Status: To Do
Fix Version/s: None
Team:Data Release Production
lsst.afw.math.Background originally supported retrieval of float, double, and int images. As part of
DM-2778, the double and int overloads were disabled because Approximate did not "yet" support values other than float. However, the templated form of Background::getImage<T> was kept, even though it now only compiles if T = float. This causes surprising compilation errors in new code that is unaware of the implementation details.
Try to re-enable double and int support. This will likely require work on Approximate. Replace getImage<T> with ordinary methods. This will break (or at least require a deprecation period for) both C++ and Python code, but leads to a cleaner API in the end.
- Document and/or static_assert that T must be float, so that at least users who run into this issue don't have to hunt down the cause.
I don't know which of these solutions is the best, given the current state of the code (and the apparently never-finished