SURREALIGRATE Posts
I've been spending my spare time working on Node.js tools for SurrealDB. The 2 that are showing some promise are named Surrealigrate (a migration tool) and Bole ORM. I may end up changing their names, but for now that's what I'm calling them.
I've talked about both of them, so I'll try not to repeat myself. I've worked on expanding Surrealigrate some, but I need to work on some more advanced schemas and namespace setups to really push it. This week I've mostly focused on Bole ORM. The word "bole" is a kind of clay or the trunk of a tree, which both are kinda fitting for an ORM, in my opinion. My approach with all of my SurrealDB tools has been to take advantage of the built-in info queries and generate database specific code. A lot of tools, especially ORMs, are very standardized and strict with their implementation. I don't want that. I want to be able to generate query tools that are specific to the database and then be able to adapt them however I want, without the ORM getting in my way.
Bole ORM is a simple CLI tool and it generates a query class built for that specific database, down to table names and specific fields' validation. It doesn't require any kind of schema files or anything that manually needs to be done by the developer, it's all generated by inspecting the database. While it currently generates a class, to be used as object-oriented, I also plan to make it configurable, so if you prefer chaining, it'll do chaining, instead of objects. It resembles Prisma in the way the queries look, when using the class output, but it's simplier and I'm pretty sure the where and fields (Prisma's select object) objects may support more options than most Prisma methods support.
All the Flavors uses Prisma, so I'm very familiar with it and it's actually powerful when combined with a GraphQL parser. That is why I'm mimicking Prisma, to some extent, but also adapting with some things I feel Prisma could have done better. Once the feature set is, well, set, then I'll go into more details for the actual capabilities...but there's also a catch. I plan to make the generator modular, so you can plugin whatever you want, leave out whatever you don't, or add custom modules. This means you can basically design your own ORM with modules and just let Bole generate it.
I want to add some additional utilities to each. For Surrealigrate, I want it to be able to identify potential missing indexes and suggest changes. I also want it to be able to use either schema within SurrealDB or from a library, such as Zod, to generate migrations. The other side of that, I want Surrealigrate to be able to generate a Zod schema and validations for use outside of Surrealigrate, in Bole ORM or just in your application in general. As far as Bole ORM additions, I want to generate documentation for the specific database, and Bole ORM itself, and I want to do automated tests, primarily on data integrity, but also for the application itself.
The idea behind these tools is to simplify development and have something you use when you need them, but they get out of your way when you don't. There's no point in making someone run a bunch of CLI commands, when you can use Surrealist to design your database, then just run a simple command that does everything else. That's the way I see it anyway. If extra commands are involved, they are optional and they add value, like a command to generate a Bole query that you can paste into your application. That could just be a plugin though, and then it just updates a generated export and you never have to touch it again. Working for you, not against you.
My interest in SurrealDB has really grown the past few days. I really feel like it can replace ArangoDB in my stack. The only issue is the same issue I had when I adopted ArangoDB: there aren't a lot of tools for Node.js. Like ArangoDB, I'll maybe have to build some myself. The first such tool is a CLI migration tool.
First of all, the Surrealist database manager application is awesome. It's great, it really is and, while ArangoDB has it's web client, it has functionality I missed using ArangoDB that I used to have for relational databases. That is a great tool, but it's more of a development and management tool. I need a migration tool for automation and to make my applications more portable.
Today I made a big push toward creating this tool and it's named Surrealigrate. I've built tools like it for other databases, such as the one we use for All The Flavors, Arangrate for ArangoDB, and I had initially built a multi-database tool for the CMS behind this site (now it uses Prisma, well, for now). Surrealgrate isn't quite ready, but like I said, it's getting very close. It's the first thing I've built from the ground up with AI assistance, which has sped up development considerably. Unfortunately, AI's aren't perfect, so there were a few issues and I had to rewrite portions of it, but most of the changes were just oversights that occurred as I asked the AI to add specific features in steps.
I need to build Surrealigrate up fairly quickly, so I can move some of my current projects to it and get them moving again. While it is already more advanced than my other tools have been, there will be more features coming later. Using AI allowed me to put in a lot of things initially that I wouldn't have had time to do right now. One different approach I'm taking with this tool is it's not intended to be an NPM package. It's intended to be included in a project as a component, potentially using git submodules. I'm hoping this approach makes it more useful per-project and will lead to others contributing.