Module tabular 

On-demand Polars conversions for reconstruction and benchmark records.

Domain types remain independent of Polars. These free functions allocate one column vector per output column only when a caller explicitly requests a table or writes a bundle.

Modules

parquet

Parquet/NPY result and benchmark bundle persistence. Parquet tables, NPY arrays, manifests, and lazy reopening for result bundles.

Structs

ReconstructionTables

All stable reconstruction tables materialized together on explicit request.

Constants

COMMON_RUN_COLUMNS

Stable identity/status columns shared by reconstruction and benchmark run tables.

Functions

algorithm_metrics_dataframe

Builds one row per algorithm-specific trace metric with stable identity columns.

benchmark_artifacts_dataframe

Builds one result-bundle artifact row per run with a supplied relative path.

benchmark_frames_dataframe

Builds one row per recorded acquisition-frame benchmark metric.

benchmark_metadata_dataframe

Builds long-form non-promoted key/value metadata rows for benchmark runs.

benchmark_runs_dataframe

Builds one stable summary row per benchmark run.

frame_calibration_dataframe

Builds acquisition-frame gain and additive-background rows.

frame_diagnostics_dataframe

Builds one row per predicted-versus-reference frame diagnostic record.

frame_evaluation_dataframe

Builds one predicted-versus-measured intensity comparison row per acquisition frame.

history_dataframe

Builds one run-ID/objective/elapsed-time row per completed iteration.

illumination_calibration_dataframe

Builds source-order (row, column) correction rows in Fourier-grid pixels.

iteration_diagnostics_dataframe

Builds one row per recorded convergence-diagnostic iteration.

metadata_dataframe

Builds long-form non-promoted run metadata rows.

raw_frame_statistics_dataframe

Builds one row per raw measured-frame statistics record.

reconstruction_tables

Materializes all required and non-empty optional tables for one result and run ID.

scalar_diagnostics_dataframe

Builds long-form run-ID/key/value rows from named scalar diagnostics.

summary_dataframe

Builds the required one-row reconstruction result summary table.