Multiple database adapters in a single workspace

[esmevane]

Hey there!

I've been attempting to set up CI for a workspace project that has the following shape:

• A core set of generic entities in one crate (where the program's persistence model lives)
• A postgres persistence query layer in another crate (for a HA mode for the program)
• A sqlite persistence query layer in another crate (for an ephemeral / local mode for the program)
• A final crate that lets you pick adapters (where logic is implemented)

Now, all of this is actually working, with some workarounds locally. The problem is a few things, which mostly make it tough to set up on new machines (or re-setup locally), and make CI very tough to reason about.

I have a repro repo here

Basically the issues I'm experiencing are these:

1. cargo sqlx prepare --workspace seems to assume that I'm using one database adapter; it wants to know what the database url is for the whole root, and it more or less ignores individual crates if that is set, as well as SQL_OFFLINE=true later.
2. The cargo sqlx db create ignores the workspace entirely, and so I have to describe the sqlite connection string in such a way that it works no matter the path origin. In other words, I can't set a relative path in the crate and have that work from the root and the crate. (Also, it tends to ignore sqlite:memory: entirely for maybe the same reason?)

I know this isn't on the happy path, and there might be some design compromises I should make. That's understandable, I just would love to make sure I'm not missing anything obvious. Is there anything I can do to make multiple database kinds in a workspace a little more maintainable?

[esmevane]

Oh I've almost forgotten. Here's the Github action file which isn't working for me:

name: Test Packages

on:
  push:
    branches:
      - main
  pull_request:

jobs:
  test:
    runs-on: ubuntu-latest

    steps:
      - name: Check out code
        uses: actions/checkout@v3

      - name: Set up Rust
        uses: actions-rs/toolchain@v1
        with:
          toolchain: stable
          override: true

      - name: Test workspace
        run: |
          SQLX_OFFLINE_DIR=$PWD/.sqlx SQLX_OFFLINE=true DATABASE_URL=postgres://localhost/repro-pg cargo test

I've tried these variations of the command and they all fail:

• SQLX_OFFLINE_DIR=$PWD/.sqlx SQLX_OFFLINE=true DATABASE_URL=postgres://localhost/repro-pg cargo test
• SQLX_OFFLINE_DIR=$PWD/.sqlx SQLX_OFFLINE=true cargo test
• SQLX_OFFLINE=true DATABASE_URL=postgres://localhost/repro-pg cargo test
• cargo test

Each with a bunch of different combinations of .sqlx directories: generated per crate, finagled into generating at the top level, etc.