Struct DataFrame 

pub struct DataFrame { /* private fields */ }

A DataFrame-centric pipeline compiled into a lazy plan.

The public API stays in this crate’s own types. The current engine implementation is Polars, but callers do not need to depend on Polars types.

Implementations

impl DataFrame

pub fn from_dataset(ds: &DataSet) -> IngestionResult<Self>

Build a pipeline starting from an in-memory DataSet.

Note: this converts the dataset into a Polars DataFrame first. The transformations after that are planned lazily.

pub fn filter(self, predicate: Predicate) -> IngestionResult<Self>

Add a filter predicate.

pub fn multiply_f64(self, column: &str, factor: f64) -> IngestionResult<Self>

Multiply a Float64 column by a constant factor (nulls remain null).

pub fn add_f64(self, column: &str, delta: f64) -> IngestionResult<Self>

Add a constant Float64 value to a column (nulls remain null).

pub fn with_mul_f64( self, name: &str, source: &str, factor: f64, ) -> IngestionResult<Self>

Add a derived Float64 column: name = source * factor (nulls remain null).

pub fn with_add_f64( self, name: &str, source: &str, delta: f64, ) -> IngestionResult<Self>

Add a derived Float64 column: name = source + delta (nulls remain null).

pub fn select(self, columns: &[&str]) -> IngestionResult<Self>

Select a subset of columns (in the provided order).

pub fn rename(self, pairs: &[(&str, &str)]) -> IngestionResult<Self>

Rename columns.

This uses Polars’ rename(..., strict=true) behavior: all from columns must exist.

pub fn cast(self, column: &str, to: DataType) -> IngestionResult<Self>

Cast a column to a target type.

Note: cast errors (e.g. invalid parses) surface at collect() time.

pub fn cast_with_mode( self, column: &str, to: DataType, mode: CastMode, ) -> IngestionResult<Self>

Cast a column with an explicit mode (strict vs lossy).

pub fn drop(self, columns: &[&str]) -> IngestionResult<Self>

Drop columns by name.

pub fn fill_null(self, column: &str, value: Value) -> IngestionResult<Self>

Fill nulls in a column with a literal.

pub fn with_literal(self, name: &str, value: Value) -> IngestionResult<Self>

Add a derived column with a literal value.

pub fn group_by(self, keys: &[&str], aggs: &[Agg]) -> IngestionResult<Self>

Group rows by keys and compute aggregations.

pub fn join( self, other: DataFrame, left_on: &[&str], right_on: &[&str], how: JoinKind, ) -> IngestionResult<Self>

Join this pipeline with another DataFrame on key columns.

Note: join planning is infallible; missing-column errors surface at collect() time.

pub fn collect(self) -> IngestionResult<DataSet>

Collect the pipeline into an in-memory DataSet.

pub fn collect_with_schema(self, schema: &Schema) -> IngestionResult<DataSet>

Collect the pipeline into an in-memory DataSet, enforcing an explicit output schema.

pub fn reduce( self, column: &str, op: ReduceOp, ) -> IngestionResult<Option<Value>>

Reduce a column using a built-in ReduceOp (Polars-backed).

Returns None if column does not exist (aligned with [crate::processing::reduce]).

pub fn sum(self, column: &str) -> IngestionResult<Option<Value>>

Reduce a numeric column by summing values (nulls ignored; all-null -> null).

Returns None if column does not exist (aligned with processing::reduce).

pub fn feature_wise_mean_std( self, columns: &[&str], std_kind: VarianceKind, ) -> IngestionResult<Vec<(String, FeatureMeanStd)>>

Single Polars collect: for each column, mean and standard deviation (std_kind maps to Polars ddof). Columns are cast to Float64 first (aligned with scalar reduces).

Returns an error if any column name is missing from the lazy schema.

Trait Implementations

impl Clone for DataFrame

fn clone(&self) -> DataFrame

Returns a duplicate of the value.

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

Performs copy-assignment from source.