Callgraph is a Python package that defines a decorator, and Jupyter magic, to draw dynamic call graphs of Python function calls.

It’s intended for classroom use, but may also be useful for self-guided exploration.

The package defines a Jupyter IPython magic, %callgraph, that displays a call graph within a Jupyter cell:

from functools import lru_cache

@lru_cache()
def lev(a, b):
    if "" in (a, b):
        return len(a) + len(b)

    candidates = []
    if a[0] == b[0]:
        candidates.append(lev(a[1:], b[1:]))
    else:
        candidates.append(lev(a[1:], b[1:]) + 1)
    candidates.append(lev(a, b[1:]) + 1)
    candidates.append(lev(a[1:], b) + 1)
    return min(candidates)

%callgraph -w10 lev("big", "dog"); lev("dig", "dog")

It also provides a Python decorator, callgraph.decorator, that instruments a function to collect call graph information and render the result.

$ pip install callgraph

In a Jupyter IPython notebook:

%load_ext callgraph

def nchoosek(n, k):
    if k == 0:
        return 1
    if n == k:
        return 1
    return nchoosek(n - 1, k - 1) + nchoosek(n - 1, k)

%callgraph nchoosek(4, 2)

As an alternative to including %load_ext callgraph in each notebook that uses %callgraph, you can add the extension to the Notebook configuration file in your IPython profile.

Your configuration file is probably called ~/.ipython/profile_default/ipython_config.py. (You can run ipython profile locate to find it.) Edit this file to include the following line:

c.InteractiveShellApp.extensions = ["callgraph.extension"]

(If your configuration file already includes an uncommented statement c.InteractiveShellApp.extensions = […], edit the list of extensions in that line to include "callgraph.extension".

See extension example notebook for additional examples.

$ pip install callgraph

from functools import lru_cache
import callgraph.decorator as callgraph

@callgraph()
@lru_cache()
def nchoosek(n, k):
    if k == 0:
        return 1
    if n == k:
        return 1
    return nchoosek(n - 1, k - 1) + nchoosek(n - 1, k)

nchoosek(5, 2)

nchoosek.__callgraph__.view()

See the API documentation for additional documentation.

See the decorator example notebook for additional instructions and examples.

Install dev tools, and set up a Jupyter kernel for the current python enviromnent:

$ pip install -r requirements-dev.txt
$ python -m ipykernel install --user

Install locally:

flit install --symlink

Callgraph uses the Python graphviz package. Python graphviz uses the Graphviz package.

MIT