quaint VS prisma-engines

That is only used in tests :) The query engine uses this: https://github.com/prisma/quaint/blob/6532d69b5aec007ad06ac6...

(I work at Prisma, could have mentioned that earlier)

Just a quick note about this point: We do use the standard Rust drivers for the DBs we support (e.g. tokio-postgres or the official MongoDB driver for Rust) and don't write/maintain our own drivers, you can see this from our Cargo.toml.

I'm using https://github.com/prisma/quaint/ with mssql. As far as I understand, it wasn't released to crates.io for two year only because of some dependencies issues with postgres (https://github.com/prisma/quaint/issues/264)

This is a very strange comment section. And this article is insanely poorly written.

> Last week, we completed a migration that switched our underlying database from MongoDB to Postgres.

Okay cool, but why? MongoDB is a very capable and fast database.

> It was a shock finding out that Prisma needs almost a “db” engine layer of its own. Read more about it here: https://www.prisma.io/docs/concepts/components/prisma-engine...

If you did any research on Prisma rather than diving in head-first, you'd realize this is a core part of why Prisma exists.

> we discovered that at a low level, Prisma was fetching data from both tables and then combining the result in its “Rust” engine. This was a path for an absolute trash performance.

Can you confirm this is actually the case? Can you show some benchmarks re: this claim? Or are you just assuming this is the case?

If you wish to auto-generate migrations, there are declarative schema change tools available for most relational databases. I'm the creator of Skeema [1] which provides them for MySQL, but there are options for other DBs too [2][3][4].

Prisma's migration system actually partially copied Skeema's design, while giving credit in a rather odd fashion which really rubbed me the wrong way: "The workflow of working with temporary databases and introspecting it to determine differences between schemas seems to be pretty common, this is for example what skeema does." [5]

While I doubt I was the first person to ever use that technique, I absolutely didn't copy it from anywhere, and it was never "pretty common". I'm not aware of any other older schema change systems that work this way.

[1] https://www.skeema.io

[2] https://github.com/djrobstep/migra

[3] https://github.com/k0kubun/sqldef

[4] https://david.rothlis.net/declarative-schema-migration-for-s...

[5] https://github.com/prisma/prisma-engines/blob/6be410e/migrat...

If you're serious about your review then you shouldn't ignore the fact that Prisma has a big blob of Rust code at its core, where other ORMs use standard database adapters from NPM. As someone who has maintained database adapters for other languages, let me tell you that the maintenance burden of that is quite significant. Especially if they ever want to support more advanced database features. If the company behind Prisma ever runs out of money, the project is probably toast.

prisma generate generates the code for the Prisma client. The code generated for the client is all JavaScript which calls into the “Prisma Engine” Rust native Node module to perform database operations. As others here have said, the Prisma Engine is pre-compiled by rustc via CI and gets dowloaded to your machine as a pre-built binary by npm, so there’s no need for you to build it yourself by running the Rust compiler locally.

Note: you may notice that it downloads some binaries when you first invoke this command. This is normal it fetches the node prisma cli and engines used by prisma. 😁

We solved this with flat vectors and just sharing index values in cheap walker objects. It is much nicer to work with compared to arc/weak pointers.

Code here: https://github.com/prisma/prisma-engines/tree/main/libs%2Fda...

Because Prisma Python currently interfaces with the Rust engine over HTTP (I am looking into changing this) and the Rust engines can be found here:

https://github.com/prisma/prisma-engines

Thanks! I know of a couple Postgres tools that work in a declarative fashion: migra [1] and sqldef [2].

Migra is Postgres-specific. Its model is similar to Skeema's, in that the desired-state CREATEs are run in a temporary location and then introspected, to build an in-memory understanding of the desired state which can be diff'ed against the current actual state. (This approach was also borrowed by Prisma Migrate [3]). In this manner, the tool doesn't need a SQL parser, instead relying on the real DBMS to guarantee the CREATE is interpreted correctly with your exact DBMS version/flavor/settings.

In contrast, sqldef supports multiple databases, including Postgres and MySQL (among others). Unlike other tools, it uses a SQL parser-based approach to build its in-memory understanding of the desired state. As a DB professional, personally this approach scares me a bit, given the amount of nonstandard stuff in each DBMS's SQL dialect. But I'm inherently biased on this topic. And I will note sqldef's author is a core Ruby committer and JIT author, and is extremely skilled at parsers.

[1] https://databaseci.com/docs/migra

[2] https://github.com/k0kubun/sqldef

[3] https://github.com/prisma/prisma-engines/blob/main/migration...

Prisma 2 is a program, written in Rust that exposes a GraphQL API on top of your database of choice. Here's a link to the "engine": https://github.com/prisma/prisma-engines