Call Map

About

Call Map is a tool for navigating call graphs in Python, with plans to support other languages. Below is a screen shot after running call_map -m toolz, then clicking on some functions.

While Call Map is intended to help in gaining a general understanding of a codebase, it is also a natural fit for tracing code paths, which constitutes an important security concern. Many security issues are revealed by finding a code path that connects user input to dangerous coding patterns.

See the blog post for more.

Installation

Call Map is distributed as a python package for Python 3.5. To install the gpl-licensed version with pip3, run:

# From a directory where you want to install call_map
git clone -b gpl-licensed https://github.com/ajylee/call_map
pip3 install -e call_map

Note that some Python 3 distributions call the package manager pip, and some call it pip3, so you may need substitute pip3 with pip.

The above commands should install the executable call_map to the same path that the Python interpreter is in.

If you want to install the bsd-licensed version, you will need to install PySide2 yourself. In contrast, the gpl-licensed version depends on PyQt5, which pip should be able download and install automatically.

For details and alternative methods see INSTALL.rst.

Quick Start

Open files to start exploring:

call_map -f example.py

# you can also open multiple files
call_map -f *.py

You can also add to the module search path (sys.path):

call_map -f *.py -p .

Call Map will try to resolve the files as modules whenever they can be found in the module search path. For more documentation on command line arguments, call_map -h.

Configuration

To configure Call Map, set the environment variable CALL_MAP_RC_DIRECTORY. The path to the Call Map configuration file will be:

$CALL_MAP_RC_DIRECTORY/call_map_rc.py

At this time the configuration options are:

• `open_in_editor(path: pathlib.Path, line: int): if you define this function it will be called whenever you open a file. For example, the following can be used to open files in an Emacs server if have you called `server-mode in a running Emacs session:

  def open_in_editor(path, line):
      import subprocess as sbp
      sbp.call(['emacsclient', '+{}'.format(line), str(path)])

  The following can be used to open files in the GVim server if you started a GVim server with `gvim --servername my_vim_server`:

  def open_in_editor(path, line):
      import subprocess as sbp
      sbp.call(['gvim', '--servername', 'my_vim_server', '--remote', '+{}'.format(line), str(path)])

• `MULTITHREADING`: Whether to use a separate thread for the GUI and searching the call graph. Defaults to True. Turning it off is for debug purposes.

Quirks

There are a couple of quirks in the UI design, due to the fact that I haven't arrived at a better solution or the tradeoff of additional complexity is unfavorable.

• Usages can appear to show up more than once, but actually they are different usages in the same scope.
• You may notice some bulitins such as help, id, and filter are ignored. See call_map/config.py in the source code for the full list of ignored functions.
• If the position to be highlighted is at the start of a file, it won't be highlighted. This is because typically only modules and scripts are positioned at the start of their respective files.

Quirks Inherited from Jedi

Some quirks are inherited from the jedi Python analysis backend.

• The search scope for usages is the set of modules that have been loaded by the jedi backend. That means that the scope will change as you explore new modules. If you want to explicitly include a module in the search scope, add it to the initial list of modules to be inspected.
• jedi always searches the interpreter's sys.path, even when it is not explicitly included. However, it will prioritize the user-defined sys_path.

Caveats

Any caveats that exist for static analysis backends also apply to Call Map. (At this time only the jedi backend is integrated into Call Map.)

Jedi

The jedi backend for Python analysis does not always find all usages. The dynamic nature of Python makes it impossible to always determine the definition of a function with static analysis. Sometimes jedi resolves a call to multiple possible functions. For example os.path.abspath depends on the platform. In this case call_map lists both possibilities as abspath and abspath (2). jedi also does not resolve all calls and usages. Jedi's own documentation also has a list of caveats.

Sometimes when Jedi throws an error when analyzing one item, the error affects other items. For example, when finding usages, if Jedi raises an error on one usage, the other usages it has found may be unrecoverable (as of Jedi v0.10.0).