Popularity
5.9
Declining
Activity
0.0
Stable
1,393
45
504

Description

PySonar2 is a type inferencer and indexer for Python, which performs sophisticated interprocedural analysis to infer types. It is one of the underlying technologies that power the code search engine Sourcegraph, where it has been used to index hundreds of thousands of open source Python repositories, producing a globally connected network of Python code. An older version of PySonar is used internally at Google, producing high-quality semantic code index for millions of lines of Python code.

To understand its properties, please refer to my blog post:

Code Quality Rank: L1
Programming language: Java
License: Apache License 2.0

pysonar2 alternatives and similar packages

Based on the "Code Analysis" category.
Alternatively, view pysonar2 alternatives based on common mentions on social networks and blogs.

Do you think we are missing an alternative of pysonar2 or a related project?

Add another 'Code Analysis' Package

README

PySonar2 - a semantic indexer for Python with interprocedural type inference

PySonar2 is a semantic indexer library for Python, designed for batch processing of large code bases. The resulting index can be used to build code browsers and code search engines.

Python is a highly dynamic language. To achieve high accuracy and quality, PySonar2 performs interprocedural analysis to infer types. PySonar2 generally produces better index than Python IDEs, while at the same time sacrificing real-time update capabilities of IDEs.

Notice the term "type inference" here is descriptive rather than prescriptive, meaning the inferred types describe how the code is actually used, but does not prescribe how they must be used. So the way type inference works here is different from type systems (e.g. Hindley-Milner system).

PySonar2 has been the underlying indexing engine for several large-scale code navigation services, such as Google's internal Code Search, sourcegraph.com and insight.io (now part of elastic).

How to build

mvn package -DskipTests

Demo

To have a feel of what PySonar2 produce, you can build a simple code browser of the Python 2.7 standard library with the following command line:

java -jar target/pysonar-<version>.jar /usr/lib/python2.7 ./html

This may take a few minutes depending on your machine. You should find some interactive HTML files inside the html directory after this process. You can move your mouse on the variables and click on them to jump to definitions etc.

Note that this is just a simple demo program based on the library. PySonar2 is not meant to be an end-user tool. It is mainly designed as a library for Python IDEs, developer tools and code search engines, so its interface may not be as appealing as an end-user tool.

If you have problems with it, please feel free to contact me.

System requirements

  • Python 2.7.x
  • Python 3.x
  • Java 8+
  • maven

Environment variables

PySonar2 uses CPython's built-in ast package to parse Python code, so please make sure you have python or python3 installed and pointed to by the PATH environment variable. If you have them in different names, please make symbol links.

PYTHONPATH environment variable is used for locating the Python standard libraries. It is important to point it to the correct Python library, for example

export PYTHONPATH=/usr/lib/python2.7

If this is not set up correctly, references to library code will not be found.

Contribute

You are welcome to make code contributions.

Because of the highly complex and unpublished theory behind PySonar2, things may go wrong easily with even an innocent-looking change. If you hope to contribute to PySonar2, please discuss with me first before making significant changes, otherwise I may not be able to review your changes.

For basic verification, you can run the unit tests. PySonar2 has a basic test framework. You can run the tests using this command:

mvn test

If you modify the code or tests, you need to generate new expected results. Run these command lines:

mvn package -DskipTests
java -classpath target/pysonar-<version>.jar org.yinwang.pysonar.TestInference -generate tests

To write new tests, you just need to write relevant Python code demonstrating your change, put them into a directory named tests/testname.test(test directory name must end with ".test"). Please look at the tests directory for examples.

Please don't expect the tests to catch all bugs. Be very careful :)

License

Apache 2.0 License. See LICENSE file.


*Note that all licence references and agreements mentioned in the pysonar2 README section above are relevant to that project's source code only.