Struct AsyncStreamingProfiler 

pub struct AsyncStreamingProfiler { /* private fields */ }

Async streaming profiler that accepts AsyncDataSource instead of file paths.

Uses a bounded channel between a blocking CSV reader task and an async processing loop to provide natural backpressure: when the processor falls behind, the reader pauses, which propagates TCP pressure back to the source.

Architecture

AsyncDataSource ──► [spawn_blocking: csv::Reader] ──► bounded mpsc ──► [processor loop]
                      SyncIoBridge + csv crate           capacity N      StreamingColumnCollection
                      (RFC 4180 compliant)                               ──► ProfileReport

Implementations

impl AsyncStreamingProfiler

pub fn new() -> AsyncStreamingProfiler

pub fn chunk_size(self, chunk_size: ChunkSize) -> AsyncStreamingProfiler

pub fn sampling(self, strategy: SamplingStrategy) -> AsyncStreamingProfiler

pub fn memory_limit_mb(self, limit: usize) -> AsyncStreamingProfiler

pub fn channel_capacity(self, capacity: usize) -> AsyncStreamingProfiler

pub fn progress( self, sink: ProgressSink, interval: Duration, ) -> AsyncStreamingProfiler

pub fn stop_condition(self, condition: StopCondition) -> AsyncStreamingProfiler

pub fn analysis_options( self, options: AnalysisOptions, ) -> AsyncStreamingProfiler

Set the whole analysis selection at once.

Preferred over the individual setters: the async pipeline must apply the same selection the synchronous engines do, and passing it as one value is what keeps a new option from reaching only some of them.

pub fn quality_dimensions( self, dims: Vec<QualityDimension>, ) -> AsyncStreamingProfiler

pub fn metric_packs(self, packs: Vec<MetricPack>) -> AsyncStreamingProfiler

pub fn locale(self, locale: Locale) -> AsyncStreamingProfiler

Set the locale used to rank detected patterns.

pub fn semantic_hints(self, hints: SemanticHints) -> AsyncStreamingProfiler

pub fn json_error_policy( self, policy: JsonErrorPolicy, ) -> AsyncStreamingProfiler

Set how malformed JSON/JSONL records are handled (default: skip and count).

pub fn csv_flexible(self, flexible: bool) -> AsyncStreamingProfiler

Set how CSV records whose field count differs from the header are handled.

true (default) recovers them — extra fields dropped, missing fields padded to null — and counts every one in ExecutionMetadata::ragged_row_count, matching the incremental file path. false rejects the first such record with a CSV parsing error.

pub fn csv_delimiter(self, delimiter: u8) -> AsyncStreamingProfiler

Set the CSV field delimiter.

When unset, it is detected from the head of the stream the same way the file path detects it, so a semicolon- or tab-separated source profiles identically over either transport.

pub async fn analyze_stream( &self, source: impl AsyncDataSource, ) -> Result<ProfileReport, DataProfilerError>

Profile data from an async source, returning a ProfileReport.

Supports CSV, JSON, and JSONL formats. Parquet async support is tracked separately.

Trait Implementations

impl Default for AsyncStreamingProfiler

fn default() -> AsyncStreamingProfiler

Returns the “default value” for a type.