Trait mz_repr::stats::PersistSourceDataStats

pub trait PersistSourceDataStats: Debug {
    // Provided methods
    fn len(&self) -> Option<usize> { ... }
    fn err_count(&self) -> Option<usize> { ... }
    fn col_min<'a>(
        &'a self,
        _idx: usize,
        _arena: &'a RowArena
    ) -> Option<Datum<'a>> { ... }
    fn col_max<'a>(
        &'a self,
        _idx: usize,
        _arena: &'a RowArena
    ) -> Option<Datum<'a>> { ... }
    fn col_null_count(&self, _idx: usize) -> Option<usize> { ... }
    fn row_min(&self, _row: &mut Row) -> Option<usize> { ... }
    fn row_max(&self, _row: &mut Row) -> Option<usize> { ... }
}

Provides access to statistics about SourceData stored in a Persist part (S3 data blob).

Statistics are best-effort, and individual stats may be omitted at any time, e.g. if persist cannot determine them accurately, if the values are too large to store in Consensus, if the statistics data is larger than the part, if we wrote the data before we started collecting statistics, etc.

Provided Methods

fn len(&self) -> Option<usize>

The number of updates (Rows + errors) in the part.

fn err_count(&self) -> Option<usize>

The number of errors in the part.

fn col_min<'a>(&'a self, _idx: usize, _arena: &'a RowArena) -> Option<Datum<'a>>

The part’s minimum value for the named column, if available. A return value of None indicates that Persist did not / was not able to calculate a minimum for this column.

fn col_max<'a>(&'a self, _idx: usize, _arena: &'a RowArena) -> Option<Datum<'a>>

(ditto above, but for the maximum column value)

fn col_null_count(&self, _idx: usize) -> Option<usize>

The part’s null count for the named column, if available. A return value of None indicates that Persist did not / was not able to calculate the null count for this column.

fn row_min(&self, _row: &mut Row) -> Option<usize>

A prefix of column values for the minimum Row in the part. A return of None indicates that Persist did not / was not able to calculate the minimum row. A Some(usize) indicates how many columns are in the prefix. The prefix may be less than the full row if persist cannot determine/store an individual column, for the same reasons that col_min/col_max may omit values.

TODO: If persist adds more “indexes” than the “primary” one (the order of columns returned by Schema), we’ll want to generalize this to support other subsets of columns.

fn row_max(&self, _row: &mut Row) -> Option<usize>

(ditto above, but for the maximum row)

Implementors