thvnx / langkit

Language creation framework.

Langkit

Langkit (nickname for language kit) is a tool whose purpose is to make it easy to create syntactic and semantic analysis engines. Write a language specification in our Python DSL and Langkit will generate for you an Ada library with bindings for the C and Python programming languages.

The generated library is meant to provide a basis to write tooling, including tools working on potentially changing and incorrect code, such as IDEs.

The currently main Langkit user is Libadalang, a high performance semantic engine for the Ada programming language.

Dependencies

To use Langkit, you will need:

• A Python 3.9 or Python 3.10 interpreter. Python2 is no longer supported.
• Some python libraries, including the Mako template system for Python (see REQUIREMENTS.dev for the full list).
• A recent version of the GNAT Ada compiler, either from your OS's packages, or from here.
• The gnatcoll-core library.
• Ada bindings for GMP and Libiconv, from gnatcoll-bindings.
• Clang-format, optionally, if you want C/C++ files to be formated.

Install

Langkit uses standard Python packaging, so to install it, just run the following command from the root directory:

$ pip install .

Build

Libraries generated by Langkit depend on an Ada project: Langkit_Support. To build and install that project, run:

# Build the "langkit_support.gpr" project
$ python manage.py build-langkit-support --library-types=static,static-pic,relocatable

# Install it. Replace $PREFIX below with the directory where you want to
# install the langkit_support.gpr project.
$ python manage.py install-langkit-support $PREFIX --library-types=static,static-pic,relocatable

If you are interested in shared (relocatable) libraries only, you can omit the --library-types argument.

Testing

Unlike the rest of Langkit, the testsuite framework requires Python 3.8 or later versions. Make sure the langkit package is available from the Python interpreter (see Install), then build the Libpythonlang/Liblktlang projects:

$ python manage.py make

You can then make them available to the environment:

$ eval `python manage.py setenv`

Finally, in order to run the testsuite, launch the following command from the top-level directory:

$ python manage.py test

This is just a wrapper passing convenient options to the real testsuite driver that is in testsuite/testsuite.py.

Note that even though the testsuite framework requires Python 3.8, it is possible to run the tests themselves using a different Python interpreter. For instance, to run them using Python 3.7, run:

$ python manage.py test --with-python=python3.7

If you want to learn more about this test driver's options (for instance to run tests under Valgrind), add a -h flag.

Documentation

The developer and user's documentation for Langkit is in langkit/doc. You can consult it as a text files or you can build it. For instance, to generate HTML documents, run from the top directory:

$ make -C doc html

And then open the following file in your favorite browser:

doc/_build/html/index.html

Bootstrapping a new language engine

Nothing is more simple than getting an initial project skeleton to work on a new language engine. Imagine you want to create an engine for the Foo language, run from the top-level directory:

$ python scripts/create-project.py Foo

And then have a look at the created foo directory: you have minimal lexers and parsers and a manage.py script you can use to build this new engine:

$ python foo/manage.py make

Here you are!

Developer tools

Langkit uses mako templates generating Ada, C and Python code. This can be hard to read. To ease development, Vim syntax files are available under the utils directory (see makoada.vim, makocpp.vim). Install them in your $HOME/.vim/syntax directory to get automatic highlighting of the template files.