new type support in query_as

[0xdeafbeef]

fixes #2173

use sqlx::PgPool;


struct Wrap(bigdecimal::BigDecimal);

impl From<bigdecimal::BigDecimal> for Wrap {
    fn from(x: bigdecimal::BigDecimal) -> Self {
        Wrap(x)
    }
}

impl From<Wrap> for bigdecimal::BigDecimal {
    fn from(x: Wrap) -> Self {
        x.0
    }
}

struct Data {
    pub amount: Wrap,
}


#[tokio::main]
async fn main() {
    let db = PgPool::connect("postgres://postgres:mysecretpassword@localhost:5432/teststore").await.unwrap();

    sqlx::query_as!(Data, "select amount from test").fetch_one(&db).await.unwrap();
    // previously it will expand like this:
     {
         {
             #[allow(clippy::all)] {
                 use ::sqlx::Arguments   as _;
                 let query_args = <sqlx::postgres::Postgres as ::sqlx::database::HasArguments>::Arguments::default();
                 ::sqlx::query_with::<sqlx::postgres::Postgres, _>("select amount from test", query_args).try_map(|row: sqlx::postgres::PgRow| {
                     use ::sqlx::Row   as _;
                     let sqlx_query_as_amount = row.try_get_unchecked::<sqlx::types::BigDecimal, _>(0usize)?;
                     Ok(Data { r#amount: sqlx_query_as_amount })
                 })
             }
         }
     }

 now it will expand like this:

let sqlx_query_as_amount = row.try_get_unchecked::<sqlx::types::BigDecimal, _>(0usize)?.into(); <-----  

This change shouldn't break anything, because From<T> is implemented for T.

Maybe we can also use From for binded arguments in sqlx::query

example:

let t = "string";
sqlx::query!("select $1",  t)

Here we can use t.into() internally, but it can break inference

[abonander]

There was a reason I didn't do this and I believe it had to do with compile errors. Can you test what happens when you try to use an incompatible type?

[0xdeafbeef]

There was a reason I didn't do this and I believe it had to do with compile errors. Can you test what happens when you try to use an incompatible type?

   |
27 |     sqlx::query_as!(Record2,r#"select 1 as "amount!""#).fetch_one(&db).await.unwrap();
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the trait `From<i32>` is not implemented for `std::string::String`
   |
   = help: the following other types implement trait `From<T>`:
             <std::string::String as From<&mut str>>
             <std::string::String as From<&std::string::String>>
             <std::string::String as From<&str>>
             <std::string::String as From<Box<str>>>
             <std::string::String as From<Cow<'a, str>>>
             <std::string::String as From<char>>
             <std::string::String as From<url::Url>>
   = note: required for `i32` to implement `Into<std::string::String>`
   = note: this error originates in the macro `$crate::sqlx_macros::expand_query` which comes from the expansion of the macro `sqlx::query_as` (in Nightly builds, run with -Z macro-backtrace for more info)

I'll try to add some type checking to make errors more obvious.

The current error looks like this:

  --> src/main.rs:27:5
   |
27 |     sqlx::query_as!(Record2, r#"select 1 as "amount!""#).fetch_one(&db).await.unwrap();
   |     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |     |
   |     expected struct `std::string::String`, found `i32`
   |     help: try using a conversion method: `.to_string()`
   |
   = note: this error originates in the macro `$crate::sqlx_macros::expand_query` which comes from the expansion of the macro `sqlx::query_as` (in Nightly builds, run with -Z macro-backtrace for more info)

[abonander]

Yeah, neither error is great. Something I want to do shortly but won't make the 0.7.0 release is rewrite the typechecking code entirely so that it also works with FromRow. I figure I can use panic!() messages in const to give much more useful error messages.

[CosmicHorrorDev]

rewrite the typechecking code entirely so that it also works with FromRow. I figure I can use panic!() messages in const to give much more useful error messages.

Yes please! Both using FromRow and giving better error messages would be amazing!

[abonander]

The annoying part is that panic!() in const doesn't support formatting AFAIK, so I'd have to use a dirty hack like the following: https://github.com/Vurich/const-concat

As-implemented, I think that macro would trigger #[deny(unsafe_code)] if the user has it set. I think it can be cleaned up with const-generics though.

[abonander]

I suppose concatenation isn't really necessary as I could just emit two different consts per check, e.g.:

const _: () = panic!("field `foo` of struct `Bar` expected type `i32`");
const _: () = panic!("column `foo` of query has SQL type `TEXT`");

gives something like:

   Compiling playground v0.0.1 (/playground)
error[[E0080]](https://doc.rust-lang.org/stable/error_codes/E0080.html): evaluation of constant value failed
 --> src/lib.rs:1:15
  |
1 | const _: () = panic!("field `foo` of struct `Bar` expected type `i32`");
  |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the evaluated program panicked at 'field `foo` of struct `Bar` expected type `i32`', src/lib.rs:1:15
  |
  = note: this error originates in the macro `$crate::panic::panic_2021` which comes from the expansion of the macro `panic` (in Nightly builds, run with -Z macro-backtrace for more info)

error[[E0080]](https://doc.rust-lang.org/stable/error_codes/E0080.html): evaluation of constant value failed
 --> src/lib.rs:2:15
  |
2 | const _: () = panic!("column `foo` of query has SQL type `TEXT`");
  |               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the evaluated program panicked at 'column `foo` of query has SQL type `TEXT`', src/lib.rs:2:15
  |
  = note: this error originates in the macro `$crate::panic::panic_2021` which comes from the expansion of the macro `panic` (in Nightly builds, run with -Z macro-backtrace for more info)

For more information about this error, try `rustc --explain E0080`.
error: could not compile `playground` due to 2 previous errors

The formatting isn't ideal but at least all the relevant information is there.

[heksesang]

Is this feature supposed to work for Options?

[beyarkay]

For future travellers, these sqlx docs solved my problem. Basically:

#[derive(sqlx::Type)]
#[sqlx(transparent)]
struct UserId(i64);