:warning: Notes

Forked from https://github.com/vitsalis/pycg

Changes: See https://github.com/deeplime-io/PyCG/tree/onecode

Essentially added code tracks the order of the calls and the code associated to it. It is used by OneCode to properly interpret code based on the excellent PyCG.

Why a new PyPi package? Well, the not-so-great PyPi doesn't allow to have forked public repositories as part of the dependencies. Nice right?

PyCG - Practical Python Call Graphs

PyCG generates call graphs for Python code using static analysis. It efficiently supports

• Higher order functions
• Twisted class inheritance schemes
• Automatic discovery of imported modules for further analysis
• Nested definitions

You can read the full methodology as well as a complete evaluation on the ICSE 2021 paper.

You can cite PyCG as follows. Vitalis Salis, Thodoris Sotiropoulos, Panos Louridas, Diomidis Spinellis and Dimitris Mitropoulos. PyCG: Practical Call Graph Generation in Python. In 43rd International Conference on Software Engineering, ICSE '21, 25–28 May 2021.

PyCG is archived. Due to limited availability, no further development improvements are planned. Happy to help anyone that wants to create a fork to continue development.

Installation

PyCG is implemented in Python3 and requires Python 3.10–3.14. It also has no dependencies. Simply:

pip install onecode-pycg

Usage

~ >>> pycg -h
usage: __main__.py [-h] [--package PACKAGE] [--product PRODUCT]
                        [--forge FORGE] [--version VERSION] [--timestamp TIMESTAMP]
                        [--max-iter MAX_ITER] [--operation {call-graph,key-error}]
                        [--as-graph-output AS_GRAPH_OUTPUT] [-o OUTPUT]
                        [entry_point ...]

positional arguments:
  entry_point           Entry points to be processed

optional arguments:
  -h, --help            show this help message and exit
  --package PACKAGE     Package containing the code to be analyzed
  --max-iter MAX_ITER   Maximum number of iterations through source code. If not specified a fix-point iteration will be performed.
  --operation {call-graph,key-error}
                        Operation to perform. Choose call-graph for call graph generation (default) or key-error for key error detection on dictionaries.
  --as-graph-output AS_GRAPH_OUTPUT
                        Output for the assignment graph
  -o OUTPUT, --output OUTPUT
                        Output path

Call Graph Output

Simple JSON format

The call edges are in the form of an adjacency list where an edge (src, dst) is represented as an entry of dst in the list assigned to key src:

{
    "node1": ["node2", "node3"],
    "node2": ["node3"],
    "node3": []
}

FASTEN Format

Dropped - not useful for OneCode and requires porting of pkg_resources

For an up-to-date description of the FASTEN format refer to the FASTEN wiki.

Key Errors Output

We are currently experimenting on identifying potential invalid dictionary accesses on Python dictionaries (key errors). The output format for key errors is a list of dictionaries containing:

• The file name in which the key error was identified
• The line number inside the file
• The namespace of the accessed dictionary
• The key used to access the dictionary

[{
    "filename": "mod.py",
    "lineno": 2,
    "namespace": "mod.<dict1>",
    "key": "key2"
},
{
    "filename": "mod.py",
    "lineno": 8,
    "namespace": "mod.<dict1>",
    "key": "nokey"
}]

Examples

All the entry points are known and we want the simple JSON format

~ >>> pycg --package pkg_root pkg_root/module1.py pkg_root/subpackage/module2.py -o cg.json

All entry points are not known and we want the simple JSON format

~ >>> pycg --package django $(find django -type f -name "*.py") -o django.json

Running Tests

From the root directory, first install the mock package:

pip3 install mock

Τhen, simply run the tests by executing:

make test