colstore.ColStoreInfo¶

class colstore.ColStoreInfo(path, format_version, n_rows, n_records, columns, file_size)[source]¶

Bases: object

Summary of a colstore file’s contents and on-disk shape.

Returned by info(). Fields are populated from the file header without scanning any record bodies, so the call is fast even on large files.

Parameters:

path (Path)
format_version (int)
n_rows (int)
n_records (int)
columns (list[dict[str, Any]])
file_size (int)

path¶

Filesystem path the info was read from.

Type:: pathlib.Path

format_version¶

On-disk format version. Currently always 1.

Type:: int

n_rows¶

Total committed row count across all records.

Type:: int

n_records¶

Number of records in the file. A one-shot write or a fully compacted file has n_records == 1.

Type:: int

columns¶

Schema: one {"name": ..., "dtype": ..., "encoding": ..., "nullable": ...} per column, in declaration order.

Type:: list[dict]

file_size¶

Size of the file in bytes (os.path.getsize).

Type:: int

property needs_compaction: bool¶

True if collapsing the file to one record would help.

A multi-record file pays a per-pattern dispatch cost on reads (cheap for slice and sorted-fancy, expensive for unsorted-fancy as records accumulate). After compact(), all reads take the single-record fast path.