default_scalars_pybind.h File Reference

Detailed Description

Helpers for defining scalars and values.

#include "drake/bindings/pydrake/autodiff_types_pybind.h"
#include "drake/bindings/pydrake/common/cpp_template_pybind.h"
#include "drake/bindings/pydrake/common/type_pack.h"
#include "drake/bindings/pydrake/pydrake_pybind.h"
#include "drake/bindings/pydrake/symbolic_types_pybind.h"
#include "drake/common/default_scalars.h"

Include dependency graph for default_scalars_pybind.h:

Namespaces
	drake
	drake::pydrake
	For more high-level information, see the Python Bindings technical notes.

Typedefs
using	CommonScalarPack = type_pack< double, AutoDiffXd, symbolic::Expression >
	Type pack defining common scalar types. More...
using	NonSymbolicScalarPack = type_pack< double, AutoDiffXd >
	Type pack for non-symbolic common scalar types. More...

Functions
template<typename T , typename PyClass , typename UPack = typename internal::CastUPack<T>::Pack>
void	DefCast (PyClass *cls, const char *doc, UPack U_pack={})
	Binds cast<T>() explicitly. More...
Handling special non-POD scalar types.
Because we use dtype=object in NumPy, we cannot alias (share references to) the data underlying matrix objects when passing data between NumPy and Eigen (see #8116). The simple policy these functions help enforce: When T is double, allow referencing (do not require copying). When T is not double (i.e., AutoDiffXd, Expression), copy data (do not reference the data).
template<typename T >
py::return_value_policy	return_value_policy_for_scalar_type ()
	Permits referencing for builtin dtypes (e.g., T == double), but then switches to copying for custom dtypes (T ∈ {AutoDiffXd, Expression}). More...
template<typename SomeBlock >
decltype(auto)	CopyIfNotPodType (const SomeBlock &x)
	A no-op for builtin dtypes (e.g., T == double), but then switches to copying for custom dtypes (T ∈ {AutoDiffXd, Expression}). More...