gsort

gsort provides stable per-key ascending and descending ordering for the active dataset. It is inspired by Stata's gsort, SAS descending sort keys, and SPSS SORT CASES, while retaining TabDat's native scalar ordering and explicit null semantics.

Syntax

gsort [+|-]varlist

Each key may begin with + (ascending), - (descending), or no prefix (ascending):

gsort -date +patient_id
gsort +site -score

Keys are applied left-to-right. Ties preserve their previous active-row order. Quoted identifiers are not interpreted as direction prefixes, so a variable literally named -score can be quoted.

Semantics

• Numeric, text, boolean, date, timestamp, decimal, and other supported scalar keys use existing native comparison semantics.
• Missing/null keys are always placed after nonmissing values, including descending keys.
• Sorting is stable; an internal ordinal tie-breaker is not retained in the public schema.
• Unknown variables are rejected before any active-data change.
• Eager, DuckDB-lazy, and Polars-lazy execution use native stable backends. Polars-lazy remains lazy.
• Labels, valid panel metadata, named-table synchronization, and transform result behavior follow the existing sort command.

Output

gsort returns the existing transform result:

Sorted by: -group_id +label

JSON output uses TransformResult with the same message and updated dataset metadata.

Invalid forms

gsort
gsort group_id, stable
gsort group_id if active == true
gsort -
gsort --score

These forms fail with deterministic parse errors. An unknown key produces the standard execution error without mutating the active dataset.