Fixed the determination of scale and rotation matrix from the CDij matrix #6573
Conversation
ayshih
changed the title
ayshih
changed the title
sunpy/map/mapbase.py
Outdated
| If the CDij matrix is defined but no CDELTi values are explicitly defined, | ||
| effective CDELTi values are constructed from the CDij matrix. |
There was a problem hiding this comment.
I think it would be good to explain how here, because as you pointed out in #6570 there's not one singular "correct" way to do this. (Which is why it didn't used to do it)
There was a problem hiding this comment.
Do you want this in this PR or another one?
There was a problem hiding this comment.
+1 to documenting how .scale is calculated without CDELT metadata present in this PR.
What's the reasoning for this choice? Does it make the resulting scale have some nice mathematica; properties? |
|
The reason to make this choice (which is the optional constraint suggested in Paper I, equation 4) is so that if the rows of the CDij matrix are orthogonal, then the rows of the PCij matrix are orthonormal. That's your best shot at getting to a matrix that is a pure rotation (or reflection), if that is possible. |
|
Doc fail is due to another PR that I merged that broke it. |
… matrix from the CDij matrix
… matrix from the CDij matrix
.rotation_matrixproperty where the PCij matrix was not determined correctly from the CDij matrix in the case of non-square pixels..scaleproperty to return reasonable values when a map has a CDij matrix and no CDELTi values. The returned pixel scales are the values that are needed so that each row of the inferred PCij matrix has unity norm.