Function stringAgg

String aggregation (GROUP_CONCAT / STRING_AGG / LISTAGG / ClickHouse array-concat), dispatched per-dialect. This is the single most error-prone aggregate across dialects, so each branch below documents the exact verified syntax.

• MySQL / MariaDB: GROUP_CONCAT([DISTINCT] expr [ORDER BY col [ASC|DESC]] SEPARATOR 'delim') Clause order is fixed: DISTINCT, then the expression, then ORDER BY, then SEPARATOR — SEPARATOR must come last and is a keyword, not a comma argument.

• SQLite: GROUP_CONCAT(expr, 'delim') (2-argument form; the separator defaults to , if omitted). SQLite's GROUP_CONCAT does not support an ORDER BY clause inside the function call at all — if options.orderBy is passed for SQLite this throws, since silently ignoring it would produce non-deterministic ordering that looks intentional. Pre-sort via a subquery/CTE if ordering is required. SQLite's DISTINCT variant (GROUP_CONCAT(DISTINCT expr)) is supported but forces the default comma separator (a custom separator combined with DISTINCT is not supported by SQLite) — this also throws if both distinct and a non-default delimiter are requested together.

• PostgreSQL / CockroachDB: STRING_AGG([DISTINCT] expr, 'delim' [ORDER BY col [ASC|DESC]]) — the ORDER BY clause sits inside the same parentheses as the delimiter argument (no WITHIN GROUP).

• MSSQL: STRING_AGG(expr, 'delim') WITHIN GROUP (ORDER BY col [ASC|DESC]) — note this is a genuinely different clause placement than Postgres: the ORDER BY is a separate WITHIN GROUP (...) clause appended after the function call, not inline inside the argument list. MSSQL's STRING_AGG also does not support a DISTINCT argument at all (unlike Postgres) — this throws if distinct: true is requested.

• Oracle / Db2: LISTAGG([DISTINCT] expr, 'delim') WITHIN GROUP (ORDER BY col [ASC|DESC]) — same WITHIN GROUP placement as MSSQL. Oracle supports DISTINCT inside LISTAGG since 19c; Db2 LUW's LISTAGG also supports DISTINCT.

• Redshift: LISTAGG([DISTINCT] expr, 'delim') WITHIN GROUP (ORDER BY col [ASC|DESC]) — Redshift's aggregate string function has always been LISTAGG (matching Oracle/Db2 syntax), not STRING_AGG. AWS did add a Postgres-compatible STRING_AGG alias to Redshift more recently, but LISTAGG is the long-standing, universally-available, and ORDER-BY-capable form, so it's used here for correctness.

• Snowflake: LISTAGG([DISTINCT] expr, 'delim') WITHIN GROUP (ORDER BY col [ASC|DESC]) — Snowflake's string aggregation function is LISTAGG (Oracle-style), not STRING_AGG — Snowflake has no STRING_AGG function.

• ClickHouse: no native GROUP_CONCAT/STRING_AGG/LISTAGG. The verified, version-portable idiom is arrayStringConcat(groupArray(expr), 'delim') (distinct: arrayStringConcat(groupUniqArray(expr), 'delim')). Newer ClickHouse releases (24.x+) also ship a MySQL-compatible groupConcat(expr) function, but arrayStringConcat(groupArray(...), ...) works on every supported version so it's used as the default here. ClickHouse's groupArray has no inline ORDER BY sub-clause (it preserves block processing order, not a guaranteed sort) — options.orderBy therefore throws for ClickHouse; pre-sort via a subquery if ordering matters.