CODEWITHSUNDEEP
c++eigen
Saeid
Saeid

Reputation: 558

The correct indexation in Eigen

Suppose I have a Matrix <float, Dynamic, Dynamic, RowMajor> A in Eigen. When i write following code:

cout << "Number of Columns of A is: "<< A.cols() << endl;
cout << "Number of Rows of A is: "<< A.rows() << endl;

I get following result:

Number of Columns of A is: 129
Number of Rows of A is: 600

According to the above results, i expect when i write following code, i get Exception Error but that does not happen and it print a value!!! WHY??!!

cout << A(500,140);

Upvotes: 0

Views: 1328

Answers (1)

Massimiliano Janes
Massimiliano Janes

Reputation: 5624

As explained in Eigen docs, matrix coefficients can be accessed either via m(i,j) or m.coeff(i,j)/m.coeffRef(i,j) (plus m[i] and m.x(),... for the special vector case).

Now, the 'm(i,j)' variant is range-checked unless the NDEBUG or EIGEN_NO_DEBUG macros are defined. Typically, the former macro is defined for 'release' builds, so no range checks will be performed in that case. The rationale is that Eigen is a performance oriented library, checks have a cost, so it makes sense enabling them for debugging purposes only.

The m.coeff(i,j) form is never checked.

When no check is performed, any attempt to invoke a coefficient accessor out of range is a precondition violation, this means the behaviour is undefined.

Generally speaking, you should minimize the use of indexed access in favor of higher level, block/linear algebra operations (Eigen has plenty of them); your code will result more compact, more readable (well, at least, to an algebra-aware reader), more correct (less risk of out of range accesses) and (possibly much) faster.

Upvotes: 5

Related Questions