Skip to content
NOQTIS

nqdsyn(3)

nqdsyn — write a synthetic CSV file from its schema, marked synthetic

libnqd, -lnqd -lnoqtis

#include <nqd.h>

int
nqdsyn(nqdrec *r, const char *dir, const char *path, uint64_t seed, uint64_t rows, char *why, size_t wcap);

() writes a synthetic stand-in for the CSV file path to dir/path, made from the record's schema of path (noqtis-data(5)), its recipe, and records it in the sources of r with its SHA-256, size and synthetic true, so that every check of it says ‘[synthetic]’. Save the record with nqdsave(3). It never replaces real data: a path whose source is not marked synthetic, or a file there that the record does not mark synthetic, is refused.

The file has the schema's delimiter, a header of the column names unless header is false, and rows rows (0: the schema's rows.min, else 100), each line ended by a line feed; a field that holds the delimiter, a quote or a line break is quoted. The values are the same for the same seed on every platform: they are drawn from the 64-bit big-endian words of SHA-256("noqtis-syn-1" || 0x00 || seed || j), j = 0, 1, ..., each 8 bytes big-endian, four words a block, and a whole number from 0 to k - 1 is a word x taken when x < 2^64 - (2^64 mod k), as x mod k (a word not taken is skipped). Row by row, column by column: when the column's nulls is true, a draw from 0 to 9 of 0 makes the value missing (the schema's first na code, else empty); otherwise

‘int’
+ a draw from 0 to max - min (defaults 0 and 1,000; at most 2^53 values)
‘real’
+ (max - min) × (x >> 11) / 2^53 (defaults 0 and 1), written as nqdjson(3) writes numbers
‘date’
+ a draw of days to max (defaults 2000-01-01 and 2025-12-31)
‘bool’
‘true’ for a draw from 0 to 1 of 1, else ‘false’
‘text with a set’
the element of set at a draw from 0 to its size - 1
‘text’
eight letters ‘a’ to ‘z’, a draw from 0 to 25 each

Every value fits its column's type, range and set, so the file passes the schema's columns; rules (‘unique, ‘incr, ...’’) are not part of the recipe.

NQOK, or with a sentence in why: NQEARG for no schema with columns, path not a capsule path, a source or file not marked synthetic, or an ‘int’ range with no whole number or more than 2^53; NQEIO when the file cannot be written (its folder must exist); NQELIM over 1 GiB; NQEMEM.

noqtis-data(1), nqdinfer(3), nqdnew(3), noqtis-data(5)