Spiral SQL is parsed with a dialect compatible with common analytical SQL, then bound into Spiral’s relation plan. The current user-facing surface is a focused subset for local inspection, data selection, batch shaping, and extension work.

• SELECT queries over one table or one table function.
• VALUES queries.
• Projection, filtering, grouping, aggregate functions, HAVING, ORDER BY, LIMIT, and OFFSET.
• Scalar and aggregate function calls through the SPQL function catalog, including named arguments.
• Built-in table functions for metadata, local Vortex files, Arrow IPC files, Parquet files, and Hugging Face dataset handles.
• EXPLAIN and EXPLAIN ANALYZE.
• CREATE VIEW and materialized CREATE TABLE name AS query catalog entries.
• COPY TO local Vortex and Arrow IPC files.
• SET trace diagnostics.
• REPL-local CREATE TABLE name (...) and INSERT for in-memory tables.

Unsupported syntax should fail with an explicit error rather than silently changing query semantics.

Example query

SELECT
  category,
  count(*) AS rows,
  sum(score) AS total_score
FROM examples
WHERE active IS NOT NULL
GROUP BY category
HAVING count(*) > 10
ORDER BY total_score DESC
LIMIT 20;

Query paths

Query forms

Spiral supports standard SELECT:

SELECT id, score
FROM events
WHERE score > 0.8
LIMIT 10;

It also parses DuckDB-style FROM-first queries:

FROM events
SELECT id, score
WHERE score > 0.8
LIMIT 10;

Use the standard form in docs and examples unless FROM-first syntax is the point of the example.

Namespaces

Scalar and aggregate functions resolve through the engine callable catalog. Unqualified scalar calls are mapped to the spql. namespace when possible:

SELECT add(a, b) FROM t;
SELECT spql.add(a, b) FROM t;

Table and source functions use their full registered id:

SELECT *
FROM spql.meta.table_functions();
 
SELECT *
FROM spql.meta.source_functions();
 
SELECT *
FROM spql.parquet.read('./scores.parquet');

Use SQL support when you need an exact current implementation matrix.