Pierre-Marie De Rodat

Langkit is a framework that aims to make it easy to create a source code analyzer for a language, by specifying its syntax and high level semantics in a domain specific language, and compiling this specification to a library that clients can use to implement any tool that needs syntactic and information about source code for that language.

Modern programming environments are heavily multi-languages. Where programmers had to use one or two languages at most, most developpers today have to juggle with an order of magnitude more languages. But people are still using pretty basic text editors and tools, with very little source awareness. Libraries for automated source code analysis are still pretty scarce. Solutions for analysis of incorrect code are still often ad-hoc and proprietary, and every language implementor has to re-invent the wheel several times.

Langkit is a framework that aims to change that to some extent, making it easy to specify your language's syntax and high level semantics in a domain specific language, and compiling this specification to a library that clients can use to implement any tool that needs syntactic and information about source code for that language.

Generated engines support out of the box includes: Ada, C, and Python APIs for the engine, support for embedding into long lived applications, high performance parsing. Also, since Langkit knows a lot about your language, it can be used to auto-generate, or ease the generation of:

• Source code unparsers (pretty-printers)
• Source code highlighters
• Auto completion servers
• Language servers (implementing for example the Microsoft Language Server Protocol)
• Static analyzers

This talk will go over the Langkit architecture, what is doable with Langkit, and the design choices and compromises that have been made.

A lot of developers consider that a language is only as good as the tooling that accompanies it. Ada has been conceived as a language pretty well amenable to tooling, yet the tooling offer besides AdaCore's is not very extensive, at least when compared to other languages like Java, despite the existence of the ASIS project (Ada Semantic Interface Specification).

One of Libadalang's aims is to help solve that by providing an easy way to build new Ada-aware tools. Libadalang is a library that allows the user to query information about Ada code, including: Syntactic information. Query the token stream, the syntax tree, find syntax patterns, etc. Semantic information, such as which declaration an identifier references, the type of expressions, all references to a declaration, etc.

In addition, one of the aims is to allow the users to modify the trees, and propagate the changes to the source.

This talk will go over what Libadalang can already do today, how it differs from ASIS, future plans for the library, and potential exciting use cases.