Implementing HasTable for n-uples of tuples of columns from the same table

[LucaCappelletti94]

Backend

PostgreSQL, MySQL, SQLite, Others

What do you want to add to diesel?

I’d like to enforce multi-column constraints using a custom ForeignKey marker trait. For that to work, tuples of columns need HasTable impls — but the orphan rule prevents implementing them outside Diesel.

Example desired behavior:

impl<C1: Column, C2: Column<Table = C1::Table>> HasTable for (C1, C2) {
    type Table = C1::Table;
    fn table() -> Self::Table { C1::table() }
}

Would Diesel accept adding these macro-generated HasTable impls?

Happy to submit a PR.

Implementation notes

A macro could generate these impls by splitting the first column from the rest:

macro_rules! impl_has_table {
    ($first:ident $(, $rest:ident)* ) => {
        impl<$first: Column, $( $rest: Column<Table = $first::Table>, )*>
            HasTable for ($first, $($rest,)*)
        {
            type Table = $first::Table;
            fn table() -> Self::Table { $first::table() }
        }
    }
}

impl_has_table!(C1);
impl_has_table!(C1, C2);
// … up to for instance 12

Additional details

The aforementioned ForeignKey marker trait could reasonably be placed in diesel. Would that be desirable?

Checklist

• I have already looked over the issue tracker and the discussion forum for similar possible closed issues.

[weiznich]

That sounds like a reasonable addition. We should use the existing __diesel_for_each_tuple! macro to generate these impls, instead of having a custom macro like this to prevent having different maximal tuple sizes for different impls. By using the existing macro you get the default supported tuple size without any effort.

[LucaCappelletti94]

Ok - any comment on the desirability of having a ForeignKey marker trait in Rust?

[weiznich]

I would like to see a bit more details on how exactly you imaging such a trait first. So something like a rough draft of how the trait definition would look like and what use-case you have in mind

[LucaCappelletti94]

Ok, I will prepare another discussion post describing what my current impl looks like.

[LucaCappelletti94]

Closing as completed with PR #4874