This document aims to answer the questions:
What needs does Dallycot fill?
Why would I use Dallycot?
How does everything broadly fit together?
Dallycot is designed to work with the asynchronous nature of the web. When running a program in Dallycot, the web is your memory, providing data and code storage.
N.B.: Almost everything about Dallycot is subject to change. For now, Dallycot provides read-only access to information.
Dallycot provides a custom functional language designed to think the way linked data thinks. See the W3C Linked Data Platform specification for an example of the kind of data services Dallycot will target. Dallycot prefers JSON-LD when communicating with services.
Dallycot is not a query language. Projects like Marmotta provide linked data query languages.
These examples probably don't work yet, but indicate what we're working towards.
The following snippet describes walking the DBPedia graph from the resource representing The Lord of the Rings to anything that shares a common subject with that book, and extracting the label (imagine how the directional edges point from node to node in the data graph).
<http://dbpedia.org/data/The_Lord_of_the_Rings.rdf> -> :dcterms:subject <- :dcterms:subject -> :rdf:label
This uses simple recursion to calculate the greatest common divisor.
gcd := ( gcd_f(self, a, b) :> ( (a = 0) : b (b = 0) : a (a > b) : self(self, a mod b, b) ( ) : self(self, a, b mod a) ); gcd_f(gcd_f, ___) )
For now, a symbol such as
gcd_f isn't defined until after the expression has finished running, so recusion requires a form of the Y-combinator, such as provided by Dallycot::Library::Core::Functions.
Because anything might result in retrieving information from the web, Dallycot makes extensive use of promises. Promises are great for managing asynchronous execution, but introduce overhead that makes programs slower if all of the information is local to the processor.
Dallycot is not designed to be good at:
Some programs take a while. When coupled with ad hoc information retrieval over the web, programs can seem to slow to a crawl. This is the nature of linked data in general, not Dallycot. If you already know exactly which data you will need, and how the data fits together, then consider processing it locally with a SPARQL service or a general purpose programming language designed to work exclusively with local data.
Use programs such as Matlab, Mathematica, or R. Consider the scientific computing libraries available to Perl, Python, or Ruby.