paulocsanz / arraystring Goto Github PK

Hello,
I'm trying to use ArrayString with diesel and it fails to derive AsChangeset.

table! {
    ...
    users (id) {
        ...
        password_hash -> Varchar,
        ...
    }
}

type PasswordHash = ArrayString<U64>;

#[derive(AsChangeset, Default, Debug)]
#[table_name = "users"]
pub struct UserSettingsChangeset {
    pub password_hash: Option<PasswordHash>,
    ...
}

will lead to:

type mismatch resolving

• <&arraystring::ArrayString<arraystring::typenum::UInt<arraystring::typenum::UInt<arraystring::typenum::UInt<arraystring::typenum::UInt<arraystring::typenum::UInt<arraystring::typenum::UInt<arraystring::typenum::UTerm, arraystring::typenum::B1>, arraystring::typenum::B1>, arraystring::typenum::B0>, arraystring::typenum::B0>, arraystring::typenum::B1>, arraystring::typenum::B0>> as diesel::Expression>::SqlType == diesel::sql_types::Nullable<diesel::sql_types::Text>
• expected struct diesel::sql_types::Text, found struct diesel::sql_types::Nullable
• expected struct diesel::sql_types::Text
  found struct diesel::sql_types::Nullable<diesel::sql_types::Text>

It works with normal String...Option<T> in AsChangeset should just skip the field during update.

Any ideas what I'm doing wrong? Thank you

lots of unchecked functions check things anyway, it makes no sense to have an unsafe version... Can i remove them while i do a refactoring for #5 ?

I just merged my master with yours, since i didn't touch the master and the tests passed when i pushed my const N PR i can only wonder what happened...
Have you run the tests lately, did they succeed ?

This is a personal reminder, but if anyone wants to do it first feel free.

• 100% safe panic free truncate with split_at (if the optimizer would be so kind as to know that we already checked if mid > len)
• faster copy with small types creating a custom copy loop: it looks like the use of ptrs in the std function forces the ptrs to actually exist, where an inlined loop would probably be optimized into register operations (please benchmark exhaustively the variato)
• base as many operations as possible with a variation of replace_range if it is properly able to optimize it into the derived ones (remove, insert, push, etc...) (Please see if assembly is identical, or at least performance identical)

there is a lot of redundant code in the CacheString

we could use delegate to do some of that, or other means.

I wonder whether it'd be possible to substitute the typenum crate with the subset of const generics stabilized since Rust 1.51.