Hi David Reiss – do you have time to perform a review of this work, please?

As you can see from the description, it's a port of work which was originally performed on HSC. The original HSC commit is available here, but, while the basic logic remains the same, I've had to perform a fairly comprehensive rewrite to make it work within LSST's new measurement framework. I've also added some simple tests.

All the work is on PR #30 in meas_base. A Jenkins build (#7248) was successful.

I am adding Simon Krughoff, who has graciously offered to help, as a reviewer.

We went over this and things look good. We had a couple of small questions. One general comment is that I like to see the test as a different commit from the implementation. It just splits things up for me logically a little better.

Feel free to merge.

Thanks both! Comments on the PR look very fair; I'll respond shortly

If I were to split the test and the implementation into separate commits, what order would you like to see them in? Are you suggesting a traditional TDD-style "write a failing test then write the code to make it pass" approach, or would you rather see the implementation first? I prefer the former, but I'm not sure how much of a sin it is to have a known-bad state on a branch.

In theory, I agree that having the test first is nice, but I think the benefits of sticking to the policy that all commits should build has a greater benefit in practice. As a reviewer, I'm not sure I care whether it comes first or second as long as the code implementation and test implementation are separate.

Our only guarantee is that master must build; branches need not at this time. I believe it is true that having all commits build helps things like git bisect. One possibility (but one that mixes some file changes) is to add the test but have it skipped or marked "should fail" until the implementation is added.

Split test into a separate commit (after the implementation, for the record). Responded to comments on GitHub, notably adding some commentary around the choice of default value for the scale factor. Created DM-4861 to request some meas_base documentation.

Merged.

Release notes updated.