Things to improve upon

[funlennysub]

I always been using sea-orm as my go-to ORM for Rust but as every other project it has some cons and pros. Im making this discussion in hopes of helping toasty have less problems I personally deal with in sea-orm.

Things I like, but can be improved:

• As mentioned in this discussion, migrations as code are a good thing, but the way sea-orm does migrations is too verbose imo, I feel like using toasty's schema language can be useful for migrations, because you don't have to write hundreds of lines simply to create a few tables. In case users need something toasty doesn't yet support (create a postgres function for example), there should probably be a way to have both, schema-defined migration and a code one at the same time in one migration
  

Things I don't really like:

• The way it handles relations in select statements. Currently, if you want to get all posts made by a certain user, you have a few ways to do that: Find a user you want, then use its field used for relation to find all posts made by them; OR Use a JOIN statement. Both ways have their own drawbacks. If you use the first one, then you create multiple queries which might not be always acceptable, and toasty already does a better job with how you load related tables, then the second method might be better, which brings me to my second point:
• The way you have to handle a response that doesn't match any of your defined schemas. Again there's 2 ways to do that, a struct that implements FromQueryResult or an "unstructured tuple", in case of your own struct, you end up defining some fields multiple times which can lead to problems when table definition changes, as of writing this I don't have a solution that can solve this problem, but I hope people reading this might. For example, in the image below the only difference between, Query schema and that struct, is the query field, because in the table it is stored as i64 and linked to a table with all names
  

Things sea-orm doesn't have but I needed:

• A way to do COPY statements, for that I have to add another crate just to execute one query which is just pointless. I also have to convert sea-orm column types into tokio-postgres', I also have to make sure I didn't make a mistake in COPY statement since that crate doesn't know my schema, for that I have my own method that creates that string with all columns from that table

For now, this seems to be it, but if I remember something else, I will add it here. I think this discussion can be used as a general list of things that can be improved from other ORMs, so feel free to write them down here

[carllerche]

Thanks, this is useful. For context, there was a bunch of discussion related to this post here: https://discord.com/channels/500028886025895936/1300487772301168640.

[bsodmike]

OT: @funlennysub - Did you write your own driver for Postgres, just wondering and I couldn't find it in the examples.

[funlennysub]

I'm not sure what you're talking about