2/2024

This post details the use of React Native for building mixed reality applications for Meta Quest, including Instagram and Facebook for Meta Quest, and the Meta Horizon mobile app. It highlights code sharing capabilities with StyleX and React Strict DOM, performance improvements in the Meta Horizon app using Relay for earlier network queries, and the development of the Meta Spatial Editor using React Native for Desktop with custom native component integrations. It also mentions the adoption of React Compiler and the New React Native Architecture.

6/11/2024

This post details the expansion of Meta's MobileConfig platform, initially used for core mobile apps, to support mixed reality (MR) devices like Meta Quest and Ray-Ban Meta smart glasses. It highlights how MobileConfig enables centralized configuration management, feature flagging, and A/B testing for these devices, addressing challenges related to extended development cycles and shared configuration needs within the MR ecosystem. The post also describes adaptations of MobileConfig for low-power devices and microcontrollers, including optimizations for limited connectivity and scarce resources, demonstrating a significant evolution in Meta's approach to device configuration and developer velocity across a broad spectrum of hardware.

2/6/2023

This post details the architectural evolution of Facebook for iOS (FBiOS) over a decade, highlighting key shifts from 2014 to 2018. It describes the move away from Apple's Core Data and MVC patterns towards a custom declarative UI framework (ComponentKit) inspired by React, leading to significant performance improvements. It also explains the strategic adoption of dynamically loaded libraries (dylibs) to combat unbounded growth in app size and startup time, and the subsequent reliance on the Buck build system for managing code generation and dependencies to address the complexities introduced by dylibs.

3/2/2020

Project LightSpeed involved a complete rewrite of the Messenger iOS codebase, reducing code lines by 84% (from 1.7M to 360K) and resulting in a 4x smaller app size and 2x faster startup time. Key architectural changes include leveraging native OS capabilities for rendering, threading, and logging; reusing UI components with dynamic templates driven by SQLite; and using SQLite as a universal system for data storage, concurrency, caching, and transactions, eliminating the need for custom in-memory caching and transaction subsystems. A server broker was introduced as a universal gateway between the client and server features.

7/12/2019

This post introduces Hermes, an open-source JavaScript engine specifically designed to improve the performance of mobile applications, particularly those built with React Native. Key technical contributions include bytecode precompilation for faster startup and smaller app size, the absence of a Just-In-Time (JIT) compiler to avoid impacting TTI and memory usage, and a specialized garbage collector strategy optimized for mobile device constraints. It also details developer experience improvements like lazy compilation for faster iteration during development and Chrome remote debugging support. The post also highlights the role of JSI in enabling further React Native infrastructure improvements like Fabric and TurboModules.

5/1/2019

This post from F8 2019 highlights several key advancements in Meta's mobile development and backend services. It showcases continued innovation with React Native, ComponentKit, and Litho for cross-platform mobile app development. It also introduces a mobile AI platform for on-device machine learning, a focus on rebuilding Messenger for improved performance, and the use of Docusaurus for open-source documentation. Additionally, it mentions FBT for internationalization at scale and the use of GraphQL and Relay in building the new Facebook.com, indicating a strong emphasis on modern web and mobile development practices.

2/22/2019

This post announces the release of React Hooks (React 16.8 and React Native 0.59), a significant new feature that allows developers to use React state and lifecycle features from function components. Hooks provide a more organized and reusable way to manage state and side effects, improving code structure and avoiding common pitfalls of previous patterns like render props and higher-order components.

2/20/2019

This post introduces SPARTA, a C++ library designed to simplify the creation of high-performance static analyzers using abstract interpretation. It provides building blocks for engineers to develop scalable and sound static analysis tools, demonstrated by its successful integration into ReDex for Android bytecode optimization, resulting in significant speedups and memory reductions. SPARTA's language-independent nature and generic algorithms also allow for broader applications beyond Android.

2/13/2019

This post announces that Mark Harman, an engineering manager at Facebook London, has won the 2019 IEEE Harlan D. Mills Award for his contributions to software engineering, including search-based software engineering, slicing, testing, and genetic improvement. It highlights the Sapienz and SapFix tools developed by his team for automated testing and bug fixing of Facebook's Android and iOS apps. Sapienz uses automated test design to find faults, and SapFix uses machine learning to propose and deploy fixes, reducing manual testing and freeing up engineers for more creative tasks.

2/8/2019

This post highlights the significant achievement of the Infer team winning the POPL 2019 Most Influential Paper Award for their foundational work on Infer, Facebook's static program analyzer. It details Infer's capabilities in detecting subtle bugs at scale, its open-sourcing in 2015, and subsequent advancements like the AL declarative language, interprocedural bug detection, concurrency bug detection, and application to multithreaded rendering. The post emphasizes Infer's impact on improving product quality and saving engineering hours, as well as its adoption by other major tech companies.

10/19/2018

This post introduces MobileLab, a new system designed to prevent mobile performance regressions by enabling highly accurate testing. It addresses the challenge of detecting very small performance changes (as low as 1 percent) by creating a deterministic testing environment. This involves establishing a validation framework, eliminating variance through consistent device performance (locking CPU/GPU frequencies), consistent application performance (using an HTTP proxy), controlling application state across trials (backing up and restoring on-disk state), and further optimizations like using RAM disks for disk writes, resetting device clocks, dismissing crash dialogs, and using direct socket connections for metric communication instead of logcat tailing. MobileLab significantly reduces variance and allows for reliable detection of sub-1 percent regressions.

6/11/2018

This post introduces Flipper, an extensible cross-platform debugging tool for iOS and Android applications, open-sourced by Meta. Flipper builds upon the learnings from Stetho, offering a more extensible and richer user experience. It comprises a desktop client (built with Electron, React.js, Flow, Metro, RSocket, and Yarn) and a mobile SDK (using Folly and RSocket). The architecture allows for custom plugin development for both the desktop client (React components) and mobile SDK (native code), enabling bidirectional communication for debugging tasks like inspecting view hierarchies (Litho, ComponentKit), GraphQL requests, and performance markers. The post highlights Flipper's extensibility and its potential to improve mobile app development workflows.

3/13/2018

This post introduces Profilo, a high-throughput, mobile-first performance tracing library for Android (with plans for iOS). It addresses the challenges of diagnosing performance regressions in real-world conditions by providing detailed telemetry, a powerful configuration system for targeted data collection, and a custom Java stack unwinder for more accurate and efficient stack trace analysis. Profilo significantly improves the turnaround time for root-causing performance issues, enabling deeper understanding of app responsiveness, scroll fluidity, and startup performance.

1/23/2018

This post details a scalable solution developed by Meta to overcome the native library limit in Android versions prior to 4.3. It introduces a build system integration with Buck to merge native libraries by combining objects and specifically addresses the challenge of duplicate JNI_OnLoad symbols using custom ELF sections and a registration object. The solution also integrates with SoLoader for seamless library loading and includes strategies for handling edge cases and determining optimal merging policies. A demonstration repository is provided to showcase the techniques.

12/14/2017

This post details significant advancements in 2017 that optimized the software development lifecycle for engineers, particularly for Android. Key contributions include the development of Jupiter for high-performance job matching, reimaging Java compilation with Buck to reduce build times, the creation of the One World system for large-scale testing resource management, and the expansion of the Infer static analyzer to detect concurrency bugs, which enabled multi-threaded rendering on Android. The post also highlights the release of several modern open-source frameworks like React 16, Relay, Litho, and Yarn 1.0, and new machine learning projects, underscoring Meta's commitment to open-source contributions and improving developer efficiency and product performance.

11/9/2017

This post introduces significant advancements in Buck, Meta's build tool for Android development. It details how Buck was re-architected to improve Java compilation performance by: 1. Ignoring unused classes, reducing rebuilds by 35%. 2. Implementing rule pipelining, allowing dependent rules to start compiling earlier by generating stub JARs mid-compilation, reducing build times by 10%. 3. Introducing source-only stub generation, which flattens the dependency graph by generating stub JARs directly from source code without waiting for all dependencies, leading to further parallelism and faster builds.

10/26/2017

Introduced Sections, a declarative and composable API built on top of Litho, for building highly-optimized list surfaces. Sections abstract away manual adapter management and provide automatic data diffing and granular UI refreshes, leading to significant performance improvements (e.g., 42% scroll performance improvement for the Comments surface). The API supports composition of different data sources and offers support for rendering with traditional Views.

10/19/2017

This post introduces RacerD, a new open-source static race detector for Java programs, built on top of the Infer static analysis platform. RacerD addresses the challenges of concurrent programming by quickly identifying data races without running the program, significantly aiding in debugging and preventing multi-threading issues in large codebases like the Android app's News Feed. It represents a practical approach to concurrency analysis at scale, focusing on high signal detection and fast reporting to integrate into developer workflows.

5/24/2017

This post introduces AL, a new declarative language designed to simplify the creation of custom checkers for Infer, Meta's static code analyzer. AL allows engineers to define bug detection rules using a high-level, syntax-aware language without needing deep knowledge of Infer's internal workings. This significantly lowers the barrier to entry for adding new bug detection capabilities, leading to the identification of thousands of bugs and improving the overall quality of mobile code.

4/19/2017

This post announces the open-sourcing of several key developer tools: Litho, a declarative UI framework for Android that improves scroll performance; React Fiber, a rewrite of React core for better scheduling and future innovation; Relay Modern, an updated GraphQL client framework for data-driven applications; and React VR, a library for building virtual reality experiences on the web. These contributions aim to improve app experiences for both users and engineers, enabling richer and more immersive experiences, even on low-power devices and slow networks.

4/18/2017

Introduced Litho, a declarative UI framework for Android, inspired by React. Litho aims to improve UI performance by decoupling layout operations from Android views, moving them to a background thread using the Yoga layout system. It also enables incremental rendering by breaking down complex views into smaller pieces and rendering them across multiple frames, leading to up to 35% improvement in scroll performance and better memory behavior.

4/18/2017

This post introduces React VR, a new library that extends the React and React Native ecosystem to enable developers to build virtual reality (VR) experiences for the web. It leverages standard web tools and JavaScript, making VR content creation more accessible. React VR allows for the composition of 3D scenes with 2D UI, text, and images, and supports distribution across the web, mobile phones, and PCs. It adopts React Native's architecture for performance and is open-sourced.

12/7/2016

Introduced Yoga, a standalone, cross-platform layout engine implemented in C, derived from css-layout. Yoga aims to simplify cross-platform UI development by providing a unified layout system based on the Flexbox spec. It offers performance improvements (33% on Android) and has bindings for Java, Objective-C, and C#. It is being integrated into projects like React Native, Components for Android, Oculus, and is being migrated into Instagram views and ComponentKit.

7/15/2016

This post highlights Meta's ongoing commitment to open-sourcing development tools and frameworks, particularly those that enhance mobile development. It details the release and traction of React Native, Draft.js (a React-based rich text editor), ReDex (an Android bytecode optimizer), and a suite of iOS memory profiling tools (FBRetainCycleDetector, FBAllocationTracker, FBMemoryProfiler). It also mentions open-sourcing AI-related Torch implementations and a CTF platform for security training, underscoring a broad strategy to improve developer productivity, app performance, and security through community collaboration.

7/13/2016

This post details the creation and evolution of a large-scale mobile device lab at Meta's Prineville data center. It describes the journey from early, less scalable testing setups ('sled', 'gondola', 'slatwall') to a robust, data center-based solution. Key contributions include the design and implementation of custom, electromagnetically isolated racks for testing thousands of devices, the integration of automated testing tools like CT-Scan and Chef for configuration management and monitoring, and the ongoing efforts to improve density, expand support for iOS, and generalize the platform for broader team use. This directly addresses the need to understand and mitigate performance regressions on diverse mobile devices, a critical aspect of mobile development at scale.

6/27/2016

Introduced LongArraySet and IntArraySet for Android, which leverage LongSparseArray and SparseArrayCompat respectively. These implementations provide efficient set operations for primitive long and int data types, offering performance benefits over traditional HashSets for smaller collections by using binary search and an optimization for removed entries.

6/27/2016

Introduced `LongArraySet` and `IntArraySet` data structures to mitigate memory overhead caused by Java autoboxing when storing primitive `long` and `int` values in collections like `HashSet`. These custom structures leverage `LongSparseArray` and `IntSparseArray` respectively, offering a more memory-efficient alternative for scenarios with hundreds of items, thereby reducing garbage collection pauses and improving scrolling performance on Android.

4/13/2016

This post introduces a suite of tools (FBRetainCycleDetector, FBAllocationTracker, and FBMemoryProfiler) developed to automate the detection and fixing of memory leaks on iOS. It details the technical challenges of retain cycles in Objective-C, explains the runtime introspection techniques used for detection, and describes how these tools integrate for both automated and manual profiling, enhancing the robustness of iOS applications.

4/13/2016

Introduced Remodel, an Objective-C code generation tool for creating and editing model objects. Remodel addresses the drawbacks of using dictionaries or handwritten model objects by enforcing type safety, reducing boilerplate code for common methods like description andencodeWithCoder, and enabling the creation of immutable value models. This simplifies data flow, improves maintainability, and supports advanced concurrency patterns.

4/13/2016

This post details the one-year anniversary of React Native's open-sourcing. It highlights the framework's evolution from a hackathon project to a widely adopted cross-platform mobile development solution. Key technical aspects include the challenges faced in early development (touch negotiation, performance, debugging), the decision to make it cross-platform, the development of the Android runtime and core components, and the significant community adoption with over 650 contributors. The post also describes the development of internal tools like GitHub bots (mention bot, shipit bot, issues bot) to manage the high volume of pull requests and issues, and the integration of React Native into other platforms like Windows and Tizen.

4/12/2016

This post announces the open-sourcing of ReDex, a tool developed by Meta to reduce the size and improve the performance of Android applications. It details new optimizations implemented in ReDex, including feedback-directed class layout to optimize bytecode loading during cold starts, removal of single-implementation interfaces to save space and improve method call efficiency, and stripping of unnecessary metadata like source file names and annotations. The post also presents performance metrics showing significant reductions in app size (25%) and cold start times (up to 30%) for Facebook for Android. It highlights ReDex's design principles for developers: speed, a simple interface for APK handling, and configurability to exclude specific code from optimization. The availability of ReDex on GitHub is announced, encouraging community contributions.

3/29/2016

This post summarizes the Mobile @Scale London event, highlighting advancements in mobile development at scale. Key technical contributions discussed include: Google's use of Material Design, protocol buffers, C++, and Java on iOS; Facebook's development of Nuclide (based on Atom and Chromium) and contributions to IntelliJ for scalable mobile IDEs; SoundCloud's lessons learned in scaling mobile development; Spotify's approach to backend-driven native UIs; Facebook's Infer static analysis tool for finding deep bugs on mobile; Twitter's image read/write path evolution and adoption of Facebook's Fresco library for Android; and the benefits of React Native for cross-platform native development.

3/28/2016

This post details performance optimizations for React Native app startup, reducing it by half. Key improvements include: reducing JavaScript bundle loading time by avoiding string copying and decoding, stripping DEV-only modules using Babel transforms, implementing lazy requires for modules, optimizing Relay's incremental cache read to reduce I/O and bridge traffic, de-batching non-UI bridge calls while batching Relay cache reads, enabling early UI flushing for parallel processing, lazy loading of native modules, lazy touch bindings on text components, and deferring popular events queries. These optimizations were applied to the Events Dashboard feature in the Facebook for iOS app.

2/26/2016

This post introduces Draft.js, a React-based rich text editor framework developed by Facebook. It addresses the challenges of building rich text input fields within Facebook products, which previously relied on complex DOM hacks and <textarea> elements. Draft.js leverages React's declarative approach and immutable data structures to provide a customizable and robust solution for rich text composition, enhancing user experience and developer efficiency. Its open-sourcing signifies a contribution to the broader React ecosystem.

10/26/2015

This post details the development of a performance instrumentation system for Android apps that uses bytecode rewriting (via ASM) to inject logging code before Dalvik VM conversion. It addresses limitations of manual instrumentation and built-in profilers by automatically capturing detailed performance telemetry, including asynchronous execution across threads, to diagnose performance regressions and subtle bugs. Key technical challenges overcome include minimizing runtime overhead, handling asynchronous tracing, and managing a global stream of instrumentation events using a lock-free ring buffer.

10/1/2015

Introduced Redex, a pipeline for optimizing Android .dex files. Redex applies customizable transformations at the bytecode level to reduce .dex file size and improve runtime performance. Key optimizations include minification and compression of strings, inlining of functions to reduce overhead, and dead code elimination. The pipeline is designed to be extensible, allowing for the addition of new optimization plugins.

9/14/2015

This post introduces Relay, a JavaScript framework designed to simplify data retrieval for React applications. It addresses challenges with coordinating server data fetching by enabling declarative data requirements, which allows for efficient batching of network requests and colocation of data requirements with rendering logic. This reduces bugs related to under-fetching and improves client performance and server load by reducing over-fetching. Relay is already in use on facebook.com, messenger.com, and mobile apps using React Native.

9/14/2015

This post details the development and release of React Native for Android, building upon the existing React Native for iOS. It highlights the technical decisions and challenges in creating the Ads Manager app as Facebook's first cross-platform React Native application. Key contributions include strategies for maximizing code reuse (85%) between iOS and Android versions through platform-specific file extensions and component refactoring, and the development of a system to manage code across separate iOS and Android repositories, including a cronjob for syncing JavaScript code and enabling JavaScript developers to test on both platforms from the iOS repository. It also discusses the implementation of the Navigator component for complex navigation flows and animations, and the use of InteractionManager to prevent JavaScript thread blocking during data processing.

8/31/2015

This post details the use of React Native for building an iPad point-of-sale application for Facebook's 'Out the Window' restaurant, highlighting React Native's Flexbox layout and stylesheet coupling for rapid development. It also describes the integration of Parse for backend services, including order queue management via a Raspberry Pi connected to the database and serial printer, and the out-of-the-box analytics provided by Parse.

8/13/2015

This post details the internal architecture of the Parse SDK, focusing on its asynchronous API implementation using the Bolts framework's `Tasks` for managing parallelism and concurrency, and its decoupled architecture comprising public API object instances, object states, controllers, and REST protocol. It highlights how this design enables API consistency, testability, and extensibility, exemplified by the integration of the Local Datastore feature.

3/27/2015

This post announces the open-sourcing of React Native for iOS, enabling native UI development with modern web techniques. It also introduces ComponentKit for declarative UI on iOS, Year Class and Connection Class for performance and network segmentation on Android, and Fresco for efficient image handling on Android. Additionally, it mentions the upcoming release of Nuclide, a unified IDE for React Native, Flow, and Hack.

3/26/2015

Introduced React Native, a framework that allows developers to build native mobile applications for iOS and Android using JavaScript and the React paradigm. This enables a 'learn once, write anywhere' approach, leveraging web development techniques for mobile, while still producing native user experiences. It addresses the challenges of native mobile development by providing a faster development cycle, access to native UI components, and a more predictable programming model.

10/2/2014

This post details advancements in mobile development architectures and tools discussed at @Scale 2014. Key contributions include Facebook's custom iOS persistence layer replacing Core Data, a new declarative/functional view architecture named Components for News Feed UI, Dropbox's cross-platform C++ development with Djinni for bridging languages, LinkedIn's transition to native languages and Rest.li for RESTful services, Facebook's Android build tool Buck and Exopackage for faster builds, Uber's mobile scaling strategies, Microsoft Office's cross-platform C++ architecture, and Facebook's redesign of the Android app for emerging markets focusing on fragmentation, device performance, and data plans, including image download optimization and prefetching.

3/5/2014

This post introduces and open-sources Bolts, a library of mobile developer tools from Parse + Facebook, starting with Tasks. Tasks are presented as a solution to manage complex asynchronous code, offering advantages over callbacks such as chaining, automatic error propagation, guaranteed single execution, and reduced retain cycles in Objective-C, thereby improving memory management and code organization.

3/4/2013

This post details a critical patch applied to the Dalvik Virtual Machine's LinearAlloc buffer for Facebook for Android. To overcome installation failures and runtime crashes on older Android versions (Froyo/Gingerbread) caused by an insufficient buffer size for a large number of Java methods, Facebook engineers developed a JNI extension. This extension dynamically replaced the LinearAlloc buffer with a larger one and included robust heuristics to locate the buffer across different device manufacturers and Android versions, ensuring the app's functionality on a wide range of devices.

10/11/2012

This post details the implementation of the first mobile advertising interface for Facebook within the Pages Manager for iOS application. It highlights the use of BoltJS for client-side UI development, the design of scalable API endpoints, and considerations for internationalization and payment flows. It also describes the collaborative effort with another intern to create a streamlined campaign management interface with an insights view.

8/23/2012

This post details the complete rewrite of the Facebook for iOS application, moving from an HTML5-based approach and the Three20 framework to a native iOS implementation. Key technical contributions include: performing computationally expensive tasks (networking, JSON parsing, NSManagedObject creation, disk saving) on background threads to keep the main UI thread responsive; using Core Text for asynchronous string layout calculations and caching CTFramesetters; optimizing news feed startup by storing story heights in Core Data and splitting story model objects to fetch data asynchronously; and adopting a modular architecture for integrating components like Facebook Messenger, utilizing dependency injection for shared objects. A fallback renderer was introduced to allow for phased rollout of new features without requiring app updates.

9/30/2011

This post details the transition from Watir to WebDriver for browser-based integration testing at Facebook. It describes the technical challenges with the existing Watir framework, including its Ruby dependency and lack of integration with the PHP unit test system. The author then explains the development of a new WebDriver client in PHP, leveraging the language-agnostic nature of the WebDriver protocol to integrate browser tests into the existing PHP unit test framework, improving test execution speed and developer workflow.