Skip to main content

A light-weight, Extendable, high level, universal code parser built on top of tree-sitter

Project description

Code mining at scale - tree hugger

Downloads PRs Welcome Support Python Version PyPI version autosoft-dev

For People in a Hurry :)

Open In Colab


Mine source code repositories at scale. Easily. Tree-hugger is a light-weight, high level library which provides Pythonic APIs to mine trough Git repositories (it works on any collection of supported code files, actually).

Tree-hugger is built on top of tree-sitter.

Covered languages:

  • Python
  • PHP
  • Java
  • JavaScript
  • C++

System Requirement: Python 3.6


Table of contents


From pip:

pip install -U tree-hugger PyYAML

From Source:

git clone

cd tree-hugger

pip install -e .

The installation process is tested in macOS Mojave, we have a separate docker binding for compiling the libraries for Linux and soon this library will be integrated in that as well

You may need to install libgit2. In case you are in mac just use brew install libgit2


Getting your .so files

From onwards tree-hugger 0.9 we ship a new command download_libs.

If you are working on Debian based Linux or Newer version of MacOS then you should probably just use this command to get the library. At any point of time we will maintain a .so file for both those OSs with all the supported languages in it.

To get the .so file for your platform you can simply do the following


Here is the full usage guide of the command

usage: download_libs [-h] [--local_file_name LOCAL_FILE_NAME]

optional arguments:
  -h, --help            show this help message and exit
  --local_file_name LOCAL_FILE_NAME
                        The local file where you want to save the library.
                        Default -

Building the .so files

Please note that building the libraries has been tested under a macOS Mojave with Apple LLVM version 10.0.1 (clang-1001.0.46.4)

Please check out our Linux specific instructions here

Once this library is installed it gives you a command line utility to download and compile tree-sitter .so files with ease. As an example -

create_libs python

Here is the full usage guide of the command

usage: create_libs [-h] [-c] [-l LIB_NAME] langs [langs ...]

positional arguments:
  langs                 Give the name of languages for tree-sitter (php,
                        python, go ...)

optional arguments:
  -h, --help            show this help message and exit
  -c, --copy-to-workspace
                        Shall we copy the created libs to the present dir?
                        (default: False)
  -l LIB_NAME, --lib-name LIB_NAME
                        The name of the generated .so file

Environment variables

You can set up TS_LIB_PATH environment variable for the tree-sitter lib path and then the libary will use them automatically. Otherwise, as an alternative, you can pass it when creating any Parser object.

Hello world example

  1. Generate the librairies : run the above command to generate the libraries.

    In our settings we use the -c flag to copy the generated tree-sitter library's .so file to our workspace. Once copied, we place it under a directory called tslibs (It is in the .gitignore).

    ⚠ If you are using linux,you will need to use our tree-sitter-docker image and manually copy the final .so file. Unless you are in a debian based distro and in that case you should probably use our pre-compiled version via download_libs command as described above

  2. Setup environment variable (optional) Assuming that you have the necessary environment variable setup. The following line of code will create a Parser object according to the language you want to analyse:


# Python
from tree_hugger.core import PythonParser
pp = PythonParser()
['first_child', 'second_child', 'say_whee', 'wrapper', 'my_decorator', 'parent']


from tree_hugger.core import PHPParser
phpp = PHPParser()
Out[5] :
['foo', 'test', 'simple_params', 'variadic_param' ]


# Java 
from tree_hugger.core import JavaParser
jp = JavaParser()
Out[6] :
['HelloWorld','Animal', 'Dog' ]


# JavaScript
from tree_hugger.core import JavascriptParser
jsp = JavascriptParser()
Out[7] :
['test', 'utf8_to_b64',	'sum', 'multiply' ]


from tree_hugger.core import CPPParser
cp = CPPParser()
Out[8] :
['foo', 'test', 'simple_params', 'variadic_param' ]

API reference

Language Functions Methods Classes
Python get_all_function_names get_all_function_doctrings get_all_function_names_and_params get_all_function_bodies get_all_class_methods get_all_class_method_docstrings get_all_class_method_bodies get_all_class_names get_all_class_docstrings
PHP get_all_function_names get_all_function_names_and_params get_all_function_bodies get_all_class_methods get_all_class_names
Java get_all_class_methods get_all_method_names_and_params get_all_method_bodies get_all_class_names
JavaScript get_all_function_names get_all_function_names_and_params get_all_function_bodies get_all_class_methods get_all_class_names
C++ get_all_function_names get_all_function_names_and_params get_all_function_bodies get_all_class_methods get_all_class_names

Extending tree-hugger

Extending tree-hugger for other languages and/or more functionalities for the already provided ones, is easy.

  1. Adding languages:

Parsed languages can be extended through adding a parser class from the BaseParser class. The only mandatory argument that a Parser class should pass to the parent is the language. This is a string. Such as python (lower case). Each parser class must have the options to take in the path of the tree-sitter library (.so file that we are using to parse the code) and the path to the queries yaml file, in their constructor.

The BaseParser class can do few things:

  • Loading and preparing the .so file with respect to the language you just mentioned.
  • Loading, preparing and parsing the query yaml file. (for the queries, we internally use an extended UserDict class)
  • Providing an API to parse a file and prepare it for query. BaseParser.parse_file

It also gives you another (most likely not to be exposed outside) API _run_query_and_get_captures which lets you run any queries and return back the matched results (if any) from the parsed tree.

We use those APIs once we have called parse_file and parsed the file.

  1. Adding queries:

Queries processed on source code are s-expressions, they are listed in a queries.ymlfile for each parser class. Tree-hugger gives you a way to write your queries in yaml file for each language parsed.

Query structure: A name of a query followed by the query itself. Written as an s-expression. Example:

            name: (identifier) @function.def
            body: (block(expression_statement(string))) @function.docstring

You have to follow yaml grammar while writing these queries. You can see a bit more about writng these queries in the documentation of tree-sitter.

Some example queries, that you will find in the yaml file (and their corresponding API from the PythonParser class) -

* all_function_names => get_all_function_names()

* all_function_docstrings => get_all_function_documentations()

* all_class_methods => get_all_class_method_names()


  • Documentation: tutorial on queries writing

  • Write *Parser class for other languages

Languages Status-Finished Author
Python Shubhadeep
PHP Clément
Java Clément
JavaScript Clément
C++ Clément

If you are using tree-hugger in your project, please consider putting parssr: tree-hugger in your prject :)

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

tree-hugger-0.10.1.tar.gz (18.8 kB view hashes)

Uploaded source

Built Distribution

tree_hugger-0.10.1-py3-none-any.whl (24.7 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page