April 2026 - ruby-talk - ml.ruby-lang.org

[ruby-talk:444810] What took 1.5 months to build in React JS code took 1.5 days in Glimmer DSL for Web Ruby code
by Andy Maleh 28 Apr '26

28 Apr '26

What took 1.5 months to build in React JS code took 1.5 days in Glimmer DSL for Web Ruby code When I reimplemented a React component as the first Glimmer DSL for Web Component in my job's Fintech Rails web app, I thought I did what would have taken 1 week in React in 1.5 days in Glimmer, which seemed like a great improvement in productivity at the time (333.33% the productivity if we count 1 week of React as 5 business days). I recently inspected the Git timeline of the original React component, and discovered that it actually took 1.5 months to develop initially!!! Continue reading: https://andymaleh.blogspot.com/2026/04/what-took-15-months-to-build-in-reac…

2 2

[ruby-talk:444809] [ANN] llm.rb v5.1.0 Released
by Robert 25 Apr '26

25 Apr '26

Hello llm.rb is the most capable runtime for building AI systems in Ruby. It is built on the standard library (zero dependencies) but it can still support advanced features through optional dependencies that are loaded on-demand. For example, it can integrate with both ActiveRecord and Sequel, and supports persistent connection pools with net-http-persistent. Since my last email, skill support has been added. It behaves in a similar way to Claude or Codex skills, and launches a sub-agent to execute a skill - with a limited tool subset. [GitHub] https://github.com/llmrb/llm.rb#readme [RG] https://rubygems.org/gems/llm.rb [CL] https://0x1eef.github.io/x/llm.rb/file.CHANGELOG.html

1 0

[ruby-talk:444808] wroclove.rb 2026 workshop exercises - Building Rails SPAs in Frontend Ruby w/ Glimmer DSL for Web
by Andy Maleh 25 Apr '26

25 Apr '26

I published a GitHub repo containing the exercises of the wroclove.rb 2026 Ruby conference workshop "Building Rails SPAs in Frontend Ruby with Glimmer DSL for Web". This GitHub repo provides the simplest and fastest way to learn Glimmer DSL for Web (the Ruby-in-the-Browser Rails Frontend Framework that won a Fukuoka Prefecture Future IT Initiative 2025 award after getting judged by Matz and other Fukuoka judges). And, just in time for wroclove.rb 2026, Glimmer DSL for Web now has official Rails 8 setup instructions. https://andymaleh.blogspot.com/2026/04/exercises-for-wrocloverb-2026-ruby.h…

1 0

[ruby-talk:444807] [ANN] hoe 4.7.0 Released
by Ryan Davis 22 Apr '26

22 Apr '26

hoe version 4.7.0 has been released! * home: <https://zenspider.com/projects/hoe.html> * code: <https://github.com/seattlerb/hoe> * bugs: <https://github.com/seattlerb/hoe/issues> * doco: <https://docs.seattlerb.org/hoe/> * clog: <https://github.com/seattlerb/hoe/blob/master/History.rdoc> Hoe is a rake/rubygems helper for project Rakefiles. It helps you manage, maintain, and release your project and includes a dynamic plug-in system allowing for easy extensibility. Hoe ships with plug-ins for all your usual project tasks including rdoc generation, testing, packaging, deployment, and announcement. See class rdoc for help. Hint: `ri Hoe` or any of the plugins listed below. For extra goodness, see: https://docs.seattlerb.org/hoe/Hoe.pdf ## Features/Problems: * Includes a dynamic plug-in system allowing for easy extensibility. * Auto-intuits changes, description, summary, and version. * Uses a manifest for safe and secure deployment. * Provides 'sow' for quick project directory creation. * Sow uses a simple ERB templating system allowing you to capture your project patterns. Changes: ### 4.7.0 / 2026-04-22 * 2 minor enhancements: * Added cov_branches to enable branch level coverage reporting. * Merged hoe/gemcutter into hoe/package. Distinction doesn't make sense.

1 0

[ruby-talk:444806] [ANN] JRuby 10.1.0.0 Released
by Thomas E Enebo 21 Apr '26

21 Apr '26

The JRuby community is pleased to announce the release of JRuby 10.1.0.0. - Homepage: https://www.jruby.org/ - Download: https://www.jruby.org/download JRuby 10.1.0.x targets Ruby 4.0 compatibility. Thank you to our contributors this release, you help keep JRuby moving forward! JRuby 10.1: Experimenting and Evolving JRuby 10.1.0.0 is our first major release since catching up with Ruby compatibility. We have used this opportunity to explore many experimental optimizations and improvements we have been chasing for many years. Ruby 4.0 compatibility is largely complete, but there will be additional features added in update releases. We also plan to continue with large scale experiments in subsequent updates, and recommend production users continue to use JRuby 10.0 while we work toward stabilization of the 10.1 series. Here’s an overview of the major enhancments already in JRuby 10.1.0.0 and others planned for update releases in the coming months. Reduced Object Size In JRuby 10.1 all objects have been reduced in size by 8 bytes, bringing the base object size down from 32 bytes to 24 bytes. This translates into smaller objects across the system, reduced GC load, higher allocation throughput, and better cache locality. On JVM versions that support Compact Object Headers (JEP 450: “Project Lilliput”) the base object size can be further reduced to 16 bytes, 40% smaller than objects in the standard Ruby implememtation. Evolving Strings, Arrays, and Hashes During the upcoming JRuby 10.1 update releases, we plan to introduce new versions of String, Array, and Hash that are optimized for memory compactness, concurrency, or integration with other JVM languages. To support these features, we have made the internal implementations RubyArray and RubyHash into abstract classes and deprecated the direct constructors for RubyString. Please test your JRuby extension code and make sure it does not call abstract or deprecated constructors. Smaller Numeric Representations JRuby’s internal class for representing 64-bit signed Integer values, RubyFixnum, has been split into two versions: LongFixnum and ShortFixnum. Combined with the reduced size of objects overall, this reduces the size of these long and short-ranged Integer objects from 40 bytes to 32 bytes or 24 bytes (24 bytes or 16 bytes if using Compact Object Headers). The size of Float objects have similarly been reduced from 40 bytes to 32 bytes. These size reductions help improve the performance of numeric algorithms and reduce impact to the heap and GC of boxing numeric values. Cross-platform Prism Parser JRuby 10.1 ships with support for the Prism Ruby parser, packaged as a set of pre-optimized native builds (for Windows, Linux, and MacOS on x86_64 or arm64) or as a WASM build on other platforms. We’re still working out a few remaining features, but we encourage experimenters to enable it by passing -Xparser.prism to JRuby. We plan to make this the default parser once the remaining issues are solved. Direct Argument Passing We have also laid the groundwork to start passing all forms of method arguments on the call stack, avoiding boxed collections for most optional and keyword arguments. In upcoming JRuby 10.1 releases, we will begin to connect these call paths up for both pure-Ruby and native (Java) method targets, eliminating box allocation and drastically improving the peformance of such forms. Watch this space for rapid improvements. New Policy for EOL and LTS Releases With the release of JRuby 10.1, we are moving JRuby 9.4 into a “soft” EOL status. This means we will no longer be doing regular OSS release updates of JRuby 9.4, but users may get in contact with us to sponsor continued support. JRuby 10.0 will be our stable LTS release until April 2028 and we recommend all users upgrade or get in contact with us for assistance. JRuby 10.1 will be our “tip” release line, with many exciting optimizations and enhancements due to land over the next several months. Going forward, we will provide a new LTS baseline release every two years, starting in spring of 2027, and alternating years will be considered as edge or development versions without LTS guarantees. This will allow us a free hand to keep bringing you more experimental features while still providing a stable baseline for more conservative users. Try JRuby Today! Users can install JRuby 10.1 through the usual means, and we encourage you to give it a try and let us know how it handles your applications and libraries. We have much more planned for the 10.1 updates and hope to move fast and push the edges of what’s possible with Ruby on the JVM. 59 Issues and PRs resolved for 10.1.0.0 #8716 dup a lot #8747 Launcher script chokes on GraalVM release file #9055 Array experiments #9066 Move Set to a core class [Ruby 4.0] #9069 Implement Ruby 4.0 support #9074 Avoid array allocation for *nil, by not calling nil.to_a #9075 Selective inspect of instance variables #9077 No numbered parameters in Binding#local_variables #9078 Add Math.{expm1,log1p} #9079 Align IO.select timeout logic with CRuby #9083 Implement missing raise features #9085 Specialize Fixnum #9090 Reduce the size of the varTableStamp #9091 Shrink varTableStamp #9095 Eliminate global flags #9096 Cache object ID and hash in low-count immediate values #9102 Add a fast path for hashing numerics and immediates #9113 Performance regression in RubyHash for large hashes (~8-10x slower than JRuby 1.7) #9119 Implement fast-path builtin method flags #9123 Fast builtin validation experiment #9124 Reduce load factor for Hash from 5 to 2 #9136 Marshal.load 2.9x regression vs JRuby 1.7 #9143 Cache both hash forms for FString #9148 Allow logical ops to begin on beginning of next line #9154 Mask internal Ruby backtrace elements like native masking #9156 Small marshal fixes #9174 Add complete thread-local builtin tracking system (Issues #9116, #9119) #9180 Implement implicit param logic for Binding #9181 Fix Data subclasses with ivars #9202 A Data object should be frozen even if it has no members #9209 Add ruby_bug for spec of Data with no members being frozen #9210 Remove Process::Status#& and Process::Status#» #9233 Load boot scripts from classloader URI #9235 Use capitalized Class and Module for frozen errors #9236 [Properly handle keywords in {Kernel Thread Fiber}#raise]#9236 #9238 Cloned class’s singleton class has too many classes in hierarchy #9239 Fix for cloned class singleton class has too many ancestors #9250 Update Unicode to Version 17.0.0 and Emoji Version 17.0 #9262 Abstract RubyHash #9263 Implicit param fixes for proc and binding #9266 Endless method with modifier method does not parse #9273 Cleanup recursion guards #9277 Use live counts for event hook stats #9279 Move fiddle back to default temporarily for resolv #9286 Proposing RISC-V architecture support #9287 Add riscv64 platform support #9298 Ensure we have a classloader to boot from #9299 Delete methods deprecated prior to JRuby 9.4.0.0 #9300 Replace now-deprecated wait_timeout #9313 Time being returned as ASCII-8BIT encoding in JRuby 10 #9318 Process.detach raises ArrayIndexOutOfBoundsException if called from a thread that is not the main thread #9325 [fix] NullPointerException from closed IO #9362 Allow endless methods to be an argument to methods like visibity #9365 Marshal load perf #9371 Deprecate all public RubyString constructors #9374 Integrate WASM-based Prism parser into standard build #9383 Disable IntFixnum #9384 Set.new(a_set) throws LocalJumpError #9385 Fixes #9384. Set.new(a_set) throws LocalJumpError -- blog: http://blog.enebo.com twitter: tom_enebo mail: tom.enebo(a)gmail.com

1 0

[ruby-talk:444805] How to build a platform of agents, in Ruby
by Robert 20 Apr '26

20 Apr '26

Hi I wrote a post: https://0x1eef.github.io/posts/how-to-build-a-platform-of-agents/ It documents a pattern where a user can have many agents, and each agent can be specialized in a domain. An agent can have its own tools, and even its own skills. An agent can execute tools and skills with multiple different concurrency strategies: fiber, thread, async-task, ractor. This pattern is integrated with both Sequel, and ActiveRecord. It is for building serious AI applications. Robert

1 0

[ruby-talk:444804] [ANN] minitest 6.0.5 Released
by Ryan Davis 20 Apr '26

20 Apr '26

minitest version 6.0.5 has been released! * home: <https://minite.st/> * code: <https://github.com/minitest/minitest> * bugs: <https://github.com/minitest/minitest/issues> * rdoc: <https://docs.seattlerb.org/minitest> * clog: <https://github.com/minitest/minitest/blob/master/History.rdoc> minitest provides a complete suite of testing facilities supporting TDD, BDD, and benchmarking. "I had a class with Jim Weirich on testing last week and we were allowed to choose our testing frameworks. Kirk Haines and I were paired up and we cracked open the code for a few test frameworks... I MUST say that minitest is *very* readable / understandable compared to the 'other two' options we looked at. Nicely done and thank you for helping us keep our mental sanity." -- Wayne E. Seguin minitest/test is a small and incredibly fast unit testing framework. It provides a rich set of assertions to make your tests clean and readable. minitest/spec is a functionally complete spec engine. It hooks onto minitest/test and seamlessly bridges test assertions over to spec expectations. minitest/benchmark is an awesome way to assert the performance of your algorithms in a repeatable manner. Now you can assert that your newb co-worker doesn't replace your linear algorithm with an exponential one! minitest/pride shows pride in testing and adds coloring to your test output. I guess it is an example of how to write IO pipes too. :P minitest/test is meant to have a clean implementation for language implementors that need a minimal set of methods to bootstrap a working test suite. For example, there is no magic involved for test-case discovery. "Again, I can't praise enough the idea of a testing/specing framework that I can actually read in full in one sitting!" -- Piotr Szotkowski Comparing to rspec: rspec is a testing DSL. minitest is ruby. -- Adam Hawkins, "Bow Before MiniTest" minitest doesn't reinvent anything that ruby already provides, like: classes, modules, inheritance, methods. This means you only have to learn ruby to use minitest and all of your regular OO practices like extract-method refactorings still apply. ## Features/Problems: * minitest/autorun - the easy and explicit way to run all your tests. * minitest/test - a very fast, simple, and clean test system. * minitest/spec - a very fast, simple, and clean spec system. * minitest/benchmark - an awesome way to assert your algorithm's performance. * minitest/pride - show your pride in testing! * minitest/test_task - a full-featured and clean rake task generator. * Incredibly small and fast runner, but no bells and whistles. * Written by squishy human beings. Software can never be perfect. We will all eventually die. Changes: ### 6.0.5 / 2026-04-20 * 2 bug fixes: * Avoid circular requires in lib/minitest/server_plugin.rb. * Raise TypeError if assert_raises is passed anything but modules/classes.

1 0

[ruby-talk:444803] [ANN] debride 1.15.2 Released
by Ryan Davis 20 Apr '26

20 Apr '26

debride version 1.15.2 has been released! * home: <https://github.com/seattlerb/debride> * rdoc: <http://docs.seattlerb.org/debride> Analyze code for potentially uncalled / dead methods, now with auto-removal. ## Features/Problems: * Static analysis of code. Can be easily hooked up to a CI. * As with all static analysis tools of dynamic languages, can't be 100%. * Whitelisting known good methods by name or regexp. * Use --rails for Rails-specific domain knowledge. * Use debride_rm to brazenly remove all unused methods. BE CAREFUL. * Use `debride_rails_whitelist` to generate an emperical whitelist from logs. * Uses path_expander, so you can use: * dir_arg -- expand a directory automatically * @file_of_args -- persist arguments in a file * -path_to_subtract -- ignore intersecting subsets of files/directories Changes: ### 1.15.2 / 2026-04-20 * 3 bug fixes: * Fix #process_hash when there's a kwsplat in the middle. (carlos-musetti) * Rescue StandardError in .run to handle wider gamut of errors gracefully. * Add #process override to handle errors as well.

1 0

[ruby-talk:444801] [ANN] llm.rb v4.17.0 released
by Robert 18 Apr '26

18 Apr '26

Hi == Links - https://github.com/llmrb/llm.rb#readme - https://0x1eef.github.io/x/llm.rb/file.CHANGELOG.html == Release llm.rb has been released. Notable features in this release are Sequel and ActiveRecord support. Turn any ActiveRecord model into an LLM-capable model. It only requires one table. Same is true for Sequel. See CHANGELOG for details. Robert

1 1

[ruby-talk:444800] New tutorial series for wxRuby3
by martin＠corino.nl 16 Apr '26

16 Apr '26

Want to see what wxRuby3 is all about? Checkout this new (and excellent) tutorial at https://www.rehali.com/wxruby.

1 0