[Haskell-cafe] Selda: confused about type signtures

[Tom Ellis]

On Sun, Apr 22, 2018 at 08:21:11PM +0200, Marc Busqué wrote:
> If I add `FlexibleContexts` and `AllowAmbiguosTypes` extensions, I can
> compile the program and ask for the type of `list`:
> 
> ```
> :t list --- list
> ---   :: (selda-0.1.12.1:Database.Selda.Compile.Result
> ---         (selda-0.1.12.1:Database.Selda.Column.Cols s a),
> ---       selda-0.1.12.1:Database.Selda.Column.Columns
> ---         (selda-0.1.12.1:Database.Selda.Column.Cols s a)) =>
> ---      selda-0.1.12.1:Database.Selda.Table.Table a
> ---      -> IO
> ---           [selda-0.1.12.1:Database.Selda.Compile.Res
> ---              (selda-0.1.12.1:Database.Selda.Column.Cols s a)]
> ```
> 
> However, manually adding this same type:
> 
> ```
> list :: (Result (Cols s a), Columns (Cols s a))  => Table a -> IO [Res (Cols s a)]
> list table = withDB $ query (select table)
> ```
> 
> results in a compilation error:

Unless I am much mistaken `Cols s a` is just `a`[1,2].  That explains why
the type is ambiguous.  There's nothing that fixes `s`.

> But I'm not sure about how I could fix that. At the end of the article
> it is said that this only happen in programs that "they could never
> really work". But, if I'm not missing something, that function does work.
> I have tried it with the auto-inferred signature way and it indeed can
> list rows for different database tables.

Can you show us a minimal example of `list` working?  I'm quite confused
about how it could work.

Tom


[1] https://www.stackage.org/haddock/lts-9.14/selda-0.1.11.1/src/Database.Selda.Compile.html#line-126)

[2] https://www.stackage.org/haddock/lts-9.14/selda-0.1.11.1/Database-Selda.html#t:Cols