Arnaud Berthomier @oz@ruby.social

Slowly spreading the word about Mastodon...😄

Testing out Ruby 3.1 with #yjit vs Ruby 3.0.

This shows a sharp decrease of CPU usage for a small Sinatra-based API.

Improvements in response times aren't as visible because the app is mostly IO bound. The app is small, so the memory penalty is light too. I like it.

Thanks Shopify. 👍

So, the solution is to not use this Ruby 2 feature anymore. I've taken to using pattern-matching instead.

Yes, it is more verbose in some cases. Not a big deal.

I'm porting 2.7 Ruby to Ruby 3, and I'm having trouble iterating an array of hashes with de-structuring keyword args.

Previously, this worked:

a = [{:foo=>1, :bar=>1}, {:foo=>2, :bar=>2}]
a.each { |foo:, bar:| puts "foo: #{foo}, bar: #{bar}" }

With #ruby3, how do you de-structure each item inline? 🤔

Oh. I managed to miss the RailsConf drama, until I opened the Ruby Weekly newsletter.

I'm sad to say that all the shits of the weeks had already been given at this point. 🤷

Are you WIP-ing son? This is (part of) the commit-logs from a pull-request I was reviewing today.

It made me smile. 😄

Nice take on Fibers in Ruby from #Polyphony author, Sharon Rosner.

https://noteflakes.com/articles/2021-10-20-explaining-ruby-fibers

Finally upgraded these API to run on Ruby 3. 👌
#ruby3

On the announcement post for Crystal 1.2: "On a bit of a sour note, we are demoting support for 32-bit x86 architectures, moving it from Tier 1 to Tier 2 (that is, /expected to build/). As a result, we will not be shipping i386 packages this time. The reason is that the specs are failing, likely due to the amount of memory they consume."

A little bit sad to read this. Of course, you could build the compiler, but I read it as a warning against Crystal where memory matters.

#crystallang

The HN comments about the CoC change for Ruby are really fun to read. So many privileged people taking offense as they're required to be nice.

https://news.ycombinator.com/item?id=28712821

I really want to use the nice dry gems but you can't nest their validators, and there's no way to add custom rules to their schemas (which you can nest 🥲). There are issues and PRs about it since 2019... 😖

I'm pretty sure I could replace the legacy validators with something simpler like Scrivener, because I don't want to reinvent this wheel. 🙃

#validation

During a code review, I found code where we were using `Enumerable#find` to choose a value from an array, against data ; building a new Regex on each loop. When your array is sorted `#bsearch` is orders of magnitude faster:

```
❯ ruby bm.rb
user system total real
justfind 25.743317 0.048932 25.792249 ( 25.791471)
bsearch 3.192201 0.003215 3.195416 ( 3.195469)
cache 0.285086 0.000304 0.285390 ( 0.285395)
```

Also, cache things when possible. 😅

I was looking at Babosa (https://github.com/norman/babosa), when I noticed the project is used "a lot" by python devs. 😆

This may not last long, but I find it nice that robots are using Github to auto-generate stuff ; after all Github's opened the box with their own "AI" bot. Still, probably just annoying marketing.

I really like having several "legs" to deploy new features or bug fixes progressively ; but having feature flags to target a few users would also help. 🤔

So yeah, code reviews are hard.

... then there's the dilemna of "I don't want to be a douche" VS "it's nice to have a clean code base to work on".

I'm annoyed when devs add helpers methods to a Ruby class named `AbstractSomething`, but I'm not sure it's actually wrong...

ruby was clearly not the bottle neck here. I guess you need to keep these fork/dump ops to a minimum if you want a fast redis.

I recently moved ±28M redis keys between two servers with a ruby script that had to update its source record after migration, and it got slow as fuck while #redis did bg-saves. :/