Building a "Rails-like" framework in Go is honestly totally antithetical to the "Go way" of doing things.

Rails has a ton of magic, implicit behaviours, monkey patching, ERB, and so on.

Go is a touch below Java verbose, explicit, no magic, every function call can be very easily traced without having to do meta programming and code generation in your head to understand what's going on.

In my 8 years of writing Go, I would liken it the most to C, where you had to spell everything out, except without the manual memory management and the macro preprocessor.

Doing magic with Go via reflection or other implicit behaviours is generally annoying to deal with. One example is some libraries using struct tags, most of the time they work as expected, but sometimes you get some weird failure and these kinds of implicit behaviours are the culprit.

Overall, I don't rate these "all in one" frameworks highly in Go. The standard library is excellent for most applications, you only need to add some code to remove some boilerplate. For most apps that I have worked on in Go that involved web components, we maybe had to import e.g a websockets library or a more elegant routing library, but that's pretty much it.

Incidentally, I've found golang to be more verbose than Java.

Generics help a little bit, but the fact remains that error checking is pervasive and verbose, and you can't chain calls (e.g. slice.Map(func(a int) { return a * 2 }).Filter(func (a int) { return a % 2 == 0}).Reduce(0, func(a int, i int) { return a + i }) is not something that can be supported by golang when errors are involved, not to mention the extreme verbosity since it does not have a short way to pass functions).

Yes, writing function chains that compose from left to right, combined with the error checking convention, is a no-starter. So maybe you embed an error, and then check for error non-nil everywhere ? Or does KISS imply that you just panic and recover ? Some intensive googling finds that there's been many attempts to design an easy to use, easy to understand pipelining convention, but I have not found any one of them to be convincing.

Go needs a better ORM story, it is unfortunate Prisma abandoned their Go port.

SQL is an incredible language with decades of theory and implementation behind it. I have never found an ORM to be better or faster or more useful than hand-written SQL for anything but the simplest of queries, and even then the benefit is debatable.

99% of a CRUD app -- which is most of them -- is going to be simple CRUD stuff. Load a thing from database, let user edit it, save it. Doing that in 'plain' sql, even with templating, is repetitive and error prone.

An ORM that didn't let you escape to SQL when you wanted to do more complex things would be a total failure, of course -- but to suggest that they're not more convenient for the 80/20 or even 90/10 use case is just really hard to understand. Must be NIH syndrome.

| Must be NIH syndrome.

I strongly disagree that not wanting to add yet another abstraction layer onto a proven technology would be NIH.

If you're using postgres, I'm a major fan of the anti-ORM (ROM?) that is sqlc. Nothing I've used comes in Go close in productivity or safety (and ability to write proper queries but use them very simply, and stay up-to-date without too much extra toil). Last I checked they're also working on adding sqlite support.

https://github.com/kyleconroy/sqlc/

They're on top of my list to add to Copper. As soon as we get sqlite support!

As the post you're replying to describes, magic is antithetical to Go idioms. An ORM is basically database-as-magic, every ORM in Go is just a nightmare.

Something like sqlc[0] is leagues better than any ORM in terms of simplicity, complexity and maintainability.

0: https://github.com/kyleconroy/sqlc

ORM's help do simple stuff well enough, but they sure make hard tasks harder. Ever try to construct a 30 line report query using this years language flavor of an ORM?

https://sqlc.dev/ makes your SQL the focus, not your Go-specific query code.

Lots of ORMS have an escape to SQL for querying, or you can choose to ignore the ORM for querying. As an example, ActiveRecord allows you to do this:

  Client.find_by_sql("
  SELECT * FROM clients
  INNER JOIN orders ON clients.id = orders.client_id
  ORDER BY clients.created_at desc
  ")

Which returns Client objects. Insert your own vastly more complex query above.

I think an ORM like https://github.com/xo/xo with https://sqlc.dev/ as a fallback for complex queries will be a killer combo!

Two high level abstractions that compile into another high level abstraction called SQL.

Why increase the complexity of something that's already a high level abstraction? Abstractions are about simplification. An orm and sqlc are not it.

https://sqlc.dev/ is not an ORM. It's not an abstraction over SQL. It is a brilliant way to make SQL the focus of your models by generating the code from your SQL, instead of the other way around.

Ah I see. My mistake.

What about the Ent ORM library?

https://entgo.io/

Ent can replace an ORM in your architecture, and it's easiest to explain in two seconds as "an ORM for Go", but it's not really that. It's a way to describe a persistent object graph, without any reference to persistence or mapping details. When your chosen persistence layer is an RDBMS, as it often is, then it uses an ORM - but you rarely interact with it at that level even when specify your entity schemas. You can back it with an object DB or REST API instead and then it wouldn't need the RM part at all.

Orms are a huge mistake.

SQL is already a very high level API that compiles to low level algorithms? Why put another very high level API on top of that?

To top it off SQL is a leaky abstraction by nature. Two high level SQL queries with equivalent results can both compile into algorithms with COMPLETELY different performance profiles. You have to manipulate the SQL query to generate the correct performance profile. This means understanding things below the SQL abstraction.

You put a ORM or any new abstraction on top of that guess what? That abstraction must compile into hacked SQL. You have to be able to manipulate the ORM such that it generates the correct SQL such that the correct SQL generates the algorithm with the correct performance profile. The leak from the sql abstraction must propogate into the ORM layer which in itself must be a leaky abstraction. It's like dealing with a leaky pipe embedded within another leaky pipe.

Optimizing SQL is already a domain knowledge thing. Now you have to use new tricks to optimize the abstraction on top of it. Two Leaky abstractions on top of each other and both very high level is a bit of a head scratcher.

Why do people even make these abstractions that only make life harder? I think it's an illusion. It's to satisfy an OCD thing but people don't realize the OCD is an illusion. People want to deal with a single language, not have strings of another language living in the code. For example, SQL strings are seemingly kind of ugly in something like GO code.

Databases are the classic bottleneck of web development in terms of speed. Optimization is a very important part of writing SQL queries as a result. Having orms and other high level abstractions on top of this area is much much more harmful then it otherwise would be if databases did not exist in this bottle-necked area of web development.

In actuality it's also questionable whether or not a leaky abstraction in the first place was the right design decision. Is SQL the right abstraction for database queries? Or should we design another high level API that has a more 1 to 1 correspondence with optimization as in a High level API that's not leaky, like What Rust is to systems programming.

This is a good observation. One trend I'm noticing is that the "old way" (PHP, Rails, etc.) of doing things is making a comeback. Go is very well positioned for this but lacks the frameworks.

I'm hoping to add something like Phoenix to Copper. It should help with the "heavily API only side" problem. I've already added integrations for Tailwind and added some utilities on top of the templating (going to add more) to fix the lack of good templating.

Go and Java can never use the same framework patterns as Python/Php/Elixir/Ruby/JS because they are not dynamic enough. The Rails-style request mapped dispatching into active record ORM pattern requires a lot of flexibility on the host language side. For Go and Java, you basically end up with code generation or reflection, and the latter is a killer for performance. On the other hand, the performance of Go and Java is better by several orders of magnitude.

Models change pretty slowly, so codegen is a good solution! Ent uses codegen for its query API which looks pretty nice: https://entgo.io/docs/tutorial-todo-crud

I’m considering writing a Typescript clone of Ent with codegen powered by Typescript types. I like codegen over dynamic magic because the runtime behavior is often easier to understand. In Rails, I need to traverse a lot of space in Pry’s debugger mode to figure out WTF is happening. I would much rather have codegen.

Reflection is killer for performance because it brings Java/Go down to the level of a JIT-less language like Python (and ORMs aren't too kind to JITs in e.g. Ruby/JS either). The problem with reflection in Go/Java is mostly that it's hard to read, since the reflection-ful APIs are hosted rather than native syntax.

Agree reflection has (often big) trade offs but its not as bad as it was years ago, especially if its used during construct time rather than runtime, if you combine it with runtime caches you still get the benefits imo

They need to solve the same problems. They don't have to solve them in the same way.

Bud author here, thanks for including Bud on your list. That's a really good overview of the landscape!

My take is that Remix + Next.js + SvelteKit are going to continue to innovate fast in the Frontend and "Backend for Frontend" space. Rails and Laravel don't hold a candle to the experience you get in that ecosystem.

But the JS ecosystem is massive and, as a result, fragmented. As you mentioned, there's no consensus on ORMs, mailers, queues, etc.

I don't see those frameworks trying to push too far in that direction, they'll remain "UI focused". This is nice for their focus, but not great for someone who wants to launch a web app and doesn't want to figure out all the surrounding ecosystem tooling.

This is where Laravel, Rails (and soon Bud) (and I assume Copper) will shine. They provide more tools and interfaces out of the box for building full-featured backends. These frameworks definitely need to keep an eye on the best ideas coming out of the JS framework ecosystem though!

A rails-like framework will never happen in a language that doesn’t have the same meta programming capabilities as Ruby. Rails exists because Ruby exists not because DHH just happened to be a Ruby programmer. There is a reason people have tried to recreate it in other languages and it always feels jank - because Rails is designed specifically and enabled by the Ruby language.

Code generation provides the same kind of flexibility you get with meta programming, you just need to do more work to keep generated code in sync and out of the way.

Like I said: jank.