redis-rs VS sqlx

staticpi, is a websocket forwarding service. Basically, it enables one to keep a Raspberry pi, or any computer, “connected”, in order to send and receive messages to and from any client, without having to deal with a static IP address, open ports on your router, or similar. Built in Rust, using axum, which in turn uses tungestine-rs for the websocket connections, tokio, sqlx, redis-rs and others.

redis (use streams and pubsub)

Are you addressing the long-standing issue that the redis client has related to dropped connections?

redis: by far the most popular and definitely the most ergonomic library. However, I was frustrated by its pubsub support; while generally acceptable, it's impossible to perform subscription operations while consuming messages. I like the library, but the workarounds to this issue were unacceptable to me.

I switched to LALRPOP for gluon but I still use combine in https://github.com/mitsuhiko/redis-rs and some other projects which need to parse "protocols" (less need for good error messages/error recovery and more need for speed).

The project is implemented with MongoDB Rust driver and redis-rs crate.

redis-rs, a Rust library for Redis with both high and low-level APIs

I think the redis crate was a well-organized API library. It's still sometimes hard to know the right things to make public and I think they nailed it.

Based on redis-rs, this library is a port of the Elixir Redix.Stream library.

sqlx

Thanks for sharing your experiences. I've been looking to get into snowflake for some time (even opened this https://github.com/launchbadge/sqlx/issues/986) . We thought to move to it years ago, but there's no way it was going to replace a significant amount of the mariadb we're running, and unless it did I wasn't looking forward to managing 2 DBs.

however, a much worse problem is its inability to do dynamic queries. If you have two queries that only differ in a column name, you're forced to copy paste the entire query. You can try and hack around it, but it's ugly either way. Here's a good overview, with no really good solutions.

Rust allows me to mainly only run the application to confirm things work from a business perspective.

For people starting out building stuff in rust - understand that there is a distinction of async code and libraries and can lead to confusing compiler errors if you don't realize there is a distinction. It's simple in hindsight but did cause me to waste hours barking up the wrong trees at first. Other wise just learn about `match` and Result/Option types asap, they're fundamental.

https://github.com/http-rs/tide tide is great to create an http server / routes

https://github.com/djc/askama I use this to template out HTML and it checks all my boxes, dynamic data, passing in functions, control flow.

https://github.com/launchbadge/sqlx sql interface for a variety of backend, async safe.

https://github.com/seanmonstar/reqwest http client to make requests

Rust is amazing, don't let the initial few speed bumps discourage you - building real things with rust is no more challenging today than any other modern language stack.

Perhaps Rust has some nice crates that will assist your development, but you're going to have to evaluate whether their design decisions meet the needs of your design, especially if you want to handle powerloss without data loss whilst maintaining high performance. Here's an example of where a create chose design over performance https://github.com/launchbadge/sqlx/issues/2007

SQLx's query macros act like format macros, and the database is apparently smart enough to not treat bind parameters as part of the SQL.

> Doesn't it need to run introspection queries to check the table structures?

Yes, but it can also run in 'offline' mode where it builds against saved metadata [0]

0: https://github.com/launchbadge/sqlx/blob/main/sqlx-cli/READM...