postgres-typed VS kysely-codegen

This is really cool, will look into using it in future projects!

I also made a tool (https://github.com/vramework/schemats) that generates the types directly from the db, which means whenever you do a DB migration your database types automatically update. Was forked from the original schemats library a couple years ago.

I also created a lightweight library ontop of pg that is less of a query builder and more of a typed CRUD + SQL for non trivial queries (https://github.com/vramework/postgres-typed). Most queries I deal with in a day to day is usually crud so I find it a little easier, but it's much less powerful then Kysely! I fall more into the camp of writing complex queries in SQL with small helpers and writing simple ones with util functions and typescript

I'm working on a few projects, from one/two days to platforms.

The first is OS and is a simple nodeJS environment to deploy applications via lambda and express quickly. Sort of like nestJS except less decorators and more functional (https://vramework.io/). I already know of a few other colleagues that rolled their own propriety versions of this to support enterprise and cloud deployments so decided to OS it.

The other OS project is a strongly typed postgres/mysql driver. The idea is to generate typescript definitions directly from postgres (https://github.com/vramework/schemats) and then have a think layer ontop of pg-node that gives you strongly typed queries (https://github.com/vramework/postgres-typed).

An open-source project I spent a few years on the core team is https://deepstream.io/, a realtime-server that allows you to mix and match multiple streaming protocols (mqtt/websocket/others) and allow those clients to talk to each other using pub-sub and records. I'm not longer working for it but wanted to give it a shout out!

On a non OS project, I have been working on an immersive audio platform for a while now. The main goal is to allow users to pick and choose how audio books progress, and also have a live session mode which allows users to record their pulse / answer questions and a few other metrics and associate it with sentences. I pretty much built and deployed all of it but require some advice/brainstorming on how to proceed now. I built it to satisfy an itch when I was practicing shamanism during the first lockdown when I was in-between contracts / taking time off.

I also want to build a simple web-pages strategy game based around eco-education, but don't have the bandwidth . If anyone is interested in mixing together gamification and eco-village building might be a fun conversion to bounce ideas!

All the OS projects above were used to support my personal/a couple professional projects over the last few years.

Email in profile

I agree with your point that adding multiple layers = more attack vectors and abstraction of a really good domain specific language. But what seems to happen on most of the projects I work on is we end up hiding away extremely common logic behind helper functions. It always starts off with SQL and then slowly gets moved into higher level functions that offer a better developer experience.

Shameless plug, but I just posted a library I wrote (for node https://github.com/vramework/postgres-typed/blob/master/READ...) which pretty much is a tiny layer ontop of pg-node (which is query based / with value parameters) and provides small util functions with typescript support derived straight from postgres tables.

In an ideal world (one I think we are getting very close to) I think we will end up having SQL queries validated in our code against the actual DB structure the same way we have any other compilation error. But until then we'll need to rely on tests or helper libraries, and for the purpose of refactoring and development I find the latter more enjoyable (although still far from perfect).

Hello HN! I was recently introduced to Kysely (https://github.com/kysely-org/kysely), a type-safe Typescript SQL query builder, and instantly fell in love! I got inspired and wanted to make something similar for AWS DynamoDB.

Thus, I developed Tsynamo! Instead of calling it type-safe, I decided to go with type-friendly, because the library is still in an early stage, and is not 100% type-safe.

Under the hood, Tsynamo compiles the built queries into AWS SDK v3 commands. I feel that Tsynamo simplifies the AWS SDK API quite a lot since the developer doesn't have to mess around with condition/filter expressions or attribute names/values themselves, and as a bonus gets autocompletion for building the queries!

There's also a playground to test it out in your browser: https://try.tsynamo.dev. It might not have the most up-to-date API of the library in use yet, but you can get the library's main idea from there.

Since the project is still in its early stages, it doesn't yet have 100% support for all DynamoDB features, like querying indexes. The next steps will be increasing the support coverage and perhaps adding automatic type generation as inspired by kysely-codegen (https://github.com/RobinBlomberg/kysely-codegen).

Would love to get some feedback, thanks in advance!

Querying the database (PostgreSQL) should not be ground breaking. Personally I like to have full type-safety so we can easily catch bugs during the development time without introducing any tests that are just testing the data type from our datastore to match the data type our API expects. I like to go database schema first, which means that we generate types from the database schema and work with those. Any change to the schema of the database is made with SQL migrations and after that, the typescript types are regenerated. Another approach is to use a code-first tool like TypeORM or Prisma. However in my experience such tools often produce not efficient SQL queries and are less easy to extend. In my projects I use library kysely (https://github.com/koskimas/kysely) with kysely-codegen (https://github.com/RobinBlomberg/kysely-codegen) to have a full type-safe SQL builder.

The main difference to me is that with Prisma I don't need to do any SQL to get a project up and running. I can define the tables and relations from my code or from Prisma Studio, and in one swoop I get the db structure done and the types defined.

I know my way around a database, but I'd rather not leave my code editor whenever I need to add a new column to a table.

With Kysely you have to create the DB schema, and then write the types; with every change you need, you gotta do both again.

(At least this seems true by default; as the project's readme mentions, there is a code generator[1] to generate the types from the DB schema; not quite the same but at least it's better than nothing.)

Commenters say that writing complex queries with Kysely is easier, which makes me wonder if I could use Prisma except for those, since Kysely should be able to just generate the SQL query for me to handle to Prisma...

[1]: https://github.com/RobinBlomberg/kysely-codegen

kysely kysely-codegen kysely-d1 allows access to D1 while using SQL Builder.

For what it's worth, I've been using kysely-codegen to generate the types based on the database schema and it's been working great for me.

I dislike ORMs and have been using Kysely https://github.com/koskimas/kysely and https://github.com/RobinBlomberg/kysely-codegen for a small project and so far I love it. It's a mapper/query builder that is equivalent to Dapper in C# (without the query builder) but with the power of typescript's type system to enforce proper queries. You can also bypass something with the sql to write raw SQL.