Big backend change coming soon!

[therealbrad]

For a few weeks I've been working on migration from zenstack v2 to v3. It's a large change with a lot of benefits. Mostly, they've moved from prisma to kysley which should provide faster DB performance and quicker build times without sacrificing the nice TypeScript dev experience. It means no more work-arounds for complex queries that prisma doesn't handle well. But it also means we might be tied to PostgreSQL longer than I wanted to. My goal has always been allowing everyone to use the tools they like and I would love to be DB agnostic, but it's difficult with such complex queries. So the plan is PostgreSQL support only for a while more.

I want to hold off on new features until this migration is complete and tested to the best of my ability.

https://zenstack.dev/docs/migrate-v2

[therealbrad]

I encountered a couple of zenstack v3 bugs that are preventing me from migrating from v2.

• findMany with distinct on one field and orderBy on a different field generates invalid PostgreSQL SQL. zenstackhq/zenstack#2529
• PolicyPlugin generates invalid PostgreSQL SQL when access rules contain nested collection predicates zenstackhq/zenstack#2538

I am holding off on migration until those issues are fixed.

[therealbrad]

I have fully upgraded zenstack from v2 to v3! This is a huge rewrite of the data layer so I have the changes in https://github.com/TestPlanIt/testplanit/tree/beta . I want to keep putting it through its paces for a few weeks before merging to main, but it means holding off on some other work until this is merged to avoid major conflicts and/or rewrites of features touching the data layer.

I am seeing much faster query execution and ACL handling which opened the door for other improvements like infinite scrolling with live column resizing on many tables I wasn't able to use it on before.

Postgresql default container updated to v18 which gives even more performance enhancement.

[therealbrad]

I am fully confident in the zenstack v3 updates and they're running incredibly well on enormous datasets. All changes are on beta if you want to see what's to come in the 1.0 release scheduled before the end of September 2026.