Enum mentat::FindSpec

pub enum FindSpec {
    FindRel(Vec<Element>),
    FindColl(Element),
    FindTuple(Vec<Element>),
    FindScalar(Element),
}

A definition of the first part of a find query: the [:find ?foo ?bar…] bit.

There are four different kinds of find specs, allowing you to query for a single value, a collection of values from different entities, a single tuple (relation), or a collection of tuples.

Examples:

  let elements = vec![
    Element::Variable(Variable::from_valid_name("?foo")),
    Element::Variable(Variable::from_valid_name("?bar")),
  ];
  let rel = FindSpec::FindRel(elements);

  if let FindSpec::FindRel(elements) = rel {
    assert_eq!(2, elements.len());
  }

Variants

FindRel(Vec<Element>)

Returns an array of arrays, represented as a single array with length a multiple of width.

FindColl(Element)

Returns an array of scalars, usually homogeneous. This is equivalent to mapping over the results of a FindRel, returning the first value of each.

FindTuple(Vec<Element>)

Returns a single tuple: a heterogeneous array of scalars. Equivalent to taking the first result from a FindRel.

FindScalar(Element)

Returns a single scalar value. Equivalent to taking the first result from a FindColl.

Methods

impl FindSpec

Returns true if the provided FindSpec returns at most one result.

pub fn is_unit_limited(&self) -> bool

pub fn expected_column_count(&self) -> usize

pub fn requires_distinct(&self) -> bool

Returns true if the provided FindSpec cares about distinct results.

I use the words "cares about" because find is generally defined in terms of producing distinct results at the Datalog level.

Two of the find specs (scalar and tuple) produce only a single result. Those don't need to be run with SELECT DISTINCT, because we're only consuming a single result. Those queries will be run with LIMIT 1.

Additionally, some projections cannot produce duplicate results: [:find (max ?x) …], for example.

This function gives us the hook to add that logic when we're ready.

Beyond this, DISTINCT is not always needed. For example, in some kinds of accumulation or sampling projections we might not need to do it at the SQL level because we're consuming into a dupe-eliminating data structure like a Set, or we know that a particular query cannot produce duplicate results.

Important traits for Box<R>

impl<R> Read for Box<R> where
    R: Read + ?Sized, impl<W> Write for Box<W> where
    W: Write + ?Sized, impl<I> Iterator for Box<I> where
    I: Iterator + ?Sized,  type Item = <I as Iterator>::Item;

pub fn columns(&'s self) -> Box<Iterator<Item = &'s Element> + 's>

Trait Implementations

impl Eq for FindSpec

impl PartialEq<FindSpec> for FindSpec

fn eq(&self, __arg_0: &FindSpec) -> bool

This method tests for self and other values to be equal, and is used by ==.

fn ne(&self, __arg_0: &FindSpec) -> bool

This method tests for !=.

impl Debug for FindSpec

fn fmt(&self, __arg_0: &mut Formatter) -> Result<(), Error>

Formats the value using the given formatter.