Hamster - Efficient, Immutable, Thread-Safe Collection classes for Ruby | Hacker News Reader

Hamster - Efficient, Immutable, Thread-Safe Collection classes for Ruby

github.com

39 pointsjustinweiss30 comments

Is it just me or is anyone else annoyed about Ruby people calling their hash table based dictionaries a "hash"? As a non-Ruby programmer, I always get confused when people talk about hashes and you can never be quite sure if it's a dictionary or an integer hash value.

They use the same name in Perl. I presume that's where Ruby got it from.

from http://en.wikipedia.org/wiki/Hash

Hash may refer to: http://en.wikipedia.org/wiki/Hash_table

i'm sure it's not just you though, people get annoyed over whole bunch of things

I first started hearing the word used that way about at the same time as Ruby started becoming more popular.

Not to be a pedant, but wouldn't thread-safe automatically follow immutable?

If a structure is immutable, how would you be able to fuck it up, even if you abuse it via threading? I would argue the relation is so strong, that something which isn't thread-safe cannot be immutable. But there may ofcourse be something which I'm missing.

Different languages and platforms handling different key concepts differently etc.

Is there some properties of Ruby which makes this possible, or gives you a need for a double-guarantee, or is the declaration here simply redundant?

Consider a balanced tree that only exposes an immutable interface but uses destructive updates internally when rebalancing. In principle it would be immutable, though it wouldn't necessarily be threadsafe without additional work.

From my point of view, that would not be something I considered a immutable data-structure.

But I guess at this point we are starting to argue technicalities and it's getting quite outside the scope of this submission.

I'll just take a mental note with regard to immutable evidently being a lossy concept ;)

Immutable simply means there is no way to change the data once you are given the object. That does not mean the impl is a clone.

In any case, ruby is really missing something as good as this. Very nice stuff. Wish that it did not require "new" syntax compared to ruby's nice [ ] and { }. But whatever. This is a problem that will exist on any collection innovation.

Hmm, I assumed this would have to be written in C so that it could use multiple threads in parallel in MRI, but it seems not. Then I remembered concurrent != parallel.

You forget that MRI isn't the only Ruby implementation. JRuby has no GIL.

    def dup
      self
    end

Weehee, immutability is great at some things.

How many people do multi threading in ruby? I'm assuming this is mostly for jruby?

Great question. It's starting to gain a more mainstream following. Here are a few resources:

* https://github.com/celluloid/celluloid

* https://github.com/mperham/sidekiq (built on Celluloid)

* https://github.com/puma/puma

* http://tenderlovemaking.com/2012/06/18/removing-config-threa...

I use threading in MRI ruby quite often… maybe I'm the odd man out here.

Threading in MRI is still entirely appropriate for juggling IO. So that makes two of us.

I guess I jumped to quickly to the conclusion that more work would be done in Ruby. I suppose it works for IO. I like non-blocking libraries like em-http-request and event machine for these things.

If I wanted to do concurrent programs, I usually chose C :)

I use it with YARV all the time (usually with the "parallel" gem) to get a Nx speedup on easily parallelisable operations like processing a set of input files, sending HTTP requests to APIs using threadsafe cURL handles, and so on.

Another decent place to find out more about multi-threaded ruby or to post info about it is http://ruby.io/

Don't forget Rubinius! No GIL neither.

Rubinius 2 when it comes out.

I use it quite a bit, across all 4 major interpreters. Not used Hamster yet, though.

As of git HEAD at the time of this comment (5edce74):

1499 lines of Ruby code.

19 lines of comments.

Nope, nope, nope. I would never touch this library.

Good amount of tests, but the moment this library doesn't work as expected and you need to jump into the code, you're completely screwed.

Not usable.

You're screwed because you won't possibly be able to follow a well composed, well tested library with an extremely shallow class hierarchy and short, clear method definitions without the aid of comments?

I don't think you're giving yourself enough credit.

A quick glance at the source shows that the size of the program is a result of implementing the many methods one expects from collection classes in Ruby, not because the code is especially byzantine. In fact, I haven't seen a single instance of class inheritance; the author seems to have chosen to solely use a few well defined modules instead.

What part of the code isn't self-explanatory that a comment would be able to band-aid? It's all very clean. Comments for comments' sake are much worse than no comments at all.

didn't look at the code but if it's good code, why would you need more comments? Good code explains itself.

Do you have an example of good code that explains rationale or "why" without comments? I rarely come across code and can't figure out what it's doing, but I routinely come across code and can't figure out why it's doing something. And I haven't found a good way to deal with that outside of comments or some external documentation. I'd love to see examples of code that's able to overcome this.

Well done code has good variables, good method/function naming and small methods/functions. If you have those 3 parts, you won't need any comments because it's self explanatory.

Comments often tend to explain the same thing as the code exactly does and what you see w/o looking at the comment. Also, they tend to be forgotten - if someone changes the code, he/she often doesn't change the comment. As a result, you'll have an outdated comment which can confuse you.

I recommend "Clean Code"[1] for more info on this subject. Beside other chapters, there's a chapter about what's a good and what's a bad comment.

[1]: http://www.amazon.com/Clean-Code-Handbook-Software-Craftsman...

I've never come across self-documenting code that can explain "why," which is what I was seeking. The closest thing I've ever come across was a PhD student working on a system called Seurat, but even that was really metadata attached to the code (something akin to an SCM). I agree clean code can help with the "what" or the "how", but that's almost never the problem to begin with.

In any event, I was looking for solid examples, not a general discussion on whether comments are good or bad. We clearly are not going to agree on that point so we'll save the Internet another pointless argument :-)

Ok I might have misunderstood you ;) Well I'm not saying comments are generally bad. But I think a lot of comments can be replaced by better code :) Solid examples I can't really show you atm (I'm not yet there where I'd say from myself that I write the cleanest code ever) ;)

1.5 kloc of code is only a few hours to grok. Get over it.

Reply on news.ycombinator.com