Here is a message from Kendrick that he sent to the hsc software list:
I've been working on PSF modeling recently, and wanted to publicize some code, now that it's getting fairly mature and integrated with the rest of the HSC pipeline:
git clone https://github.com/kmsmith137/meas_extensions_hscpsf.git
What's currently implemented on the master branch is a PCA-style PSF model, which was initially based on PcaPsf in the LSST pipeline but ended up deviating quite a bit. (Unlike previous versions of the code which some of you may have played with, this version runs as a standalone meas_extensions module in the current HSC pipeline, and doesn't require boost_python + a lot of my pet libraries!)
Another thing which is nearly complete is a reimplementation of the psfex PSF model using low-level classes and libraries from the LSST/HSC pipeline. The idea is to have the functionality of psfex in a form which is familiar and easy to modify for anyone familiar with the HSC pipeline (the reimplementation will also much smaller by lines of code). It's not quite ready yet but the current semi-complete version is on the psfex_hacking branch if anyone is curious to take a look.
It's currently unclear to me whether we'll ultimately want to use a PCA-style PSF model, a psfex-style PSF model, or something else. What I'd like to do in the short term is to have both models implemented in a cleaned-up class hierarchy which is easy to modify/extend, and to have a well-defined suite of tests for evaluating a given PSF model. Bob Armstrong at Princeton and I have been coordinating recently on this.
If you get a chance to play with the code in its current form, please let me know your feedback!
Thanks and more updates coming soon!