lab42_nested_hash

A nested hash view with dotted deep access à la I18n.t of Rails and with optional string interpolation. Typically YML loaded Hashes are used.

I developped NHash to scratch an itch that developped when creating presentations of hierarchical data with different aspects. E.g. different languages and different versions.

For that I wanted to combine the data and define strategies what to do when compound keys were not found. Typically data will be quite a big Hash instance coming from Yaml files, a Json webservice, or a document oriented database like MongoDB.

But let us start slowly:

    h = { "a" => 1,
          "b" => { "c" => 2, "d" => true } }

    nh = NHash.new h

Then get lets us access elements

    nh.get('a').assert == 1
    nh.get('b.c').assert == 2

or subhashes:

    nh.get('b').assert.kind_of? NHash

Please see the QED demos here for detailed explainations and specifications of the implemented features:

Symbol and String Keys are treated the same.

Prefixes and Suffixes can be predefined and used when a compound key starts and/or ends with a ..

Taking advantage of the Affixes features an affixed access can look in more than one affix. E.g. to fall back to English as the default translation language.

A more general concept of what to do if an access fails (raises a KeyError).

If values have ERB syntax templates as values, the special get! access can be used to interpolate them in various contexts, the NHash instance's binding being the default.

Subhashes and subarrays maintain important context information, e.g. for Interpolation.

If all else fails... We can check in other NHash instances. They are organised as a Depth First Search Tree.

This feature is particularly interesting for complex document composition out of a hierarchical data source tree (e.g. a set of yaml files, different for each new version/variation of the result).

The NHash constructor takes a Hash instance (defaulting to {} ) as param. This implies that much boiler blate code has to be written in order to bring in other data sources, as e.g. Yaml files.

The factories help to avoid this.

There are even helpers to create an NHash hierarchy from data sources.