Struct PromPlannerContext 

struct PromPlannerContext {

    start: Millisecond,
    end: Millisecond,
    interval: Millisecond,
    lookback_delta: Millisecond,
    stmt_start: Millisecond,
    stmt_end: Millisecond,
    table_name: Option<String>,
    time_index_column: Option<String>,
    field_columns: Vec<String>,
    tag_columns: Vec<String>,
    aggregation_field_labels: Vec<String>,
    use_tsid: bool,
    field_column_matcher: Option<Vec<Matcher>>,
    selector_matcher: Vec<Matcher>,
    schema_name: Option<String>,
    range: Option<Millisecond>,
    range_fold_offset: Option<Millisecond>,
}

Fields

start: Millisecond

end: Millisecond

interval: Millisecond

lookback_delta: Millisecond

stmt_start: Millisecond

Evaluation range of the whole statement, which @ start() and @ end() refer to.

Unlike Self::start and Self::end, these are never rewritten while planning, so a selector inside a subquery still resolves @ start() / @ end() against the statement.

stmt_end: Millisecond

table_name: Option<String>

time_index_column: Option<String>

field_columns: Vec<String>

tag_columns: Vec<String>

aggregation_field_labels: Vec<String>

by(...) labels of the aggregation that produced this operand that are not series tags of its input, i.e. value fields (or a label an inner aggregation already reported as one).

An aggregation reports every by(...) label it finds in its input schema among its tag columns, and a value field named there is a group key of the aggregate rather than a property of a series: its value varies between the samples of one series. Lowering a matcher on it into the scan would change which samples are selected (#9242), so matching_filters refuses to propagate such a matcher.

use_tsid: bool

Use metric engine internal series identifier column (__tsid) as series key.

This is enabled only when the underlying scan can provide __tsid (UInt64). The planner uses it internally (e.g. as the series key for [SeriesDivide]) and strips it from the final output.

field_column_matcher: Option<Vec<Matcher>>

The matcher for field columns __field__.

selector_matcher: Vec<Matcher>

The matcher for selectors (normal matchers).

schema_name: Option<String>

range: Option<Millisecond>

The range in millisecond of range selector. None if there is no range selector.

range_fold_offset: Option<Millisecond>

The offset in milliseconds the window of the last planned range selector is folded with, or None when no range selector has been planned since the last read.

Self::start and the sample timestamps are compared on the shifted timeline: a range payload carries sample_timestamp + offset, while the time index column of a folded row stays the evaluation timestamp of its step. A function that reads it right after its input plan is built, like predict_linear, consumes it instead of reading the state, so the offset cannot leak from one input to another; see PromPlanner::create_function_expr.

Implementations

impl PromPlannerContext

fn from_eval_stmt(stmt: &EvalStmt) -> Self

fn reset(&mut self)

Reset all planner states

fn reset_table_name_and_schema(&mut self)

Reset table name and schema to empty

fn has_le_tag(&self) -> bool

Check if le is present in tag columns

Trait Implementations

impl Clone for PromPlannerContext

fn clone(&self) -> PromPlannerContext

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PromPlannerContext

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

Formats the value using the given formatter.

impl Default for PromPlannerContext

fn default() -> PromPlannerContext

Returns the “default value” for a type.