Methods for handling wide sources

Sam Ritchie edited this page Dec 6, 2011 · 2 revisions

In many real world examples, the number of tuple fields produced by some generator can grow to be quite cumbersome.

Often, a subquery will only want to deal with a small subset of the fields produced. Cascalog provides a few methods that make it quite easy to deal with these wide taps.

(Further examples of these solutions can be found in Nathan Marz's cascalog workshop repository.)


Take this example query, designed to add the first two fields produced by the generator and assign the result to ?sum:

    (<- [?a ?b ?sum]
        (+ ?a ?b :> ?sum)
        (generator ?a ?b ?c ?d ?e ?f ?g ?h ?i))

Not only does this query take up a lot of space; the user also has to think up dummy names for the unused variables. One solution Cascalog allows is the use of underscores for ignored variables:

    (<- [?a ?b ?sum]
        (+ ?a ?b :> ?sum)
        (generator ?a ?b _ _ _ _ _ _ _))

This works, but is still a bit clunky. Next choice: select-fields.

Fields by Name

If generator is some other subquery with named tuple fields, we can use select-fields to pull ?a and ?b without knowing their position.

    (<- [?a ?b ?sum]
        (+ ?a ?b :> ?sum)
        ((select-fields generator ["?a" "?b"]) ?a ?b))

Note that select-fields takes a generator and a sequence of (string representations of) variables, and returns a new generator that produces only those variables.

Fields by Position

Our final method of attacking wide taps involves the predicate operator :#>. (All other predicate operators are discussed in Predicate operators.) Here's an example:

    (<- [?a ?b ?sum]
        (+ ?a ?b :> ?sum)
        (generator :#> 9 {0 ?a, 1 ?b}))

:#> sits to the right of our generator, and accepts two items: the total number of fields produced by the generator, and a map of variable names keyed by position within the generator. Variables names don't have to match the names on the original output tap, as position is already specified.