spin 3 postgres support

[michelleN]

This also generates the bindings for wasi kv and wasi config as they're now part of the spin platform world.

[fibonacci1729]

Looks good!

[itowlson]

I don't remember if we discussed feature-gating the 3.0 stuff - can we just confirm that we happy to lean on dead code elimination rather than an explicit Cargo feature?

Otherwise looks good except I was puzzled by the wrapper types for decoding date-time types. We don't wrap other types - why these?

Also wondering if we need to do anything on the write side. What does it look like to insert a date-time value via a parameter?

[itowlson]

This brought a smile to my morning!

[itowlson]

Is there a reason we need a wrapper struct here rather than decoding straight to chrono::NaiveDate? It feels like the user can't do anything the wrapper except call .0 on it to extract the NaiveDate: is it possible to cut out the middleman?

[fibonacci1729]

The ole can't implement traits on foreign types.

[fibonacci1729]

Wait maybe I gave poor advice here. is it can't implement foreign traits on foreign types?

[itowlson]

Yeah you can't implement foreign traits on foreign types but you can totally implement local traits on foreign types.

[fibonacci1729]

🤦 yep yep yep.

[itowlson]

Indeed, we do exactly that on all the other types - stdlib types still count as foreign for orphan rule purposes.

[itowlson]

Can we skip the wrapper again?

[itowlson]

Can we skip the wrapper again again?

[michelleN]

I don't remember if we discussed feature-gating the 3.0 stuff - can we just confirm that we happy to lean on dead code elimination rather than an explicit Cargo feature?

yep, relying on dead code elimination. ran it past @fibonacci1729. I think @dicej is on the same page (feel free to chime in if not Joel). Are there any benefits to feature gating other than it being more explicit?

[itowlson]

Nah, the only thing I can think of is making the choice to opt into v3-only stuff more explicit, but it's not a big deal. I'm happy to stick with what you have.

[michelleN]

Otherwise looks good except I was puzzled by the wrapper types for decoding date-time types. We don't wrap other types - why these?

you're right. they were unnecessary. updated!

[michelleN]

Also wondering if we need to do anything on the write side. What does it look like to insert a date-time value via a parameter?

Here's an example of date: https://github.com/fermyon/spin-rust-sdk/pull/39/files#diff-57b15e5f7dd54bae6c97fe6ddf112edff16dc9eb6014d269f80d320280241f4cR94
Should I add a datetime example as well?

[itowlson]

Sorry, by "as a parameter" I meant via execute parameters rather than by embedding literally into a SQL string.

E.g.

let title: String = figure_out_the_title();
let date: chrono::NaiveDate = figure_out_the_date();
conn.execute("INSERT INTO articletest(title, published) VALUES ($1, $2)", &[ /* what */ ]);

It's reasonably obvious for strings and ints and so on, I just want to see if it's still easy with dates and times where the WITs are complicated and (as you identified when we talked about it) not very obvious.

[michelleN]

@itowlson good call. We added an example of doing the writes and more helpers in the module. Lmk if you see anything else we should add.

[itowlson]

This looks great. Thank you for turning it around so quickly, and for beefing up the example - this makes it a lot more helpful!