CLI reference

$ openqvd --help

Subcommand	What it does
stat	Print a header summary: table name, row count, record size, and every field's bit offset/width/bias/symbol count.
head	Print the first N rows (default 10) as tab-separated values, with a header row.
csv	Print every row as tab-separated values. NULLs become empty fields.
json	Print one JSON object per line, one key per field.
rewrite	Read a file and re-serialise it through the writer. Useful for round-trip smoke testing.

Subcommand details

stat <file>

$ openqvd stat data.qvd
table: Orders
rows:  3
record_byte_size: 1  row_block: offset=27 length=3
fields (2):
  [ 0] OrderId                           bits@0+2   bias=0    n_sym=3       type=UNKNOWN  tags=
  [ 1] Region                            bits@2+2   bias=-2   n_sym=2       type=UNKNOWN  tags=

(Region has bias=-2 because one of the three rows is NULL for that column - see Format: NULL representation.)

head <file> [--rows N]

$ openqvd head data.qvd --rows 5

Prints a tab-separated header row followed by the first N rows (default 10).

csv <file>

$ openqvd csv data.qvd > data.tsv

Despite the name, output is tab-separated for simplicity - there is no quoting to get wrong. NULL cells are empty fields; embedded tabs and newlines in string values are replaced with a single space.

json <file>

$ openqvd json data.qvd > data.jsonl

One JSON object per line (JSON Lines), field name to value. DualInt and DualFloat symbols emit their numeric component (falling back to the text form only when the number is non-finite); NULL becomes null.

rewrite <in> <out>

$ openqvd rewrite data.qvd roundtrip.qvd
wrote roundtrip.qvd

Reads <in> and re-serialises it through Qvd::write_to_path. The output is semantically equivalent to the input (same row count, field names, and cell values) but is not guaranteed to be byte- identical, since the writer chooses its own bit-packing and symbol layout.