Charles Nutter. How to move your Ruby project to JRuby and why

Charles Nutter works on JRuby and JVM language support at Red Hat.
03 October 2018   944

— How did you get into programming and into Ruby world?

— In 2004, I was working at a government contracting firm as a Java Enterprise Architect. I was in charge of a large mainframe conversion for the United States Department of Agriculture, which meant I spent a couple weeks a month in the Washington D.C. area. One of those trips coincided with RubyConf 2004, and since a close friend had been recommending I look at Ruby, I decided to attend. So there I was sitting in a Ruby conference without ever having learned Ruby...and I understood every piece of code, every example. I was amazed and vowed to find a way to bring Ruby into my Java world.

— Which projects are you working on now?

— My primary role is as co-lead of JRuby. This also means supporting several side projects like our native-library backend (Ruby's FFI library is maintained by us using this backend on JRuby) and our String encoding subsystem (an elaborate port of the logic from CRuby). I also do much of the outreach to the community and try to make sure our users are getting their issues addressed. There's always plenty to work on!

— Which one would have the biggest general impact from you opinion?

— I like to think that JRuby, while not the most popular JVM language, has at least helped to change the JVM platform. Because of our collaborations with Sun Microsystems, Oracle, and others, we have solid dynamic language support at the JVM level along with many other projects to support alternative languages. The JVM today is a much more hospitable home for non-Java languages than it used to be, and I hope we've played some small part in that change.

— Which languages are you writing on in your everyday life? Which one do you like most? Why?

— Most days I write in a mix of Ruby and Java, since JRuby is implemented using both. I like both languages for different tasks. Ruby is a better language for building applications that need to evolve and adapt quickly. Java is a great language for writing high-speed, reliable libraries and services. JRuby gives you the best of both worlds!

— Do you like to program in Java?

— I do, especially with all the language improvements that have been added recently, like lambdas (closures or blocks in Ruby) and the new "var" syntax for declaring local variables when the static type is unambiguous.

— What do you think about Rust?

— Rust is a great language! I did a lot of C++ development in my college years, and I can tell you right now if I'd had Rust available then I would have used it. I'm especially jealous of the static-typed ownership model, which helps avoid thread-safety issues like races and data corruption. I hope to see other languages adopt this pattern in the future.

— What do you think about the Ruby language perspective? Is it or its community dying?

— After all these years working on JRuby, I do still love Ruby syntax and the Ruby way of doing things. However I worry that the language is held back too much by limitations of its primary runtime. JRuby has been fighting to make true parallel threading a reality for Ruby developers, but still today the vast majority of Ruby services are run using multiple isolated processes, wasting tremendous amounts of CPU and memory resources. I believe this is due to the C API for writing Ruby extensions being so large and so prevents many improvements -- including parallel threading -- from being developed. Hopefully we'll see this change some day.

— Which upcoming or not well-known features of Ruby language would rush in future?

— I look forward to strings becoming immutable-by-default, as they are in most other languages. Parallel programming would be much simpler if more of Ruby's objects supported pure-immutable or "deep freeze" semantics. It's a bit like the Rust ownership model...if you're going to be sharing an object across threads, choose the version of that object that you know can't be modified anymore. This extends to arrays, hashes, and just about every other mutable object in Ruby: we need to make it easier to lock down mutable data.

— Could you give me an advice on how to move my ancient monolithic project to JRuby? And should I?

— The first question really is whether such a move would benefit you. There's many good reasons to consider a move to JRuby:

  • Reducing CPU  and memory costs in a shared hosting environment by utilizing JRuby's true parallel threading
  • Deploying a Ruby application into a JVM-heavy environment, such as used by larger financial or government organizations
  • Needing access to libraries that only exist on the JVM, or that are more portable or scalable on the JVM than their Ruby or C equivalents
  • Getting a little performance boost out of CPU-heavy or concurrency-heavy applications.

— I would say if your application is scaling well and not costing you too much today, perhaps you don't need to make a move. But if you decide you need more out of Ruby, here's the process for migrating:

  • Do a clean bundle of your application, paying special attention to C extensions you may be using. You can also do this bundling *on* JRuby, and then deal with missing or unsupported libraries one by one.
  • For each extension, search for a JRuby equivalent. We have some pages on the JRuby wiki to help with this. Most popular libraries have JRuby versions. If no JRuby version exists, you may look for a pure-Ruby version (it might be fast enough on JRuby) or a JVM library (in Java or Scala or Clojure or whatever) that could be used as a replacement.
  • Once your bundle completes, you should have a working JRuby application! We've worked very hard on compatibility, and try to be responsive if users find new issues, but a successfully-bundled application is expected to work.

The steps beyond this involve deciding how to take advantage of your newfound power: how many threads to throw at a given server, what you're going to do with all the money you're saving, etc.

— What should nowadays students learn to become good programmers?

— When I was at university, my earliest computer science courses used the Scheme language, a Lisp-like functional language that's great for teaching the fundamentals of programming. I still recommend that serious new programmers work through at least some of the Structure and Interpretation of Computer Programs book from MIT. Beyond that, I'd say learn as many different and unusual languages as you can; they'll all give you new ideas and new ways to look at programming problems.

— How do you keep yourself motivated for programming? Have you ever been "burned-out"?

— Burn-out is a real problem in our industry, and working in open source brings with it huge amount of stress. We've all felt that way sometimes...too much work to do and not enough time to do it, missing out on time with family and friends, ignoring our own health so we can fix one more bug. These days I try to center myself by keeping up with hobbies: playing video and board games, learning to play guitar, studying foreign languages, and traveling around the world to meet new friends. There's always this nagging workaholic telling me to get back on the job, but I'm learning how to maintain the right balance.

— What do you think about Russia and what do you expect of the upcoming RubyRussia event?

— I love Russia, and my speaking trips the past few years have been some of the most rewarding of my life. This will be my fourth visit, having been to Saint-Petersburg, Moscow, and Novosibirsk (!!!) previously. I'm looking forward to returning to Moscow and meeting the RubyRussia community I've heard so much about!

Questions by Dmitry Matveyev PM at Evrone

Git 2.20 to be Available

Let's check updates and features of new version control system
11 December 2018   117

Distributed version control system Git has received another update. In order for Git 2.20 to appear, 83 developers made changes 962 to the zcode. According to the team, this is an order of magnitude higher than the same figure in the largest release of the 2.x.x branch.

The git branch -l command is now a shortened version of the git branch --list - it used to help run reflog during the creation of a new branch. Developers have limited the launch of git fetch: it is only possible with an indication of --force, to avoid problems with consistency when updating the link.

The git help -a and git help -av commands to help newbies display a more verbose output. To return to the old view, just type git help - no-verbose -a. In git send-email, it is possible to extract lines with addresses that end with “-by” from signatures. This is an incompatible change, and it can be disabled by adding to the --suppress-cc = misc-by command.

  • If the repository contains files whose addresses differ only in the case of letters, a warning will be displayed during the execution of git clone.
  • The git format-patch command received the --interdiff and --range-diff options, which in a note or comment list the differences between the existing and previous versions.
  • git mailinfo learned how to recover code patches sent by email with plain text and damaged due to hyphenation.
  • git multi-pack-index now fixes damage in .midx files.
  • Creating experimental commit-graph files for large repositories takes a lot of time, so the developers have provided a form of output about the state of the process.

Performance and Development Support

  • For working builds, the -Wunused-function compilation option is provided.
  • git submodule update is completely rewritten in C.
  • One of the continuous integration (CI) tests, designed to work with the unusual/experimental/random settings, now supports midx and commit-graph files.
  • A new mechanism for finding objects among a large number of pack-files. It relies on combining all .idx files into one.

The previous version of the system was released in September 2018.