Welcome to sqg Discussions!

[uwemaurer]

🎤 Help us shape the future of SQG!

Hey everyone!

We built SQG to bridge the gap between raw SQL power and type-safe code without the overhead of a heavy ORM. Now that the project is growing, we want to make sure we're building the features you actually need.

To help us prioritize, we’d love to hear about your experience (good or bad!):

🛠 Your Setup

• What database are you using? (PostgreSQL, SQLite, DuckDB, or something else?)
• Which language are you generating for? (TypeScript, Java, or are you waiting for another?)
• How are you using SQG? (Personal project, production app, data analysis?)

💬 Your Feedback

• What’s working well? Does the introspection feel accurate?
• What’s missing? Is there a specific SQL feature or generator option you're dying to have?
• Any "papercuts"? Small things that are annoying or documentation that isn't clear?

Don’t hold back. Whether it's a bug report, a "this saved my life," or a "why doesn't it do X?", every bit of feedback helps us improve the tool.

Drop a comment below—we’re excited to see what you're building! 🚀

[linux-china]

What database are you using? DuckDB, PostgreSQL, SQLite, MySQL
Which language are you generating? TypeScript with Bun, Java with JBang, Python with uv
How are you using SQG? Personal project, Ops and data analysis

[uwemaurer]

Thank you! If you run into any issues or have suggestions, please let me know

[daniel-nagy]

Hi, thanks for creating this project!

I'm currently evaluating code generation tools for an existing Java project. I really like the idea of writing raw SQL queries and generating type-safe functions for executing those queries.

We do have the unfortunate requirement of supporting multiple database vendors, including H2 (for local development), Postgres, Microsoft SQL server, and Oracle. Given SQG's design, how feasible do you think it would be add support for H2, MSSQL sever, and Oracle?

I can see that SQG is similar to SQLDelight and sqlc (with the Java generator). What would you say sets SQG apart from those other tools?

For our particular app, we currently use Flyway to manage migrations, and I have no intention of refactoring this to use a specific code generation tool. I am successfully able to generate types from a Postgres database using Testcontainers and setting the SQG_POSTGRES_URL, though figuring this out required looking at the source code of SQG.

Some other feedback:

• I would really like to see support for generating enum types, and automatic conversion between Java enums and SQL enums.
• I don't like that I am force to use @set foo = 'bar' when defining my queries, the database client I use does not support this syntax and I don't see the value in requiring these variables.

[uwemaurer]

@daniel-nagy I implemented code generation for Java enum from postgres enums. It will be used in the results from the DB and also in the parameters

[daniel-nagy]

Hi @uwemaurer, thanks for adding enum support! I just tested it and it is working!

I have some more feedback for you.

• On Mac OS I have to use npm install --force or else I will get this error about duckdb on npm install

  npm error notsup Unsupported platform for @duckdb/node-bindings-linux-x64@1.4.3-r.2: wanted {"os":"linux","cpu":"x64"} (current: {"os":"darwin","cpu":"arm64"})

• The built-in type mapping is pretty good, but is there any way to add mappings for custom types? For example, if I wanted to map Postgres uuid to MyUuid type in Java.

I wonder which database client you use. I am thinking about dropping this @set foo = 'bar' and define my own format. It was to stay compatible with DBeaver so I can actually run these queries in DBeaver.

I currently use Beekeeper Studio. Beekeeper Studio does not have a a special syntax for query parameters

The ability to define and use variables within the SQL script itself (e.g., using SET @var = 'value' in MySQL or DECLARE in SQL Server) is a feature of the specific database's SQL dialect, not a universal Beekeeper Studio feature.

To use variables in Postgres I would use $1, $2, etc. For example,

INSERT INTO hockeyPlayer(player_number, full_name)
VALUES ($1, $2);

If I ran this query in Beekeeper it will prompt me to enter values for these variables

[uwemaurer]

Thank you for your feedback! I am planning to work on type mapping next, to support own types and also to share types between queries (if they return the same fields)

[uwemaurer]

Thank you for the feedback @daniel-nagy !

• H2 will be a bit difficult to support since this is a Java library and SQG is implemented in typescript

• I don't have Microsoft and Oracle SQL, but probably it will be quite similar to the postgres integration.

• I didn't try SQLDelight, from the website I think it is for JVM only. but looks really good if you need it only in Java/Kotlin

• I tried sqlc, a great project and inspiration for SQG. It could not do DuckDb yet and I wanted the query definition file to be compatible with DBeaver, this is why I continued with my own project.

• great suggestion about the enums. They are currently supported in the queries, but it maps them to String in Java (also UUID to String)

• I wonder which database client you use. I am thinking about dropping this @set foo = 'bar' and define my own format. It was to stay compatible with DBeaver so I can actually run these queries in DBeaver.

• I am currently working on a simple SQL IDE integrated into this project, which can load the sqg.yaml and allows to run the queries. so then I don't need to stay compatible with DBeaver. ( I could always let SQG produce a DBeaver compatible file if needed)