feat(postgresql): Add support for PostgreSQL multi-dimensional arrays

[colli173]

This PR adds support for generating model fields from multi-dimensional postgresql array columns. The approach here is to pass around an additional column attribute array_bounds representing the dimensions of the postgres array column. When generating code, the array_bounds is checked along is_array to determine the dimensions of the resulting struct field. Would be happy to hear alternative approaches or solutions here as well. More info: Postgres Arrays

fixes issue: #1494

Replaces #1651 #2308 #2309

[andrewmbenton]

Thanks for your work and patience on this PR. I'd like to get it merged before the 1.20 release next week. I think all that's needed is to regenerate the test output and resolve conflicts (if there are any) one more time.

[colli173]

Thanks for your work and patience on this PR. I'd like to get it merged before the 1.20 release next week. I think all that's needed is to regenerate the test output and resolve conflicts (if there are any) one more time.

@andrewmbenton thank you for getting this PR some attention. I appreciate it. It likely needs another code review in addition to the CI checks given how things have changed since the initial PR was opened.

[andrewmbenton]

I'm re-running our automated CI checks now, and will do one more code review for sure.

[andrewmbenton]

So there are a couple of things I would change, but I don't want to make you put in more effort on this so I'm happy to take it over the finish line if you don't have time.

At a high level, I would name this new parameter ArrayDims to more accurately reflect its value. It's a small difference but its clearer I think.

The one failing test is not a big deal. I think you just need to add stderr.txt with the correct string to the directory (can copy from main branch).

There is an issue with some endtoend test output changing though. In particular I noticed that the sqlc_slice test no longer has a slice input parameter in the generated query func. So there's something fishy going on there which will definitely need to be resolved. There may be other bad test output as well.

I'm going to attempt to fix the broken test on your branch, and then I can dig in to see if I can discover what's changing the sqlc_slice test.

[colli173]

If you're able to finish up this PR, that'd be great! Otherwise, I'll have time next weekend to work on it. Given the target of 1.20, I know that doesn't line up well timeline-wise. I agree on the ArrayDims name change, it's definitely clearer. If you have any more questions on the implementation, let me know and I'll see if I can get an answer to you during the week. Thanks again