pub struct SourceDiagnostics {Show 21 fields
pub source_read_strategy: String,
pub credential_mode: String,
pub lakehouse_cache_status: String,
pub lakehouse_cache_read_mode: String,
pub lakehouse_cache_fill_mode: String,
pub direct_device_read: bool,
pub file_count: usize,
pub ranged_file_count: usize,
pub known_bytes: Option<u64>,
pub selected_row_groups: Option<usize>,
pub selected_row_group_stats: Option<SelectedRowGroupStats>,
pub predicate_disposition: PredicateDisposition,
pub byte_range_disposition: ByteRangeDisposition,
pub partition_column_disposition: PartitionColumnDisposition,
pub zero_column_count_disposition: ZeroColumnCountDisposition,
pub scan_limit_disposition: ScanLimitDisposition,
pub schema_mapping_disposition: SchemaMappingDisposition,
pub read_column_count: usize,
pub projected_output_column_count: usize,
pub pure_filter_column_count: usize,
pub rejected_features: Vec<RejectedSourceFeature>,
}Expand description
Frontend-visible source capability and lowering diagnostics.
Diagnostics explain why a source can or cannot use the native direct device path. They are intentionally structural so adapters and tests can assert on stable fields without parsing displayed plans. Only local Parquet sources produce these diagnostics.
Fields§
§source_read_strategy: StringComma-separated high-level Parquet read strategies represented by this source.
credential_mode: StringComma-separated credential modes represented by this source.
lakehouse_cache_status: StringComma-separated local byte-cache statuses represented by this source.
lakehouse_cache_read_mode: StringComma-separated foreground lakehouse cache-read modes represented by this source.
lakehouse_cache_fill_mode: StringComma-separated complete-object lakehouse cache fill modes represented by this source.
direct_device_read: boolTrue when the source is proven to use a local path/cache reader that can satisfy the direct-device source-read boundary.
file_count: usizeNumber of files represented after any frontend coalescing.
ranged_file_count: usizeNumber of source file entries that carried byte ranges.
known_bytes: Option<u64>Total known bytes across represented files, or None when incomplete.
selected_row_groups: Option<usize>Total selected row groups, or None when row-group selection is unknown.
selected_row_group_stats: Option<SelectedRowGroupStats>Selected-versus-total row-group, byte, and row evidence.
predicate_disposition: PredicateDispositionHow pushed-down source predicates were represented.
byte_range_disposition: ByteRangeDispositionHow byte-range file entries were represented.
partition_column_disposition: PartitionColumnDispositionHow partition columns were represented.
zero_column_count_disposition: ZeroColumnCountDispositionWhether metadata-only zero-column count was required and represented.
scan_limit_disposition: ScanLimitDispositionWhether a source-level scan limit is absent or represented.
schema_mapping_disposition: SchemaMappingDispositionHow logical source columns map to physical or synthetic source columns.
read_column_count: usizeNumber of physical Parquet columns read by the native source.
projected_output_column_count: usizeNumber of logical source columns expected to remain after source-adjacent native filtering/projection.
pure_filter_column_count: usizeNumber of emitted source columns used only for source filtering.
rejected_features: Vec<RejectedSourceFeature>Stable feature codes explaining why direct device read was rejected.
Implementations§
Source§impl SourceDiagnostics
impl SourceDiagnostics
Sourcepub fn set_resolved_read_labels(&mut self, files: &[LocalParquetFile])
pub fn set_resolved_read_labels(&mut self, files: &[LocalParquetFile])
Sets the read strategy, credential mode, and byte-cache labels to the
ones a source over the resolved files reports, and direct device read
to whether those files prove it and no feature was rejected.