Fix 2D Nx1/1xN inputs to eigen dense vector args
This fixes a bug introduced in b68959e822a619d1ad140f52c7197a29f4e6a06a
when passing in a two-dimensional, but conformable, array as the value
for a compile-time Eigen vector (such as VectorXd or RowVectorXd). The
commit switched to using numpy to copy into the eigen data, but this
broke the described case because numpy refuses to broadcast a (N,1)
into a (N).
This commit fixes it by squeezing the input array whenever the output
array is 1-dimensional, which will let the problematic case through.
(This shouldn't squeeze inappropriately as dimension compatibility is
already checked for conformability before getting to the copy code).
diff --git a/tests/test_eigen.cpp b/tests/test_eigen.cpp
index 17b156c..20be0aa 100644
--- a/tests/test_eigen.cpp
+++ b/tests/test_eigen.cpp
@@ -288,6 +288,13 @@
m.def("iss738_f1", &adjust_matrix<const Eigen::Ref<const Eigen::MatrixXd> &>, py::arg().noconvert());
m.def("iss738_f2", &adjust_matrix<const Eigen::Ref<const Eigen::Matrix<double, -1, -1, Eigen::RowMajor>> &>, py::arg().noconvert());
+ // test_issue1105
+ // Issue #1105: when converting from a numpy two-dimensional (Nx1) or (1xN) value into a dense
+ // eigen Vector or RowVector, the argument would fail to load because the numpy copy would fail:
+ // numpy won't broadcast a Nx1 into a 1-dimensional vector.
+ m.def("iss1105_col", [](Eigen::VectorXd) { return true; });
+ m.def("iss1105_row", [](Eigen::RowVectorXd) { return true; });
+
// test_named_arguments
// Make sure named arguments are working properly:
m.def("matrix_multiply", [](const py::EigenDRef<const Eigen::MatrixXd> A, const py::EigenDRef<const Eigen::MatrixXd> B)