Module run 

CLI dispatch logic for the readstat binary.

Structs

IndicatifProgress 🔒

ProgressCallback implementation backed by an indicatif::ProgressBar.

ReaderConfig 🔒

Inputs to the reader thread spawned by run_data.

WriteContext 🔒

State shared by every write strategy in run_data.

Constants

CHANNEL_CAPACITY 🔒

Capacity of the bounded channel between reader and writer threads. Also used as the batch size for bounded-batch parallel writes.

STREAM_ROWS 🔒

Default number of rows to read per streaming chunk.

Functions

create_progress 🔒

Create a progress bar if progress is enabled.

format_with_commas 🔒

Formats a number with comma thousands separators (e.g. 1081 -> “1,081”).

join_reader 🔒

Joins the reader thread, surfacing either its internal error or a panic.

print_write_summary 🔒

Prints the “wrote N rows” summary. The library no longer prints this; the CLI owns all user-facing output.

resolve_columns 🔒

Resolve column names from --columns or --columns-file CLI options.

resolve_stream_rows 🔒

Determine stream row count based on reader type.

run

Executes the CLI command specified by the parsed ReadStatCli arguments.

run_data 🔒

Handle the data subcommand: read SAS data and write to an output file.

run_metadata 🔒

Handle the metadata subcommand: read and display SAS file metadata.

run_preview 🔒

Handle the preview subcommand: read a limited number of rows and write to stdout as CSV.

spawn_reader 🔒

Spawns the reader thread that parses row chunks and sends them to sender.

write_empty_output 🔒

Writes a valid empty output file (header-only CSV, empty Parquet/Feather/ NDJSON) when the input contributed zero rows. Without this, a zero-row input would produce no output file at all despite a success exit code.

write_parallel_parquet 🔒

Parallel Parquet write path (only for --parallel --parallel-write with Parquet output): write each buffered batch group to a temp file concurrently, then merge the temp files into the final output.

write_sequential 🔒

Default write path: consume chunks in order, streaming each to the format writer, then finalize. Memory stays bounded because only the most recent chunk is retained — kept solely so finish can report the row total.