Reducing Memory Usage in Ruby

https://tenderlovemaking.com/2018/01/23/reducing-memory-usage-in-ruby.html

222 56 9 years ago

sam0x1710• 9 years ago

Or just use Crystal https://crystal-lang.org XD

igravious2• 9 years ago

Crystal is not a replacement for Ruby. In the same way that many Algol-like languages are not a straight swap for each other. Don't get me wrong, Crystal is a nice idea but say goodbye to all that yummy Ruby meta-programming which is one of the key things what makes the language so appealing in the first place.

eropple9• 9 years ago

Exactly. As a Ruby developer, Crystal makes me itch. It jettisons exactly what Ruby is valuable to me for: quickly and dynamically building fluent-feeling libraries and tools. It's fast, but I don't need fast--if I don't need Ruby's metaprogramming, I have other, better, established languages to use for those tasks. (Kotlin, C#, Rust, whatever.)

Elixir is kind-of in the same boat for me. I think it's neat and has its place, and BEAM is really cool, but it doesn't scratch the Ruby itch for me. I just don't care about it looking vaguely like Ruby if it's not actually gonna be Ruby.

sam0x1710• 9 years ago

Take a look at macros in crystal and I promise you 99% of the time you will be able to do what you did in Ruby. Practically everything people do that is "necessarily dynamic" could be done with macros at compile-time. You CAN do things like define the behavior for when a method name is unmatched on an Object or (at compile time) do some string processing and define a bunch of methods based on the results.

pankajdoharey• 8 years ago

Meta-programming in Crystal exists through Macros.

igravious2• 9 years ago

What I'd like to be able to do is load ActiveRecord (or any other v. heavy lib) asynchronously in non web-based apps so that I can paint the UI quicker. By the time the user will do/need anything db-related a proxy object should be swapped out for an actual ActiveRecord object. Using Rails' ORM is so nice and ergonomic but it kills start-up time with all the file requires, Bootsnap helps a bit but I'd prefer not to have to use it as it is yet one other moving part to consider.

AznHisoka• 9 years ago

I am actually facing a similar problem. I also find that Rails just creates a db connection everytime I do Model.new even though I have not yet saved it.

amichal• 8 years ago

This is because active record models learn their attributes by querying the database scheme. I haven’t tried it but there appears to be a way to cache this information in the local file system which would likely avoid the early dB connect if you really needed to do that. Also avoids querying the scheme on processs boot entirely if you are willing to trust the cached scheme matches the. database scheme. See https://github.com/rails/rails/issues/24133 for discussion

stouset• 9 years ago

I don’t quite understand why this is a problem. You pay the ActiveRecord boot time once, at app start, and then it’s no longer an issue.

ad_hominem• 9 years ago

This is orthogonal to tenderlove's GC work, but if you're using MRI and want an easy memory usage/performance win, swap out the standard glibc malloc for jemalloc: https://www.levups.com/en/blog/2017/optimize_ruby_memory_usa...

johncolanduoni• 9 years ago

Kind of a tangent, but does anyone know why glibc’s allocator loses to jemalloc on common workloads? I would expect that would be something worth rectifying unless other requirements tie their hands.

__s• 9 years ago

Quite a few complaints about jemalloc from Rust, where people wonder about higher than necessary memory usage

jemalloc likely wins more for the kinds of allocations you find in scripting language runtimes, which also happens to be similar to a browser given DOM & JS

masklinn• 8 years ago

Kind of a tangent, but does anyone know why glibc’s allocator loses to jemalloc on common workloads?

Jemalloc implements more complex strategies (per-thread caches learned from tcmalloc, preallocated arenas, …).

It tends to have a somewhat higher memory use than simpler allocators, but have faster throughput and less fragmentation (the latter is why Firefox switched to jemalloc by default IIRC).

I would expect that would be something worth rectifying unless other requirements tie their hands.

Why? If they consider ptmalloc good enough for their purpose, replacing it would be very low priority, and they may value e.g. code simplicity higher.

Incidentally, there was talk at one point to replace glibc's ptmalloc: https://sourceware.org/ml/libc-alpha/2014-10/msg00419.html

kawsper• 9 years ago

I am so happy that Github have hired Tenderlove to go nuts with these type of projects. His talks are always extremely fun and interesting.

I run Ruby with jemalloc, and I know that samsaffron have tried to include it within the Ruby releases (like Redis does), but progress seems to have been stalled https://bugs.ruby-lang.org/issues/9113

farnsworthy2• 8 years ago

I thought he was nuts already. (Smiley thing here.)

hirundo• 8 years ago

Ruby seems to have a footprint on Hacker News disproportionate to its general popularity and TIOBE index. Why?

berfarah• 8 years ago

Ruby has one of the better communities among programming languages - disproportionate to its popularity. I wish more languages had that quality of discourse.

lstamour• 8 years ago

Rails, and possibly 37signals, was very influential to many folks about a decade ago, especially in the startup community, where Rails was seen as especially useful for prototyping or rapid, test-driven web app development. https://www.quora.com/Why-do-so-many-startups-use-Ruby-on-Ra...

It’s popularity at this may have diminished with the rise of JavaScript, Go and alternatives but many folks still like and use Ruby, or at least would like to see it continue to improve and thrive.

farnsworthy2• 8 years ago

From a different perspective, having worked on numerous Ruby projects for years across the private (startups to Fortune 500), nonprofit, and government sectors, spanning many locations with worldwide team members, I don’t see where Ruby/Rails isn't at this point—it’s seemingly all-pervasive.

Similar to general hiring/salary surveys, I’m not sure that the referenced index is accurate.

abhiyerra• 8 years ago

Ruby and Rails have been around for about 10 years at this point. It is stable and for most websites it just works (up to a certain size). It is not as exciting as JS is now but people are using it to get things done since it is relatively mature and used a lot in startups. Likely why you see it referenced a lot here.

mwaitjmp• 8 years ago

Not only due to the popularity of Rails, Ruby has become popular for sysadmins as config management solutions such as puppet and chef are built on top of it. The resulting syntax is very ruby like, and in some cases you are even required to write a few ruby blocks here and there.

It also has an amazing testing framework in the form of rspec, which can be used for much more than just testing Ruby code.

ngsayjoe2• 8 years ago

Yeah, and i use Selenium with Ruby as well.

dragonwriter2• 8 years ago

Ruby seems to have a footprint on Hacker News disproportionate to its general popularity and TIOBE index.

TIOBE index is a junk metric (I'm surprised so many technical people seem to buy into it; it's the worst kind of easy to get a precise number, hard to demonstrate that that number has any more than a very distant relationship to anything meaningful, kind of metric that, IME, tech people mock bad managers for buying into.)

That said, Ruby’s got (via Rails and other web frameworks, and via various testing and admin tools) a very strong footprint in use even if it's no longer the hype-of-the-month.

flukus• 8 years ago

HN is not at all representative of the wider programming community. HN attracts a lot of people working for trendy startup companies and is very silicon valley centric, but that is a tiny portion of the programmers world wide who are much more likely to be working on a java project for a generic business than using the latest fad.

TIOBE has it's issues but as a rough guide to what the industry is actually using I think it's in the ballpark.

look_lookatme3• 8 years ago

Because Ruby has a footprint in the startup community that is disproportionate to its general popularity and TIOBE index.

sandGorgon• 8 years ago

What I'm really keen to see is truffleruby receiving a lot of community support.

The underlying JVM and the Truffle/Graal/SubstrateVM is the product of millions of man hours of research - and for the very first time, you have a compiler framework on top of that.

The Truffle framework also brings in a lot of support that all languages get for free, like zero overhead profiling https://twitter.com/nirvdrum/status/948333404122214401

Truffleruby is already in the running in benchmarks, and it will be awesome to see how far it can be pushed . http://nirvdrum.com/2017/02/15/truffleruby-on-the-substrate-...

ksec• 8 years ago

The major obstacle right now is C extension, which I think Sulong [1] is suppose to solve.

It seems right now, TruffleRuby 99% of Ruby code without much problem. C Extension may be the key to wide spread adoption.

[1] https://github.com/graalvm/sulong

crb002• 8 years ago

Is there any effort to make WASM run under Ruby? Seems like easy work to get high performance without bespoke work on Ruby's VM.

Reducing Memory Usage in Ruby

Related Stories