IDEs, Debugging, and Profiling

ChimeraX is a mostly Python project and ships mostly Python code alongside C++ extensions. There are over 260 packages in the distributed ChimeraX, including over 150 bundles. This is an unwieldy amount of code for any one person to keep in their head all at once. The use of an IDE, which can provide code completion and in-editor documentation, is highly encouraged.

IDEs

Popular IDEs include PyCharm and VS Code. Your choice of IDE is up to you; many share the same feature set thanks to the language server protocol (LSP) and the debug adapter protocol (DAP).

When you first open a new IDE, it will probably try to set its Python interpreter to the system Python. It is better to use a virtual environment based on ChimeraX’s internal Python as the Python interpreter instead.

If you set ChimeraX’s internal interpreter directly as your IDE interpreter, the IDE may become confused and report that it cannot find certain modules or that certain modules which are present in the ChimeraX distribution are not actually installed.

Virtual environments solve both problems, and many others. You can install packages into the virtual environment without corrupting the internal Python environment, and trash the virtual environment and start again if something goes wrong.

First, build ChimeraX or obtain a built ChimeraX. If you have built ChimeraX, then you can use the makefile to create a virtual environment with the correct settings in your repository directory:

make venv

If you are working off of the ChimeraX distribution, use the internal Python to create a virtual environment in your project’s folder. The internal Python can be found at the following locations:

  • Windows: C:\path\to\ChimeraX.app\bin\python.exe

  • macOS: /path/to/ChimeraX.app/Contents/bin/python

  • Linux: /path/to/ChimeraX.app/bin/python

Ensure that the virtual environment has access to ChimeraX’s site-packages directory, and then set that virtual environment’s Python as the project’s Python interpreter in your IDE.

Example commands:

# Windows
C:\path\to\ChimeraX.app\bin\python.exe -m venv .venv --system-site-packages
# macOS
/path/to/ChimeraX.app/Contents/bin/python -m venv .venv --system-site-packages
# Linux
/path/to/ChimeraX.app/bin/python -m venv .venv --system-site-packages

Debugging

Since PyCharm uses its own wrapper around pydevd instead of debugpy, we’ll start with steps for PyCharm.

  • Create a new Run/Debug Configuration

  • Set the script path field to point at a module instead

  • Set the module to “chimerax.core”

  • Set the Python interpreter to ChimeraX’s internal Python

  • Uncheck “Add content roots to PYTHONPATH”

  • Uncheck “Add source roots to PYTHONPATH”

We recommend the following profiles for programs using debugpy as the interface to pydevd (the majority of them).

{
    "name": "ChimeraX (GUI)",
    "type": "python",
    "request": "launch",
    "module": "chimerax.core",
    # Ensure we can pause ChimeraX, not just break in our module
    "justMyCode": false
},
{
    "name": "ChimeraX (NoGUI)",
    "type": "python",
    "request": "launch",
    "module": "chimerax.core",
    "args": ["--nogui"],
    # Ensure we can pause ChimeraX, not just break in our module
    "justMyCode": false,
    "console": "externalTerminal"
}

Breaking on raise leads to slow execution in the debugger, but breaking on an uncaught exception or a user uncaught exception provides a tight enough net to catch errors in ChimeraX code.

ChimeraX spends the majority of its time in its GUI event loop. Unless you set a breakpoint in your bundle then upon pausing execution you will likely be dropped into the context of the GUI module.

There, the GUI module’s reference to the program’s session will be available so that you can get an idea of what’s going on in your debugging session.

Line Profiling

Line profiling is based on Robert Kern’s line_profiler package. Support is restricted to platforms that have binaries in pypi.org (just Linux for now). Differences from the conventional setup are given in parenthesizes.

There are five parts to profiling:

  1. Decorate functions that you wish to profile with the @line_profile decorator and install them. This decorator is a no-op if ChimeraX is not profiled. (Instead of the conventional @profile.)

  2. Run ChimeraX --lineprofile from the command line. (Instead of using kernprof.) This generates a ChimeraX.lprof file in the current directory.

  3. Get your profiling results by running ChimeraX -m line_profiler ChimeraX.lprof.

  4. Analyze your results, possibly edit your code, and return to step 1.

  5. Remove the function decorators before committing your changes.

To use the timeit module, see the scripting documentation.