Rubyspec is dead, long live Rubyspec
gist.github.com
gist.github.com
Ruby/Sinatra used to be my favorite combo, but since Rails started going off in a weird direction (anti-js, asset pipeline, choosing AR over DM, insecure defaults pre rails 4) there has been less good stuff written in Ruby to draw from to augment Sinatra... Rails has become a rapidly changing but rarely improving API. That Ruby 2.2 breaks Rails 3.2 is just one more reason to look elsewhere.
I question why anyone would start a new project using Ruby at all. This is not a troll, I really don't think there are significant benefits anymore.
You specifically mentioned starting a new project. I think that's where Ruby and Rails are the strongest. Maintaining a large system might be better done using some other language and framework.
I think this is an aesthetic preference. I personally think that the simplest and cleanest design is a combination of something like Sinatra/Express to implement a simple API and React on the client to do component-ized UI.
Much of the "completeness" of rails is simply not needed when client side components can keep a bit of state around for housekeeping and make the occasional API call when data actually needs to move between client and server.
IMHO the component abstraction provided by something like react leads to much cleaner separation of concerns, better boundaries between parts of the system, and easier to change/understand code.
One might start a new project with Ruby if it were one's primary language, and there was a desire to build something without the overhead of working in another language in which one is less proficient.
How about this analogy: there are many different ways to cook a meal, but often my goal is to prepare food easily and quickly so I can eat and move on. Thus, I usually stick to the relatively few methods with which I am comfortable. Sometimes I do experiment with the goal of learning something new. But in those cases, the primary goal is expanding my skills, not building the "product" as it were.
edit: clarification
It seems to me that Ruby is doing just fine - the community is thriving, MRI is advancing quite rapidly. In reality people find Ruby to be a good solution.
> I question why anyone would start a new project using Ruby at all.
Actually there's a whole Ruby world out there besides Rails, even for web apps/services. I've been tempted numerous times to start new projects using alternatives such as Erlang, node.js or Go, only to find myself finally going back to Ruby. The reasons: it's versatile, convenient, familiar, and to be honest, for much of the stuff that Ruby does, and for many different uses, it performs just fine, and it continues to improve at a steady pace.
> This is not a troll, I really don't think there are significant benefits anymore.
A sweeping statement like this, how can it not be taken as trolling?
Note that I said significant benefits. There are still some benefits but I can't currently think of any reason to use Sinatra over Express, and I love Sinatra.
Ruby optimizes for programmer happiness - that was Matz's explicit goal when writing the language. If that doesn't matter to you as much as speed on microbenchmarks, or if writing Ruby doesn't make you happy (it makes me happy!) then Ruby isn't for you.
Consider how many heroku dynos are dedicated to running slow, non-concurrent Ruby web processes, and how many Kilowatt Hours are burned waiting for Rails to apps to start. These are just not problems people need to deal with anymore.
So while I am a big fan of Ruby (I still think it's the closest to my brain's own pseudocode of any language I've used) the ecosystem is lagging significantly these days, not because it's not improving, but because js is improving at a must faster clip.
I can't see how you can assert that the JS ecosystem is "better". It is rapidly changing which is partially detrimental. What worked or was considered a good choice at the start of 2014 is now mostly obsolete or abandoned by the end of 2014.
As so many companies have shown it really does not matter what programming language you use as there are plenty of options to make it work, it's more about choosing the language that will provide the best fit for your business.
This is happening to supposedly mature, nearly decade-old Rails!
1. Ruby is not built from the ground up for concurrency and eventing. You can use a number of great tools to effect that, but it isn't built for that from the ground up.
2. Rails continues to forge ahead with weird JS and asset management at the Ruby community's peril. We need to be taking a very serious look at tools like:
http://scorchedrb.com/ - which provides a (more) correct pattern for routing
and things like
https://github.com/apotonick/trailblazer
which provides an alternative code organization method which might make more sense.
Most of our web apps are essentially state machines and the only framework which acknowledges it is:
https://github.com/seancribbs/webmachine-ruby
3. Ruby needs to be faster. The community needs to get library support for 2.2 ASAP. We need to use Fibers and we need to use Celluloid.
https://github.com/celluloid/celluloid/
4. I'd love to see in 2015 less new projects (fragmentation?), but more forking and more collaboration. That goes for the JS world as well.
2. I definitely agree about Rails and the asset pipeline. It feels awkward and kludgey and particularly in relation to JS I'd like to see code organization and tooling stepped up; many of the apps we're writing today are heavily Javascript, and it's a lot of work to maintain a large JS app on top of a Rails app right now. That's not to say that it's prohibitive, but Rails built a name for itself by making the drudgery less painful, and this is an open pain point.
3. Agreed, though JRuby is again a trailblazer here; boot-up time sucks (I'm working on that over Christmas break!), but 9k and the Truffle+Graal projects are huge leaps forward, and once JIT kicks in, JRuby's long-term performance is really excellent. MRI continues to make strong advances here, as well; the discussions around MRI 3.0 have hinted at some really good stuff coming down the pipe.
4. Can you elaborate on what specifically you mean? Ruby as a whole is absolutely stronger for the "new projects" (RBX, JRuby, Topaz); while issues like this can become divisive, I'll argue that RBX has been a net positive to the Ruby community, both because of RubySpec (a different perspective is valuable, even if it doesn't become the official spec) and because a new approach to a problem generally ends up leaking improvements all over the ecosystem.
As for #4, I'm all for those mentioned, low-level (and fascinating) projects. And for projects surrounding programming paradigms e.g. Celluloid, EM... I was really referring to libraries, but on second thought #4 isn't that a big of an issue in Ruby-land, it's mostly a JS world problem.
I suppose my particular fear is that instead of re-organizing code within Rails as an example, the community will opt to make another new framework instead, increasing fragmentation.
This really shines in places where CRuby has lagged behind - places like numeric computing and natural language processing. The library support is iffy at best, and usually the answer is "use python", but when you can just drop in a Java library and call it a day, stuff gets really fun.
Profiling a live JRuby app with JMX and jvisualvm is amazing - I can attach to a live, remote production instance and perform heap dumps, start sample profiling, or even execute Ruby code! Going to back to fix our MRI apps has left me very discontent with the state equivalent tooling :)
My primary complaint with Java is that it's verbose and lends itself to overcomplex design, but I absolutely love that I can write high-performance code in Java and then write all my business logic in Ruby - it's the best of both worlds. To illustrate, I wrote Manticore (https://github.com/cheald/manticore) to solve my persistent discontent with the state of HTTP clients in Ruby; the Apache Commons HTTP components are extremely comprehensive, fast, powerful, etc, but the API is horrific. I spent a bit of time wrapping them up in a nice idiomatic Ruby interface, and ended up with an HTTP client that destroys the pure Ruby clients for features and performance, and even beats out even Typhoeus (which is really just a set of libcurl bindings) - and all in only 700 lines of code!
Finally, the JRuby community is so welcoming and inclusive. #jruby on Freenode is populated with all the project core members, and they are very willing to answer questions and help out. headius (Charles Nutter) has turned around bugfixes for issues in minutes (in addition to being extremely friendly and helpful), and they as a group have made it very easy for me to contribute to the project. I've gotten a few dozen pull requests accepted into master in the last few months, and thoroughly enjoy working on it. It has easily been one of my best experiences in open source.