The Role of Intermediate Representations<\/h3>\n
Compiler intermediate representations (IRs) are traditionally thought of, and designed with, a specific source language (or languages) in mind. Their historical use has primarily been as an intermediate step in the process of lowering a source language program, with all of the associated syntactic sugar, into a final executable binary. However, over time they have become increasingly important in supporting program analysis and IDE tooling tasks. In these scenarios choices which were did not matter in the context of the compilation workflow can have major negative impacts.<\/p>\n
In the Bosque project we ask the question of what happens if the IR is designed explicitly to support the rich needs of automated code reasoning, IDE tooling, etc. With this novel IR first perspective we are exploring a new way to think about and build a language intermediate representation and tools that utilize it. Our initial experiments show that this empowers a range of next-generation experiences including symbolic-testing (opens in new tab)<\/span><\/a>, enhanced fuzzing, soft-realtime compilation with stable GC support (opens in new tab)<\/span><\/a>, API auto-marshaling, and more!<\/p>\n Many features that make the Bosque IR amenable for automated reasoning involve simplifying and removing sources of irregularity in the semantics. These regularizations<\/em> also simplify the task of understanding and writing code for the human developer. Inspired by this idea the Bosque project is building a new regularized programming language<\/em> (opens in new tab)<\/span><\/a> that takes advantage of the features of the IR.<\/p>\n Bosque Calculator Sample Code<\/p><\/div>\n The Bosque Programming Language builds on the strengths of classical Functional Programming,\u00a0 modern TypeScript\/Node.js, and the power of the new IR. The result is a language that simultaneously supports the kind of high productivity development experience available to modern developers while also providing a resource efficient and predictable runtime, scaling from small IoT up to heavily loaded cloud services. In addition to bringing all the expressive power expected from a modern language, the Bosque language also introduces several novel features like Typed Strings<\/em> (opens in new tab)<\/span><\/a> and API Types<\/em> (opens in new tab)<\/span><\/a>, that directly address challenges faced by developers working in a distributed cloud based world.<\/p>\n The move into cloud based development, with architectures based around microservices, serverless functions, and RESTful APIs, brings new challenges for development. In this environment an program may interoperate with many other (remote) services which are maintained by different teams (and maybe implemented in different languages). This forces APIs to use least-common denominator types for interop and creates the need for extensive serialize\/deserialize\/validate logic. Further, issues like cold-service startups, 95% response times, resiliency and diagnostics, all become critical but have not been design considerations in most traditional languages.<\/p>\n The Bosque project takes a cloud and IoT first view of programming languages. Thus, it includes features like API Types to simplify the construction and deployment of REST style APIs. Application initialization design provides 0-cost loading for lighting fast (cold) startup. Choices like fully determinized (opens in new tab)<\/span><\/a> language semantics, keys and ordering (opens in new tab)<\/span><\/a>, and memory behavior result in a runtime with minimal performance variability and enable ultra-low overhead tracing.<\/p>\n An overarching theme of the Bosque project is increasing the ability of automated tools to reason about and transform code. This mechanization is a foundational part of unlocking the future of using AI and Synthesis in the development pipeline. The ability to mechanically reason about the semantics of a program via symbolic means is a key enabler to synthesizing (opens in new tab)<\/span><\/a> novel and useful code components using examples or conditions. Bosque’s fully determinized and loop free design can also help facilitate the development and application\u00a0of automated program differentiation (opens in new tab)<\/span><\/a>. These are open problems but, just as we saw how Bosque unlocks value in classical tooling\/compilation scenarios, we are excited to see what it can do to power the AI and Synthesis programming revolution.<\/p>\n To simplify collaboration with other researchers and the wider developer community this project is setup around an open-source (MIT) licensed GitHub repository (opens in new tab)<\/span><\/a>. This project welcomes community contributions including, issues, comments, pull requests and research based on or enhancing the Bosque language.<\/p>\nRegularized Programming<\/h3>\n
![\"Bosque](\"https:\/\/www.microsoft.com\/en-us\/research\/uploads\/prod\/2019\/04\/bosque_code_sample.png\")
Cloud First Development<\/h3>\n
Powering Programming 2.0<\/h3>\n
Get Involved<\/h2>\n