Clean up code - #30
Conversation
These changes are a no op on mrcal 2.5 anyways.
| int *c_imagersizes = imagersize; | ||
| auto point_min_range = -1.0, point_max_range = -1.0; | ||
| mrcal_problem_constants_t problem_constants = { | ||
| .point_min_range = point_min_range, .point_max_range = point_max_range}; |
There was a problem hiding this comment.
If you follow the logic for this, you'll see this struct is only accessed by function calls in an #ifdef 0 block, so this is never accessed. Future versions of mrcal empty the struct out entirely.
There was a problem hiding this comment.
penalty_range_normalization is only used in https://github.com/dkogan/mrcal/blob/c311b0acdb29d3f6c1a5abeaf17dc6a7e2ab10d9/mrcal.c#L5212 which has the comment:
TEMPORARY TWEAK: disable range normalization: I will re-add this later
But yeah, main doesn't have it at all. The if 0 was added in dkogan/mrcal@0e72718 , and was fully removed in dkogan/mrcal@e65c1b7 . "This is probably never coming back, so I'm removing the variables I'm not using anymore and the disabled code"
At the API layer there's no guarantee this is totally unused. mrcal.h makes no statement on if it's used or not. I think passing null will quickly cause a crash if he does try to access, so let's do that.
There was a problem hiding this comment.
Harder for me but sure. I had to really go digging for that Python code though. (and there's no comment indicating it was taken from Python, the one that exists is further down.)
|
This diff passes the tests in wpical on Windows. |
2 participants
Focuses on compatibility with newer mrcal and OpenCV versions, but also does some optimizations, like being more precise with headers, killing iostream, and removing 2-3 copies of point structures (which is actually a lot of memory to be shuffling around.)