I find writing sql in rust with sqlx to be far fewer lines of code than the same in Go. This server was ported from Go and the end result was ~40% fewer lines of code, less memory usage and stable cpu/memory usage over time.
It has the advantage that it implements the parsing and type checking logic in pure Go, allowing it to import your migrations and infer the schema for type checking. With SQLx you need to have your database engine running at compile time during the proc macro execution with the schema already available. This makes SQLx kind of a non-starter for me, though I understand why nobody wants to do what sqlc does (it involves a lot of duplication that essentially reimplements database features.) (Somewhat ironically it's less useful for sqlc to do this since it runs as code generation outside the normal compilation and thus even if it did need a live database connection to do the code generation it would be less of an impact... But it's still nice for simplicity.)
I ran sqlx / mysql on a 6M MAU Actix-Web website with 100kqps at peak with relatively complex transactions and queries. It was rock solid.
I'm currently using sqlx on the backend and on the desktop (Tauri with sqlite).
In my humble opinion, sqlx is the best, safest, most performant, and most Rustful way of writing SQL. The ORMs just aren't quite there.
I wish other Rust client libraries were as nice as sqlx. I consider sqlx to be one of Rust's essential crates.
Comparing and contrasting, sqlc type checking happens via code generation, basically the only option in Go since there's nothing remotely like proc macros. Even with code generation, sqlc doesn't default to requiring an actual running instance of the database, though you can use an actual database connection (presumably this is useful if you're doing something weird that sqlc's internal model doesn't support, but even using PostgreSQL-specific features I hadn't really ran into much of this.)
The sqlc authors are to be applauded for making a static analyzer, that is no small feat. But if you can get away with offloading SQL semantics to the same SQL implementation you plan to use, I think that's a steal. The usability hit is basically free - don't you want to connect to a dev database locally anyway to run end to end tests? It's great to eliminate type errors, but unless I'm missing something, neither SQLx nor sqlc will protect you from value errors (eg constraint violations).
2. Sure, the database will probably be running locally, when you're working on database stuff. However, the trouble here is that while I almost definitely will have a local database running somehow, it is not necessarily going to be accessible from where the compiler would normally run. It might be in a VM or a Docker container where the database port isn't actually directly accessible. Plus, the state of the database schema in that environment is not guaranteed to match the code.
If I'm going to have something pull my database schema to do some code generation I'd greatly prefer it to be set up in such a way that I can easily wrap it so I can hermetically set up a database and run migrations from scratch so it's going to always match the code. It's not obvious what kinds of issues could be caused by a mismatch other than compilation errors, but personally I would prefer if it just wasn't possible.
I would definitely recommend writing a Compose file that applies your migrations to a fresh RDBMS and allows you to connect from the host device, regardless of what libraries you're using. Applying your migrations will vary by what tools you use, but the port forwarding is 2 simple lines. (Note that SQLx has a migration facility, but it's quite bare bones.)
Type inference was okay, since SQLite barely has any types. The bigger issue I had was dealing with migration files. The nice part about SQLx is that `cargo sqlx database setup` will run all necessary migrations, and no special tooling is necessary to manage migration files. sqlc, on the other hand, hard codes support for specific Go migration tools; each of the supported tools were either too opinionated for my use case or seemed unmaintained. SQLx has built-in tooling for migrations; it requires zero extra dependencies and satisfies my needs. Additionally, inferring types inside the actual database has its benefits: (1) no situations where subsets of valid query syntax are rejected, and (2) the DB may be used for actual schema validation.
For an example of why (2) may be better than sqlc's approach: databases like SQLite sometimes allow NULL primary keys; this gets reflected in SQLx when it validates inferred types against actual database schemas. When I last used sqlc, this potential footgun was never represented in the generated types. In SQLx, this footgun is documented in the type system whenever it can detect that SQLite allows silly things (like NULL primary keys when the PK satisfies certain conditions).
That being said, as I understand it, SQLx does something very different. If you want dynamic queries, you'll basically have to build that module yourself. The power of SQLC is that anyone who can write SQL can work on the CRUD part of your Go backend, even if they don't know Go. Hell, we've even had some success with business domain experts who added CRUD functionality by using LLM's to generate SQL. (We do have a lot of safeguards around that, to make it less crazy than it sounds).
If you want fancy Linq, grapQL, Odata or even a lot of REST frameworks, you're not getting any of that with SQLC though, but that's typically not what you'd want from a Go backend in my experience. Might as well build it with C# or Java then.
Let's compare: SQLC - configuration file (yaml/json) - schema files - query files - understand the meta language in query file comments to generate code you want
SQLx - env: DATABASE_URL
Now does that mean that SQLx is the best possible database framework. No, it does not. Because I didn't spend my time doing things that weren't related to the exact queries I had to write I got more work done.
I want to appreciate the hard work the SQLx Devs have put in to push the bar for a decent SQL developer experience. People give them a really hard time for certain design decisions, pending features and bugs. I've seen multiple comments calling it's compile time query validation "gimmicky" and that's not nice at all. You can go to any other language and you won't find another framework that is as easy to get started with.
And of course now that I have it, the incremental cost of adding a new query is really low as well
You could compare it to people writing CSS, JavaScript and Markup in separate files Vs having just one file in React/Svelte etc. which gives the user the option to combine everything into one.
There maybe a lot of drawbacks from the latter approach but it's makes everything a hell easier for people to just get started building.
As far as building something fast, I'm with you. I always reach out for Python with UV, Litestar and Advanced Alchemy when I want to build personal web projects. I don't think SQLC is bad as such, once you've written your SQL you can essentially compile that into a CRUD application which is ready to go. As you've pointed out, however, you'd need to slam something like a GraphQL engine on top of it if you wanted rich quries easily, and you'd still not have the auto-generated OpenAPI that comes with Python web frameworks.
SQLC is for code where you want a low amount (or zero) external depedencies. Which is a very "Go" thing to want. It does scale well, but that requires you to build various CLI tools to help maintain things as well as your own Go modules to add "quality of life" like dynamic routers and get queries for low traffic requests.
I'll try SQLx eventually when I get time to look more into Rust.
I would recommend using pg_dump for your schema file which means it'll not be related to SQLC as such. This way it will be easier for you to maintain your DB, we use Goose as an example. In our setup part of the pipeline is that you write your Goose migration, and then there is an automated process which will update the DB running in your local dev DB container, do a pg_dump from that and then our dev container instance of SQLC will compile your schema for you.
The configuration file is centralized as well, so you don't have to worry about it.
I agree with you on the SQLC meta language on queries, I appreciate that it's there but we tend to avoid using it. I personally still consider the meta language a beter way of doing things than in-code SQL queries. This is a philosophical sort of thing of course, and I respect that not everyone agres with me on this. It's hard for me to comment on SQLx, however, as I haven't really used it.
What I like about SQLC is that it can be completely de-coupled from your Go code.
I referred go-jet since it introspects the database for it's code generation instead.
I've been quite happy with this setup!
FWIW, the compile-time query checking is entirely optional. If you don't use the query syntax checking then you don't need live database and you don't need `sqlx prepare`.
Can you tell me why it's a non-starter for you?
For sqlc, it isn't really a big problem because you only need to run the code generation when you're actually modifying database things. Still, with that having been said, I think just passing a database URI and having analysis work based on that is unideal. Using an actual database isn't a huge problem, but having to manage the database instance out of band is the part that I think isn't great, because it allows for the schema in the code to trivially desync with the schema used in analysis. If I used SQLx I'd probably be compelled to try to wire up a solution that spawns the database and migrates it up hermetically for the caching part. Likewise if I used this mode of sqlc.
I guess it might be possible for sqlc to add first class support for that sort of concept, but I can see holes in it. For one thing, you have to figure out where to grab binaries from and what version. An approach using Docker/Podman works, and at least partly solves this problem because you could allow specifying any OCI image, but that has caveats too, like requiring Docker or Podman to be installed. The most heroic effort would be to use some kind of solution using WASM builds of database engines: pulling down and running something like PGlite in process seems like it would be an almost ideal solution, but it sticks you to whatever things can actually be made to work in WASM in terms of features, extensions and versions, at least unless/until database servers and extension vendors miraculously decide that supporting WASM as a target is a good idea. Still, if you want some crazy ideas for how to make the UX better, I think either the Docker approach or the WASM approach could be made to work to some degree.
Barring that, though, I'd be most likely to have some kind of Docker setup for running sqlc with an ephemeral database instance. It's not pretty, but it works...
I don't think it would be a non-starter, though. I only really think that connecting to the database from within rustc invocations is a non-starter.
The more serious LoC offenders in Go were:
1. Marshalling/Unmarshalling code (for API responses, to/from external services, etc). In general, working with JSON in Go was painful and error prone. Rust's serde made this a complete non-issue.
2. Repetitive sql query code (query, scan for results, custom driver code for jsonb column marshalling/unmarshalling). Rust's sqlx made this a non-issue.
3. Go's use of context to share data through handlers was a real pain and error prone (type casting, nil checks, etc). Rust's actix-web made this a real beautiful thing to work with. Need a "User" in your handler? Just put it as an argument to the handler and it's only called if it's available. Need a db connection? Just put it as an argument to the handler.
4. Go's HTML/Text templates required more data to be passed in and also required more safety checks. Rust's askama was overall more pleasant to use and provided more confidence when changing templates. In Rust, I'd catch errors at compile time. In Go, I'd catch them at runtime (or, a user would).
I must admit I was surprised. I thought Rust would have been more lines of code because it's a lower level language, but it ended up being ~40% less code. My general sentiment around working with the code is very different as well.
In the Rust codebase I have no hesitation to change things. I am confident the compiler will tell me when I'm breaking something. I never had that confidence in Go.