Function to return number of statements (or even better, the split statements)

[ryanschneider]

Something like this would be very helpful:

> SELECT parse_statements('SELECT 42; SELECT 43;');
┌───────────────────────────────────────────┐
│ parse_statements('SELECT 42; SELECT 43;') │
│               varchar                     │
├───────────────────────────────────────────┤
│ SELECT 42                                 │
│ SELECT 43                                 │
└───────────────────────────────────────────┘

Still useful but less helpful would be a scalar function num_statements that in this example returns 2.

[zfarrell]

Thanks for the suggestion @ryanschneider . You caught me at the perfect time - I was able to knock this out pretty quickly.

One small quirk: the statement returned is not always the exact (character-by-character) statement that was provided. Because we rely on DuckDB’s parser, the output is the parser’s normalized representation of the SQL. This means it’s always semantically equivalent, but may differ syntactically. For example:

Input	Output
SELECT count(*);	SELECT count_star()
INSERT INTO test VALUES (2)	INSERT INTO test (VALUES (2))
WITH cte AS (SELECT 1) SELECT * FROM cte	WITH cte AS (SELECT 1)SELECT * FROM cte
SELECT * FROM users WHERE id > 10	SELECT * FROM users WHERE (id > 10)
WITH data AS (SELECT * FROM source) SELECT * FROM data	WITH "data" AS (SELECT * FROM "source")SELECT * FROM "data"

These formatting differences are expected behavior since DuckDB's parser normalizes the SQL according to its internal representation.

[ryanschneider]

Oh wow thanks @zfarrell this is perfect!

[zfarrell]

@ryanschneider - this should now be available. there were some build issues in duckdb's community extension ci config, but it just got merged a few hours ago: duckdb/community-extensions#618

[ryanschneider]

Thanks @zfarrell I see it now, it's working great!