with variables defined as class members as follows

Code: Select all

Eigen::Matrix<float,Eigen::Dynamic,Eigen::Dynamic,Eigen::ColMajor>      m_systemMatrix;
Eigen::ArrayXf                                                           m_columnSums;

in my code I am calling

Code: Select all

//int rowNumber (method param)
m_columnSums += m_systemMatrix.row(rowNumber).array();

Which compile-error dies as

Code: Select all

local_code\trunk\thirdparty\eigen\eigen\src/Core/ArrayWrapper.h(58): error C2440: 'return' : cannot convert from 'const float *' to 'float *'
1>          Conversion loses qualifiers
1>          c:\users\SUPERSECRET\desktop\local_code\trunk\thirdparty\eigen\eigen\src/Core/ArrayWrapper.h(58) : while compiling class template member function 'float *Eigen::ArrayWrapper<ExpressionType>::data(void)'
1>          with
1>          [
1>              ExpressionType=Eigen::Block<Eigen::Matrix<float,-1,-1>,1,-1,false>
1>          ]
1>          SystemMatrix.cpp(117) : see reference to class template instantiation 'Eigen::ArrayWrapper<ExpressionType>' being compiled
1>          with
1>          [
1>              ExpressionType=Eigen::Block<Eigen::Matrix<float,-1,-1>,1,-1,false>
1>          ]
1>

Even if I get the row back const it should be OK because it is not modified. Thoughts?
This seems like I've made a simple error somewhere. I've just tried this on Eigen/default 5297.

edit: actually adding a row vector with a column vector is not allowed, so you have to transpose one:

m_columnSums += m_systemMatrix.row(0).transpose().array();

ggael wrote:edit: actually adding a row vector with a column vector is not allowed, so you have to transpose one:

m_columnSums += m_systemMatrix.row(0).transpose().array();

Is this a little pedantic or would this behavior actually protect someone? For two vectors, being able to omit the transpose seems like a nice 'beautification' feature.