Enum Provenance 

pub enum Provenance {
    RowCount,
    ZoneMap,
    NullCount,
    Sketch,
    FrequencySynopsis,
    Quantiles,
    Dictionary,
    Sortedness,
    Distinctness,
    LinkHeader,
    DegreeDistribution,
    Sample,
    Default,
    Observed,
    Propagation,
}

What produced an answer.

Printed by EXPLAIN next to the class, so these are the words a person reads when a plan went wrong. It names the source rather than the value, and it is carried by exact answers as well as by guesses, because an exact count out of a catalog and an exact join cardinality out of a link header want different follow up questions when one of them turns out to be stale.

The list is spec/stats/02-the-catalogue.md section 2.1.1’s fourteen plus Self::Propagation, which the specification does not name because it is not a source of data. It is what a number derived from two others says about itself, and leaving it out would mean a combination inherited the provenance of whichever operand happened to be on the left.

Variants

RowCount

A maintained row count. Exact.

ZoneMap

A minimum and a maximum, either answered from or interpolated between.

NullCount

A maintained null count. Exact.

Sketch

A distinct-count sketch.

Carried by both classes, because a bottom-k sketch that has not filled up is holding every hash it was given and the count it gives back is the count. So this arrives as Exact for a column with few enough distinct values and Estimated for one with more, and the class is where that difference is written rather than here.

FrequencySynopsis

A frequency synopsis, with or without its certificate discharged.

Quantiles

A quantile summary.

Dictionary

A dictionary’s size, standing in for or answering a distinct count.

Sortedness

A persisted sortedness flag.

Distinctness

A persisted distinctness flag.

LinkHeader

The four numbers in a relationship’s link header, or a cardinality derived from them. Exact, and the most valuable kind of exact there is, because join cardinality is where every cost model in the literature goes wrong by orders of magnitude.

DegreeDistribution

A relationship’s degree distribution.

Sample

The stored sample.

Default

A constant in the source. The weakest answer that is not Unknown, and the one worth searching an EXPLAIN for, because it means nobody had a number at that node at all.

Observed

Something a previous execution measured, out of the observation log. Kept apart from every other variant here on purpose: the rest are facts about the file and this one is a fact about history.

Propagation

A rule applied over two other answers. Not a source of data, and the honest thing to say about a number that was derived rather than read.

Implementations

impl Provenance

pub const fn name(self) -> &'static str

The word EXPLAIN prints.

pub const fn is_observed(self) -> bool

Whether this is a fact about a previous execution rather than about the file.

The one question a consumer of spec/stats/06-the-reward.md’s tier 1 corrections has to be able to ask, because an observation is Exact only for the generation it was taken on.

Trait Implementations

impl Clone for Provenance

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for Provenance

impl Debug for Provenance

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Display for Provenance

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for Provenance

impl Hash for Provenance

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for Provenance

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for Provenance