I love Rails and how simple and boring[0] it is, and appreciate how it keeps evolving and growing. I've dabbled with some Javascript frameworks and Elixir/Phoenix, but always come back to Rails and forget just how fast and productive I can be. A lot of people complain about it, but I feel like Active Record just can't be beat, and helps me feel so productive. Add Stimulus and Turbo and the level of UI fidelity possible is impressive.
You can like rails, but calling it boringly simple is highly misleading. I think, generously, something like "complex but fully-featured" might be accurate.
As a reference point, the top 2 rails books on amazon are 900 and almost 1100 pages.
I say this to warn those who are looking for something minimalist and quick to learn -- Rails does not fit the bill.
I love how they focus on simplicity and are not afraid to cut stuff out. Often times, software projects get more complicated over time. Rails seems to get simpler.
it has added stuff (attachments, some wysiwyg editor, docker support) while removing other. It's neither getting smaller nor bigger, it just "moves" (maybe following Basecamp's needs).
(1) IMHO they've flubbed the progressive typing. I'm spending some time learning Typescript and React. And it's amazing. The type system is flexible, powerful, and works well with an IDE. It's just way faster to write when code gets complexer. The fact that it's null and type safe is the cherry on the top. Amazing that Javascript with all of it's warts can be turned into this.
(2) They've never really got the front end story right. Coffeescript was a dead end, then Turbolinks was the thing, and now it's Hotwire. None of them are really right. Rails will always be a primary SSR paradigm. They just need an easy way to do basic dynamic things like show/hide fields and form error checking clientside. Ideally something written in Ruby and well integrated into the form builder.
A bonus for (3) is that the Rails UI is ugly out of the box. Wish they'd have more integrated UI components that are production ready.
1 has nothing to do with Rails. Rails is a framework, Ruby is the language.
As for 2, I see this as a common complaint, but Hotwire, Turbo Streams with ViewComponent is a really power combination whose full potential is still being worked out.
As for client-side error checking... why? So I have to maintain validations in two places and ensure they always match?
I don't even understand your 3. Rails has no UI. Do you mean the CSS applied to the scaffolding views?
I agree with your number 2 especially, but think that Stimulus is doing a fairly good job at it now. The Turbo side of Hotwire still feels like a bit of a mess, especially once you try to do a lot with it. But little Stimulus controller additions here and there are (with the main exception of managing a lot of verbose data attributes) almost a delight to work with.
I was pretty sad to see the pattern matching PR reverted for 7.1 [0], but I guess the Rails team wants to make sure they nail the execution (?) even if that means being late to the pattern matching party, which is arguably one of the best additions to Ruby since the 1.9.3 hash syntax.
Yup, I was hoping this would make it to 7-1 but I suppose there were too many cases to consider.
I wonder if certain features could be marked as "experimental" by Rails core like they do in Ruby, and include them without the promise that the API would be stable (or even be present at all in the future).
I think it was reverted because the Rails team isn't sure what values of a model should be available to pattern match, and how each attribute or association should be pattern matched.
I’ve never seen a more productive stack. Enables me to build complete apps as a single dev - which would take 2x-5x of the time and effort if using separate front end apps. Ruby makes me happy when I use it. Hotwire and Stimulus let me create front end interactions without a separate SPA. The ecosystem is mature and well maintained. Absolutely love it.
Rails is really a gift. It's so productive and works so great on teams because of the heavy conventions I think. And the ruby library ecosystem is insanely good. Libraries generally do what they say they're going to do and there are libraries for everything. I've wondered a lot about why this is. Heavy air time? Is there something about ruby?
Every now and then I try some new web app framework that catches my eye in some new language and it always seems like its just getting started compared to rails -- and, it is. Worth noting -- rails works great API only as well. I bet Django is pretty good by now too of course, but Rails is really a fantastic tool well suited to task.
Because of lazy evaluation, type issues persist though. Plenty of instances of trying to call string methods on integers, and array methods on nil, and Rails devs in denial that that's a big problem as the app continues to grow.
I don't think it's denial - I've worked on a ton of rails apps of all different sizes and scales and...I've gotta say, I just don't see the issues you name in production all that often.
> Because of lazy evaluation, type issues persist though. Plenty of instances of trying to call string methods on integers, and array methods on nil, and Rails devs in denial that that's a big problem as the app continues to grow.
This is not anymore inherent in the language. Gradual typing support for Ruby is now reasonably mature (companies like Shopify with large codebases have adopted it), so it's up to the team do decide if they want to take advantage of it or not.
> and Rails devs in denial that that's a big problem as the app continues to grow.
As someone who maintains Rails 3 apps ad infinitum, this is where things break down with Rails applications. You can create applications very quickly with Rails if you stay within the conventions. But as the app grows, that discipline breaks down and the structure starts to suffer.
This is true for a lot of web applications. But I think the problem is worse for Rails because it's so easy to get started and create apps. That ease comes back to bite you -- as a giant legacy Rails app that no one on the team understands and no one is interested in fixing or maintaining.
Because of lazy evaluation or because the developers aren't properly documenting their work? A language with a formal type system essentially forces you to provide type documentation, but there is an expectation with dynamically typed languages that you will still document the types (probably in your test suite). Rails in particular makes this a core function of the framework to really push you to do so.
I love TypeScript, but Rails has a wonderful ecosystem of libraries that just feels unbeatable for the backend. For some reason (hot take incoming) the JS community seems to hate working with each other and improving existing libraries. The Ruby community seems to rally around improving our libraries (gems) rather than reinventing them every year.
I recently tried Nextjs and Remix to explore other stacks since I've been using Elixir and Phoenix since 2016.
You won't find the dev UX you take for granted in Elixir and Phoenix.
Endless routing options, background jobs? Just install Squirrel and yadabadabadoo or just use AWS SQS(https://old.reddit.com/r/nextjs/comments/qspw4v/how_to_do_ba...). What about solid backend processes? Can I call those in some kind of repl? In Elixir I can just go `iex -S mix phx.server` and `MyModule.foobar("test")` and bada bing I'm all set. What about logging? Wait I need to decide and configure a logger? I also need to decide and configure what to output to? What about a nice ORM? Prisma looks good, set that whole thing up from scratch, I need to await, but special considerations need to be thought of when iterating through a collection? Promise.all vs for..of? I iterated an array and want to save it to the DB but it's running out of connections (https://github.com/prisma/prisma/discussions/16884)? Why do I even need to think about this?
Really curious to hear your counterpoints to this because granted I am quite new to backend typescript.
The counter argument is that you don't have to babysit any of those services. AWS/some other PaaS will handle everything for pennies. Unless you are working in a low cost of labor country, developer time is going to be more expensive than anything else. It will take very large scale/growth for infrastructure fees to exceed the total compensation of a decent US staff engineer or a site reliability engineering team.
Still using it, still highly productive. Not doing as much Rails work as I was 10 years ago but that's due to a change in the nature of my business. We still use it and Sinatra for the bulk of our Internet-facing and internal stuff.
At the backend side, it makes easy to build websites. However, the UI layer is a bit complex to integrate with common framework libraries like Vue and React.
Switched to Elixir/Phoenix a few years ago and have never looked back. Absolutely a 100% improvement on Rails in every way (except the availability of work!)
I revisited rails this past week for the first time in nearly 10 years. So much easier to quickly develop an app on than what I have been doing. I feel done with the whole SPA thing unless there's no other way to build the app.
But now I'm looking at performance, and memory usage, and phoenix looks very appealing from that side.
Now I think I should try pheonix, since both rails and phoenix are new to me, so I can compare.
Writing typescript code that compiles to javascript code that runs on a platform like node.js, and writing typescript code that compiles to javascript that run in a web browser.
Rails is unfortunately much slower, and it’s quite coupled to the concepts of server-side rendering and OOP MVC. Making Rails work with a modern frontend feels pretty hacky compared to using a full-stack TypeScript framework.
I'll give you the slow, but "modern" frontend is a breeze. Webpacker (yes, the deprecated one, newer solutions are worse) handles React beautifully, and my graphql API definition gets seamlessly converted into typescript types. If anything, it often feels even easier than the code sharing I've experienced in node-land.
Rails are good enough to keep Shopify standing for another year of holiday sales. It seems to do that since 2006, so this alone demonstrates that Rails can support you a long way in your multi-billion business.
And if you think: "yes, but Shopify has a big talented team," I feel the need to say back, "Well, when you are at Shopify business size, you will have it too. And Shopify used it when they were small too."
I am not saying Rails fits everything, but if your business fits Rails, that is a great framework to build with.
I went from a full stack typescript project to a rails project and honestly I hate it.
The Ruby language is great if you want to jerk off about how concise your code is but programmers end up creating overly abstract, write-only code.
There’s so much magic that it’s hard to trace the code to see where stuff comes from.
Abstractions are over-engineered. Serializers should be simple, async functions that are easy to step through and debug. Instead you have serializer relations? Delegates, etc.
Don’t even get me started on updating rails itself. It’s a massive pain every time and you quite often see projects that are multiple major versions behind.
In typescript it’s common to integrate multiple independent libraries into your own framework. This is great because if I need to update my database library I can focus on that specific part of the code. Updating rails means everything could potentially break.
Great to see so many new features in Rails. Tbh, one of the reasons I use it for certain backend projects is how simple to do complex things.
Managing files and image variants is a pain when developing a site. Rails make it easy and even adds more features around.
However, I still think the UI layer is too coupled to the backend. Webpacker was a try, but I understand why it's deprecated. I would go on supporting common frameworks, even if it's just a template to organize the code.
Finally upgraded some internal applications to Rails 7 this month, and I was very happy with the asset pipeline changes. Rails' asset pipeline management hasn't really meshed with our development strategies for CSS and JS for a long time, and it felt like we were always fighting with trying to make our process fit Rails' expectations.
We're on Propshaft now and I expect far fewer asset issues in the future :P
I am really excited about the possibility that Rails finally get something in the way of ActiveDeployment. This tweet https://twitter.com/dhh/status/1604856555848884225 and the commited changes make me wish and wonder.
My impression is that there are two ways to use rails:
1. Your frontend js needs are pretty minimal and you don't want something approaching a single-page app. In this case, you lean into Turbo and Stimulus to have rails do most things for you that would previously have been done with some a heavy-weight react-based frontend.
2. Your frontend js needs are significant and you do want a single-page app. In this case, ignore/disable turbo + stimulus. Instead, build your rails app as just a json API and then have a separate project for your frontend built using a dedicated frontend build tool like vite/CRA/webpack/etc.
On one hand, I feel like the turbo/stimulus stuff is the rails community preferring to pretend that the JS world doesn't exist. On the other hand, maybe it's really just the rails world embracing the above dichotomy: they're tools for doing some light frontend work with no JS, and an acceptance of the fact that if you want an SPA, there are better non-rails tools that you'll wind up using anyway.
My employer fits your first criteria precisely. Turbo and Stimulus provide exactly as much interactivity as we need. The appearance is that we're trying to avoid Javascript. The reality is that we are (successfully) minimizing context switching when working on the code base.
We went with esbuild and regardless of speed (it's way faster) the main reason I like it is because the config is tiny and I actually understand what it's doing. Our Webpacker config was big and fragile by comparison.
Readit News
[0] https://boringrails.com/
As a reference point, the top 2 rails books on amazon are 900 and almost 1100 pages.
I say this to warn those who are looking for something minimalist and quick to learn -- Rails does not fit the bill.
Which is (arguably) a good thing.
(1) IMHO they've flubbed the progressive typing. I'm spending some time learning Typescript and React. And it's amazing. The type system is flexible, powerful, and works well with an IDE. It's just way faster to write when code gets complexer. The fact that it's null and type safe is the cherry on the top. Amazing that Javascript with all of it's warts can be turned into this.
(2) They've never really got the front end story right. Coffeescript was a dead end, then Turbolinks was the thing, and now it's Hotwire. None of them are really right. Rails will always be a primary SSR paradigm. They just need an easy way to do basic dynamic things like show/hide fields and form error checking clientside. Ideally something written in Ruby and well integrated into the form builder.
A bonus for (3) is that the Rails UI is ugly out of the box. Wish they'd have more integrated UI components that are production ready.
As for 2, I see this as a common complaint, but Hotwire, Turbo Streams with ViewComponent is a really power combination whose full potential is still being worked out.
As for client-side error checking... why? So I have to maintain validations in two places and ensure they always match?
I don't even understand your 3. Rails has no UI. Do you mean the CSS applied to the scaffolding views?
Please allow me to introduce Sequel (https://sequel.jeremyevans.net/) for your consideration.
For now, I'm good with the little gem that kddnewton put together (https://github.com/kddnewton/rails-pattern_matching).
[0]: https://github.com/rails/rails/pull/45553
I wonder if certain features could be marked as "experimental" by Rails core like they do in Ruby, and include them without the promise that the API would be stable (or even be present at all in the future).
What's the general vibe on RoR nowadays?
Every now and then I try some new web app framework that catches my eye in some new language and it always seems like its just getting started compared to rails -- and, it is. Worth noting -- rails works great API only as well. I bet Django is pretty good by now too of course, but Rails is really a fantastic tool well suited to task.
Most people that use it are fine and productive.
Because of lazy evaluation, type issues persist though. Plenty of instances of trying to call string methods on integers, and array methods on nil, and Rails devs in denial that that's a big problem as the app continues to grow.
It's a problem, but it's not a big problem.
This is not anymore inherent in the language. Gradual typing support for Ruby is now reasonably mature (companies like Shopify with large codebases have adopted it), so it's up to the team do decide if they want to take advantage of it or not.
I've been experimenting with Rbs, Sorbet & Contracts though, they have their potential.
As someone who maintains Rails 3 apps ad infinitum, this is where things break down with Rails applications. You can create applications very quickly with Rails if you stay within the conventions. But as the app grows, that discipline breaks down and the structure starts to suffer.
This is true for a lot of web applications. But I think the problem is worse for Rails because it's so easy to get started and create apps. That ease comes back to bite you -- as a giant legacy Rails app that no one on the team understands and no one is interested in fixing or maintaining.
As I add type annotations to my methods, VS Code and Sorbet are giving me similar feedback to what I get from Typescript projects.
Deleted Comment
You won't find the dev UX you take for granted in Elixir and Phoenix.
Endless routing options, background jobs? Just install Squirrel and yadabadabadoo or just use AWS SQS(https://old.reddit.com/r/nextjs/comments/qspw4v/how_to_do_ba...). What about solid backend processes? Can I call those in some kind of repl? In Elixir I can just go `iex -S mix phx.server` and `MyModule.foobar("test")` and bada bing I'm all set. What about logging? Wait I need to decide and configure a logger? I also need to decide and configure what to output to? What about a nice ORM? Prisma looks good, set that whole thing up from scratch, I need to await, but special considerations need to be thought of when iterating through a collection? Promise.all vs for..of? I iterated an array and want to save it to the DB but it's running out of connections (https://github.com/prisma/prisma/discussions/16884)? Why do I even need to think about this?
Really curious to hear your counterpoints to this because granted I am quite new to backend typescript.
Although, I still use it in some projects :)
Maybe InertiaJS [0] would help.
[0] https://inertiajs.com/
But now I'm looking at performance, and memory usage, and phoenix looks very appealing from that side.
Now I think I should try pheonix, since both rails and phoenix are new to me, so I can compare.
I understand that it is functional, which seems neat, but how does that translate into better app development?
Deleted Comment
Rails are good enough to keep Shopify standing for another year of holiday sales. It seems to do that since 2006, so this alone demonstrates that Rails can support you a long way in your multi-billion business.
And if you think: "yes, but Shopify has a big talented team," I feel the need to say back, "Well, when you are at Shopify business size, you will have it too. And Shopify used it when they were small too."
I am not saying Rails fits everything, but if your business fits Rails, that is a great framework to build with.
The Ruby language is great if you want to jerk off about how concise your code is but programmers end up creating overly abstract, write-only code.
There’s so much magic that it’s hard to trace the code to see where stuff comes from.
Abstractions are over-engineered. Serializers should be simple, async functions that are easy to step through and debug. Instead you have serializer relations? Delegates, etc.
Don’t even get me started on updating rails itself. It’s a massive pain every time and you quite often see projects that are multiple major versions behind.
In typescript it’s common to integrate multiple independent libraries into your own framework. This is great because if I need to update my database library I can focus on that specific part of the code. Updating rails means everything could potentially break.
Oh and there’s no type system so good luck.
Managing files and image variants is a pain when developing a site. Rails make it easy and even adds more features around.
However, I still think the UI layer is too coupled to the backend. Webpacker was a try, but I understand why it's deprecated. I would go on supporting common frameworks, even if it's just a template to organize the code.
We're on Propshaft now and I expect far fewer asset issues in the future :P
How do these new tools compare with the old? Versus Webpack, Vite etc?
1. Your frontend js needs are pretty minimal and you don't want something approaching a single-page app. In this case, you lean into Turbo and Stimulus to have rails do most things for you that would previously have been done with some a heavy-weight react-based frontend.
2. Your frontend js needs are significant and you do want a single-page app. In this case, ignore/disable turbo + stimulus. Instead, build your rails app as just a json API and then have a separate project for your frontend built using a dedicated frontend build tool like vite/CRA/webpack/etc.
On one hand, I feel like the turbo/stimulus stuff is the rails community preferring to pretend that the JS world doesn't exist. On the other hand, maybe it's really just the rails world embracing the above dichotomy: they're tools for doing some light frontend work with no JS, and an acceptance of the fact that if you want an SPA, there are better non-rails tools that you'll wind up using anyway.
37Signals have both their Basecamp and Hey.com products which act very much like a single page app but only using Turbo and Stimulus.
I think the rails devs believe that single page apps are overused and the majority of functionality can be done using just the provided tools.