tvallotton / models

A tool for automated migrations for PostgreSQL, SQLite and MySQL.

Hello folks, this is my github account. I'm a narcissist so here are my user profile stats:

models's Issues

Getters for foreign keys and primary keys

Add support for getters for forein keys, with the following API:

#[derive(Model)]
struct Post {
    #[primary_key]
    id: i32, 
    #[foreign_key(User.id, on_delete="cascade")]
    author: i32
}

let post: Post =  Post::from_key(&mut conn, &id).await?;

let user: User = post.author(&mut conn).await?;

`cargo install sqlx-models-cli` conflicts with `sqlx-cli`

Maybe add an explanation that it conflicts with sqlx-cli in the readme :)

Remove pancis from Migrations

Panics will poison the mutexes used to sync the tables.

add a Json<T> and Blob<T> type for serializable objects

create a Json type can hold any type T that implements Serialize and Deserialize. Json probably using serde_json and Blob or Bytes using bincode. they should also implement Deref and DerefMut to T.

Error when using `sqlx migrate generate` on existing project

I already had a code base with handwritten (up and down) migrations. I changed my Rust struct and annotated it properly. For some reason I get the following error:

$ sqlx migrate generate
    Finished dev [unoptimized + debuginfo] target(s) in 16.92s
thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Error("expected value", line: 1, column: 33)', /home/user/.cargo/registry/src/github.com-1ecc6299db9ec823/sqlx-models-cli-0.1.1/src/generate.rs:74:43
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

I tried to reproduce it inside a clean environment but that failed. In a new, clean project everything works fine. Sadly I can't link you the code.

new-name: Bug with constraints and moves.

There is some buggy behavior in new-name related to constraints in temporary moves. Some constraints are missed.

False positive dependecy cycles

False positive errors are triggered when checking for dependency cycles if one of the dependencies is not a model. This could be fixed either by adjusting the error message of by changing the mechanism to allow this kind of behavior. This kind of permissiveness could be desirable when the structure belongs to another crate. Example:

struct A {
   id: i32, 
}

#[derive(Models)]
struct B {
   #[foreign_key(A.id)]
   a: i32
}

How to combine this with rocket_auth

Is there an example of how to use this together with rocket_auth?

In the rocket_auth example, I see that the Users table is being created, but I want that to be part of a migration:
https://github.com/tvallotton/rocket_auth/blob/4046459ffec839d6f2cf8884c30a6d8cc4f8f52c/examples/sqlite.rs#L65

Moves with constraints cause conflicts

constraint names are global, so they should be dropped before a move in PostgreSQL and MySQL, as they otherwise conflict with constraints on the tempopary table.

Improve CLI generation interface

Ideally, show the compilation output, but remove the testing output, and replace it with a custom log.

Default literals should be compatible with all dialects.

Currently default SQLite boolean defaults require to use a 0 and 1, whereas PostgreSQL and MySQL use boolean literals. It would be better if true and false were replaced with 0 and 1 when using SQLite automatically.

new-name: Default strings are turned to lowercase

Remove unnecessary dependencies

Specifically, remove itertools, fehler, and thiserror to speed up compile times. And make chrono optional.

Include binary literals on default constraints

So far only bools, floats, integers and strings. It should also include bytestrings.

Postgresql serial type

in postgresql there is a handy serial type would be very convenient to mark some fields (e.g. id) to have this type instead of the regular i32 -> integer conversion or in general more power to choose types. Is there any other way to have auto inc ids?

Supported Databases

What databases are currently supported? I was about to play around with it and was wondering if it currently is even possible to create migrations other than sqlite that are supported by sqlx right now. I see some occurrences of postgresql in the code and dialect sections but i don't really now how to navigate the code yet to answer that myself. It is unfortunately not mentioned in the readme etc. but some places in the code made it look like it can determine the database from the connection string DATABASE_URL=sqlite:// ... DATABASE_URL=postgres:// but it looks like it is only generating sqlite compatible migrations atm.

CLI requires --database-url parameter

Make it infer it from .env file.

Generate both `up` and `down` variants.

Right now, the CLI does only generate the up variant. This could be achieved by using the same mechanism as sqlx: the -r flag.

add a `modify(&self, conn)` to update the values of a row

Maybe using a trait.

Big serial compatibility

Any plans about this type support ?
Big serial should be useful to represent i64

diesel compatibility

could be awesome to be able to use this crate with diesel.
The output files should be under
./migrations/migrationId/up.sql
./migrations/migrationId/down.sql