Drake
Drake C++ Documentation
Python Bindings

Details on implementing python bindings for the C++ code.

Overview

Drake uses pybind11 for binding its C++ API to Python.

At present, a fork of pybind11 is used which permits bindings matrices with dtype=object, passing unique_ptr objects, and prevents aliasing for Python classes derived from pybind11 classes.

Before delving too deep into this, please first review the user-facing documentation about What's Available from Python.

Module Organization

The structure of the bindings generally follow the directory structure, not the namespace structure. As an example, the following code in C++:

will look similar in Python, but you won't use the header file's name:

from pydrake.multibody.parsing import Parser
from pydrake.multibody.plant import MultibodyPlant

In general, you can find where a symbol is bound by searching for the symbol's name in quotes underneath the directory drake/bindings/pydrake.

For example, you should search for "MultibodyPlant", "Parser", etc. To elaborate, the binding of Parser is found in .../pydrake/multibody/parsing_py.cc, and looks like this:

using Class = Parser;
py::class_<Class>(m, "Parser", ...)

and binding of MultibodyPlant template instantiations are in .../pydrake/multibody/plant_py.cc and look like this:

using Class = MultibodyPlant<T>;
DefineTemplateClassWithDefault<Class, systems::LeafSystem<T>>(
m, "MultibodyPlant", ...)

where the function containing this definition is templated on type T and invoked for the scalar types mentioned in drake/common/default_scalars.h.

Some (but not all) exceptions to the above rules:

pybind11 Tips

Python Types

Throughout the Drake code, Python types provided by pybind11 are used, such as py::handle, py::object, py::module, py::str, py::list, etc. For an overview, see the pybind11 reference.

All of these are effectively thin wrappers around PyObject*, and thus can be cheaply copied.

Mutating the referred-to object also does not require passing by reference, so you can always pass the object by value for functions, but you should document your method if it mutates the object in a non-obvious fashion.

Python Type Conversions

You can implicit convert between py::object and its derived classes (such as py::list, py::class_, etc.), assuming the actual Python types agree. You may also implicitly convert from py::object (and its derived classes) to py::handle.

If you wish to convert a py::handle (or PyObject*) to py::object or a derived class, you should use py::reinterpret_borrow<>.

Conventions

API

Any Python bindings of C++ code will maintain C++ naming conventions, as well as Python code that is directly related to C++ symbols (e.g. shims, wrappers, or extensions on existing bound classes).

All other Python code be Pythonic and use PEP 8 naming conventions.

For binding functions or methods, argument names should be provided that correspond exactly to the C++ signatures using py::arg("arg_name"). This permits the C++ documentation to be relevant to the Sphinx-generated documentation, and allows for the keyword-arguments to be used in Python.

For binding functions, methods, properties, and classes, docstrings should be provided. These should be provided as described here.

Testing

In general, since the Python bindings wrap tested C++ code, you do not (and should not) repeat intricate testing logic done in C++. Instead, ensure you exercise the Pythonic portion of the API, using kwargs when appropriate.

When testing the values of NumPy matrices, please review the documentation in pydrake.common.test_utilities.numpy_compare for guidance.

Target Conventions

Names

File names should follow form with their respective target.

Visibility

Bazel

Given that libdrake.so relies on static linking for components, any common headers should be robust against ODR violations. This can be normally achieved by using header-only libraries.

For upstream dependencies of these libraries, do NOT depend on the direct targets (e.g. //common:essential), because this will introduce runtime ODR violations for objects that have static storage (UID counters, etc.).

Instead, you must temporarily violate IWYU because it will be satisfied by drake_pybind_library, which will incorporate libdrake.so and the transitive headers.

If singletons are required (e.g. for util/cpp_param_pybind), consider storing the singleton values using Python.

If you are developing bindings for a small portion of Drake and would like to avoid rebuilding a large number of components when testing, consider editing //tools/install/libdrake:build_components.bzl to reduce the number of components being built.

pybind Module Definitions

Documentation

Drake uses a modified version of mkdoc.py from pybind11, where libclang Python bindings are used to generate C++ docstrings accessible to the C++ binding code.

These docstrings are available within constexpr struct ... pydrake_doc as const char* values . When these are not available or not suitable for Python documentation, provide custom strings. If this custom string is long, consider placing them in a heredoc string.

An example of incorporating docstrings from pydrake_doc:

#include "drake/bindings/pydrake/documentation_pybind.h"
PYBIND11_MODULE(math, m) {
using namespace drake::math;
constexpr auto& doc = pydrake_doc.drake.math;
using T = double;
py::class_<RigidTransform<T>>(m, "RigidTransform", doc.RigidTransform.doc)
.def(py::init(), doc.RigidTransform.ctor.doc_0args)
...
.def(py::init<const RotationMatrix<T>&>(), py::arg("R"),
doc.RigidTransform.ctor.doc_1args_R)
.def(py::init<const Eigen::Quaternion<T>&, const Vector3<T>&>(),
py::arg("quaternion"), py::arg("p"),
doc.RigidTransform.ctor.doc_2args_quaternion_p)
...
.def("set_rotation", &RigidTransform<T>::set_rotation, py::arg("R"),
doc.RigidTransform.set_rotation.doc)
...
}

An example of supplying custom strings:

constexpr char another_helper_doc[] = R"""(
Another helper docstring. This is really long.
And has multiple lines.
)""";
PYBIND11_MODULE(example, m) {
m.def("helper", []() { return 42; }, "My helper method");
m.def("another_helper", []() { return 10; }, another_helper_doc);
}
Note
Consider using scoped aliases to abbreviate both the usage of bound types and the docstring structures. Borrowing from above:
{
using Class = RigidTransform<T>;
constexpr auto& cls_doc = doc.RigidTransform;
py::class_<Class>(m, "RigidTransform", cls_doc.doc)
.def(py::init(), cls_doc.ctor.doc_0args)
...
}

To view the documentation rendered in Sphinx:

bazel run //doc/pydrake:serve_sphinx [-- --browser=false]
Note
Drake's online Python documentation is generated on Ubuntu Jammy, and it is suggested to preview documentation using this platform. Other platforms may have slightly different generated documentation.

To browse the generated documentation strings that are available for use (or especially, to find out the names for overloaded functions' documentation), generate and open the docstring header:

bazel build //bindings/pydrake:documentation_pybind.h
$EDITOR bazel-bin/bindings/pydrake/documentation_pybind.h

Search the comments for the symbol of interest, e.g., drake::math::RigidTransform::RigidTransform<T>, and view the include file and line corresponding to the symbol that the docstring was pulled from.

Note
This file may be large, on the order of ~100K lines; be sure to use an efficient editor!
If you are debugging a certain file and want quicker generation and a smaller generated file, you can hack mkdoc.py to focus only on your include file of chioce. As an example, debugging mathematical_program.h:
...
assert len(include_files) > 0 # Existing code.
include_files = ["drake/solvers/mathematical_program.h"] # HACK
This may break the bindings themselves, and should only be used for inspecting the output.

For more detail:

Deprecation

Decorators and utilities for deprecation in pure Python are available in pydrake.common.deprecation.

Deprecations for Python bindings in C++ are available in drake/bindings/pydrake/common/deprecation_pybind.h.

For examples of how to use the deprecations and what side effects they will have, please see:

Note
All deprecations in Drake should ultimately use the Python warnings module, and the DrakeDeprecationWarning class. The utilities mentioned above use them.

Keep Alive Behavior

py::keep_alive<Nurse, Patient>() is used heavily throughout this code. Please first review the pybind11 documentation.

py::keep_alive decorations should be added after all py::args are specified. Terse comments should be added above these decorations to indicate the relationship between the Nurse and the Patient and decode the meaning of the Nurse and Patient integers by spelling out either the py::arg name (for named arguments), return for index 0, or self (not this) for index 1 when dealing with methods / members. The primary relationships:

Some example comments:

// Keep alive, reference: `self` keeps `context` alive.
// Keep alive, ownership (tr.): `return` keeps `self` alive.

Return Value Policy

For more information about pybind11 return value policies, see the pybind11 documentation.

pydrake offers the py_rvp alias to help with shortened usage of py::return_value_policy. The most used (non-default) policies in pydrake are reference and reference_internal due to the usage of raw pointers / references in the public C++ API (rather than std::shared_ptr<>).

Note
While py_rvp::reference_internal effectively implies py_rvp::reference and py::keep_alive<0, 1>(), we choose to only use it when self is the intended patient (i.e. the bound item is a class method). For static / free functions, we instead explicitly spell out py_rvp::reference and py::keep_alive<0, 1>().

Function Overloads

To bind function overloads, please try the following (in order):

Mutable vs. const Method Overloads

C++ has the ability to distinguish T and const T for both function arguments and class methods. However, Python does not have a native mechanism for this. It is possible to provide a feature like this in Python (see discussion and prototypes in #7793); however, its pros (similarity to C++) have not yet outweighted the cons (awkward non-Pythonic types and workflows).

When a function is overloaded only by its const-ness, choose to bind the mutable overload, not the const overload.

We do this because pybind11 evaluates overloads in the order they are bound, and is not possible for a user to "reach" any overloads that can only be disambiguated by const-ness.

Examples of functions to bind and not bind:

// N.B. The two free functions below aren't necessarily great in terms of coding
// and the GSG; however, we use them for illustrative purposes.
// BIND: Should be exposed.
void MyFunction(MyClass* mutable_value);
// DO NOT BIND: Identical to above mutable overload, may constrain user.
void MyFunction(const MyClass& value);
// Disambiguating an accessor solely based on `const`-ness of `this`. In
// general, you may want to avoid this.
class MyOtherClassDispreferred {
public:
...
// BIND: Even though more costly, this ensure the user has access to the
// "maximum" amount of functionality.
MyClass& value();
// DO NOT BIND: Identical to above mutable overload.
const MyClass& value() const;
};
class MyOtherClassPreferred {
public:
...
// BIND: Unambiguous.
MyClass& mutable_value();
// BIND: Unambiguous.
const MyValue& value() const;
};

Public C++ API Considerations for Function and Method Templates

The motivation behind this section can be found under the "C++ Function and Method Template Instantiations in Python" section in doc/python_bindings.rst.

In general, Drake uses techniques like parameter packs and type erasure to create sugar functions. These functions map their inputs to parameters of some concrete, under-the-hood method that actually does the work, and is devoid of such tricks. To facilitate python bindings, this underlying function should also be exposed in the public API.

As an example for parameter packs, MultibodyPlant<T>::AddJoint<JointType, Args...>(...) (code permalink) is a C++ sugar method that uses parameter packs and ultimately passes the result to MultibodyPlant<T>::AddJoint<JointType>(unique_ptr<JointType>) (code permalink), and only the unique_ptr function is bound (code permalink):

using Class = MultibodyPlant<T>;
...
.def("AddJoint",
[](Class* self, std::unique_ptr<Joint<T>> joint) -> auto& {
return self->AddJoint(std::move(joint));
},
py::arg("joint"), py_rvp::reference_internal, cls_doc.AddJoint.doc_1args)
...

As an example for parameter packs, GeometryProperties::AddProperty<ValueType> (code permalink) is a C++ sugar method that uses type erasure and ultimately passes the result to GeometryProperties::AddPropertyAbstract (code permalink), and only the AddPropertyAbstract flavor is used in the bindings, but in such a way that it is similar to the C++ API for AddProperty (code permalink):

using Class = GeometryProperties;
py::handle abstract_value_cls =
py::module::import("pydrake.systems.framework").attr("AbstractValue");
...
.def("AddProperty",
[abstract_value_cls](Class* self, const std::string& group_name,
const std::string& name, py::object value) {
py::object abstract = abstract_value_cls.attr("Make")(value);
self->AddPropertyAbstract(
group_name, name, abstract.cast<const AbstractValue&>());
},
py::arg("group_name"), py::arg("name"), py::arg("value"),
cls_doc.AddProperty.doc)
...

Matrix-multiplication-like Methods

For objects that may be represented by matrices or vectors (e.g. RigidTransform, RotationMatrix), the * operator (via __mul__) should not be bound because the * operator in NumPy implies elemnt-wise multiplication for arrays.

For simplicity, we instead bind the explicitly named .multiply() method, and alias the __matmul__ operator @ to this function.

Clone Methods

If you wish to bind a Clone() method, please use DefClone() so that __copy__() and __deepcopy__() will also be defined. (It is simplest to do these things together.)

Returning Vectors or Matrices

Certain bound methods, like RigidTransform.multiply(), will have overloads that can multiply and return (a) other RigidTransform instances, (b) vectors, or (c) matrices (representing a list of vectors).

In the cases of (a) and (c), pybind11 provides sufficient mechanisms to provide an unambiguous output return type. However, for (b), pybind11 will return ndarray with shape (3,). This can cause an issue when users pass a vector of shape (3, 1) as input. Nominally, pybind11 will return a (3,) array, but the user may expect (3, 1) as an output. To accommodate this, you should use the drake::pydrake::WrapToMatchInputShape function.

See also
https://github.com/RobotLocomotion/drake/issues/13885

Python Subclassing of C++ Classes

In general, minimize the amount in which users may subclass C++ classes in Python. When you do wish to do this, ensure that you use a trampoline class in pybind, and ensure that the trampoline class inherits from the py::wrapper<> class specific to our fork of pybind. This ensures that no slicing happens with the subclassed instances.

Interactive Debugging with Bazel

If you are debugging a unitest, first try running the test with --trace=user to see where the code is failing. This should cover most cases where you need to debug C++ bits. Example:

bazel run //bindings/pydrake/systems:py/lifetime_test -- --trace=user

If you need to debug further while using Bazel, it is suggested to use gdbserver for simplicity. Example:

# Terminal 1 - Host process.
bazel run -c dbg \
--run_under='gdbserver localhost:9999' \
//bindings/pydrake/systems:py/lifetime_test -- \
--trace=user
# Terminal 2 - Client debugger.
gdb -ex "dir ${PWD}/bazel-drake" \
-ex "target remote localhost:9999" \
-ex "set sysroot" \
-ex "set breakpoint pending on"
# In the GDB terminal:
(gdb) continue

set sysroot is important for using gdbserver, set breakpoint pending on allows you set the breakpoints before loading libdrake.so, and dir ... adds source directories. It is also suggested that you enable readline history in ~/.gdbinit for ease of use.

If using CLion, you can still connect to the gdbserver instance.

There are analogs for lldb / lldbserver but for brevity, only GDB is covered.