--DO NOT MERGE--
The purpose of this is just to let me run automated tests on these modifications to the lifting system, to see whether this preserves the functionality for IDs and embeddings, and in particular passes the tests. I'm unfamiliar, in particular, if introducing Maybes to those areas of code will break things (although the non-Maybe version could be restored easily).
Experimentation with a better lifting model.
You can run the automated tests locally that are in persistent-test. There are some instructions in the README.
Fix mistaken lift.
This is a bit of a fragile approach. I guess I see why the previous approach was
used. I can't honestly say which is better.
Oh, ok. Well, hopefully this run will be more enlightening than my stupid screwups :)
The build made it farther, but a similar error occurred.
durdle a bit
After some experimentation, I'm firmly of the opinion that a) the approach in my previous patch is wrong and b) I don't know how to fix it. I'll upload my progress and someone can try to fix the bugs that come up.
thanks! The test suite is passing now which is encouraging.
Yeah, but parts of yesod break if I try to build with it. Bug in the testsuite, I guess.
The code in question has changed a lot since this patch was created. I think some of the idea about improving SqlTypeExp was incorporated.