Gibbler: Git-like Hashes and History for Ruby Objects(rubyinside.com)

11 pointsby luccastera17 lat temu5 comments

Gibbler: Git-like Hashes and History for Ruby Objects

http://www.rubyinside.com/gibbler-git-like-hashes-and-history-for-ruby-objects-1980.html

1 comments

We use this pattern for immutable objects in our Perl applications, although I think our implementation is cleaner:

http://github.com/nothingmuch/kiokudb/blob/96f9a380c715b7aa1...

delano17 lat temu

Interesting! I'm not familiar with KiokuDB or Moose so I hope you don't mind a couple questions:

- what is the return value of $self->data in the example you link to?

- do you store the digest or calculate it each time?

- where can I find an example of the digests being used? I looked around the repo a bit but couldn't find one.

jrockway17 lat temu

$self->data returns the value of the data attribute. The class is responsible for generating the identity of instances.

The digest is only calculated when the object is stored. When it is retrieved and stored again (a no-op, since it's immutable), the ID is remembered and not recalculated. (Creating an identical new object will result in the digest being calculated again, of course.)

The basic pattern for using this, once you've written the class:

    my $foo = Class->new( ... );
    my $id = $db->store($foo); # 29837a387bffe...
    my $foo2 = $db->lookup('29837a387bffe...');

Then this is true (the linker keeps note of the in-memory object IDs):

    refaddr $foo == refaddr $foo2

delano17 lat temu

Thanks for the detailed answer. The reason I ask is b/c there are some non-trivial problems which are similar regardless of the language or implementation. For example, what would the $self->data return value look like for a non-scalar? Does it recursively process a hash?

When you say the linker keeps note of the object IDs, does this mean the digests are stored outside of the objects themselves?

jrockway17 lat temu

Yes, the class is not responsible for managing any aspect of its storage. It can provide hints, however, and that's what the "ID" and "Immutable" roles do.

It is worth pointing out that no automatic hashing is occurring. You write a function that is invoked on an instance of the class that determines the unique ID. You decide the content-addressibility of your object by choosing the address based on the content.

(If you just want a unique ID per object, KiokuDB assigns a UUID by default as the ID.)