gnorm VS sqlc

Furthermore, there can be a lot of boilerplate queries we do that it's nice to not have to write, say, the same kind of delete query over and over. In the past I've used gnorm as one way of generating all that boilerplate code based on the actual database design, and it works reasonably well, but again it plays a similar role to an ORM.

I've used gnorm for that in the past for some code generation, and I had absolute control. Gnorm took care of the database inspection side of things, and I created the templates it used to generate the code. I had full control over generated models and code.

I'm a big fan of the database first code generator approach to talking to an SQL database, so much so that I wrote pggen[1] (not to be confused with pggen[2], as far as I can tell a sqlc fork, which I just recently learned about).

I'm a really big partisan of this approach, but I think I'd like to play the devil's advocate here and lay out some of the weaknesses of both a database first approach in general and sqlc in particular.

All database first approaches struggle with SQL metaprogramming when compared with a query builder library or an ORM. For the most part, this isn't an issue. Just writing SQL and using parameters correctly can get you very far, but there are a few times when you really need it. In particular, faceted search and pagination are both most naturally expressed via runtime metaprogramming of the SQL queries that you want to execute.

Another drawback is poor support from the database for this kind of approach. I only really know how postgres does here, and I'm not sure how well other databases expose their queries. When writing one of these tools you have to resort to tricks like creating temporary views in order infer the argument and return types of a query. This is mostly opaque to the user, but results in weird stuff bubbling up to the API like the tool not being able to infer nullability of arguments and return values well and not being able to support stuff like RETURNING in statements. sqlc is pretty brilliant because it works around this by reimplementing the whole parser and type checker for postgres in go, which is awesome, but also a lot of work to maintain and potentially subtlety wrong.

A minor drawback is that you have to retrain your users to write `x = ANY($1)` instead of `x IN ?`. Most ORMs and query builders seem to lean on their metaprogramming abilities to auto-convert array arguments in the host language into tuples. This is terrible and makes it really annoying when you want to actually pass an array into a query with an ORM/query builder, but it's the convention that everyone is used to.

There are some other issues that most of these tools seem to get wrong, but are not impossible in principle to deal with for a database first code generator. The biggest one is correct handling of migrations. Most of these tools, sqlc included, spit out the straight line "obvious" go code that most people would write to scan some data out of a db. They make a struct, then pass each of the field into Scan by reference to get filled in. This works great until you have a query like `SELECT * FROM foos WHERE field = $1` and then run `ALTER TABLE foos ADD COLUMN new_field text`. Now the deployed server is broken and you need to redeploy really fast as soon as you've run migrations. opendoor/pggen handles this, but I'm not aware of other database first code generators that do (though I could definitely have missed one).

Also the article is missing a few more tools in this space. https://github.com/xo/xo. https://github.com/gnormal/gnorm.

[1]: https://github.com/opendoor/pggen

Hi HN, I’m Kyle and together with Andrew (https://news.ycombinator.com/user?id=stanleydrew) we’ve been working on Riza (https://riza.io), a project to make WASM sandboxing more approachable. We’re excited to share a developer preview of our code interpreter API with HN.

There’s a bit of a backstory here. A few months ago, an old coworker reached out asking how to execute untrusted code generated by an LLM. Based on our experience building a plugin system for sqlc (https://sqlc.dev), we thought a sandboxed WASM runtime would be a good fit. A bit of hacking later, we got everything wired up to solve his issue. Now the API is ready for other developers to try out.

The Riza Code Interpreter API is an HTTP interface to various dynamic language interpreters, each running inside a WASM sandbox without access to the outside world (for now). We modeled the API to align with a POSIX shell-style interface.

We made a playground so you can try it out without signing up: https://riza.io

The API documentation lives here: https://docs.riza.io

There are many limitations at the moment, but we expect to rapidly expand capabilities so that programs can e.g. access the network and filesystem. Our roadmap has more details: https://docs.riza.io/reference/roadmap

If you need to execute LLM-generated code we’d love to have you try the API and let us know if you run into any issues. You can email us directly at [email protected].

"Is there a way to get sqlc to use pointers for nullable columns instead of the sql.Null types?"

I came across this yesterday for golang: https://sqlc.dev which is somewhat like what you want, maybe.

Not sure it allows you to parameterize table names but the basic idea is codegen from sql queries so you are working with go code (autocompletion etc).

Agreed, but tools like https://sqlc.dev, which I mention in the article, are a good trade-off that allows you to have verified, testable, SQL in your code.

sqlc is not strictly a conventional ORM. It offers a unique approach by generating Go code from SQL queries. This allows developers to write SQL, which sqlc then converts into type-safe Go code, reducing the boilerplate significantly. It ensures that your queries are syntactically correct and type-safe. sqlc is ideal for those who prefer writing SQL and are looking for an efficient way to integrate it into a Go application.

I was browsing awesome-go for ideas on how to setup my data access layer when I stumbled on sqlc. It seemed like a great option. Code generation is a strategy often used in the Go ecosystem and making my queries safe at compile time was an idea I really liked. Knex was great, but it required of me that I test thoroughly my queries at runtime and that I sanitize my query results to ensure type safety within my application.

Now, we are going to generate the code. For this purpose, we are going to use sqlc.

https://github.com/sqlc-dev/sqlc — for use with //go:generate