feat(snowflake)!: Transpilation support for HASH function

[fivetran-ashashankar]

No description provided.

[geooo109]

Does this logic cover all the cases ?

What happens when we have a join ?

[geooo109]

Also there is no need to check for unsupported.

[geooo109]

Did you take into account this case ?

Snowflake:
  SELECT HASH(x) FROM (
      SELECT 2.0 AS x
      UNION ALL
      SELECT 2.0 AS x
  );
> -3690131753453205264
-3690131753453205264

SELECT HASH(x) FROM (
      SELECT 2 AS x
      UNION ALL
      SELECT 2 AS x
  );
> -3690131753453205264
-3690131753453205264

Both cases return the same values in snowflake.
======================================
Duckdb:
memory D   SELECT HASH(x) FROM (
               SELECT 2 AS x
               UNION ALL
               SELECT 2 AS x
           );
┌─────────────────────┐
│       hash(x)       │
│       uint64        │
├─────────────────────┤
│ 2060787363917578834 │
│ 2060787363917578834 │
└─────────────────────┘
memory D SELECT HASH(x) FROM (
               SELECT 2.0 AS x
               UNION ALL
               SELECT 2.0 AS x
           );
┌─────────────────────┐
│       hash(x)       │
│       uint64        │
├─────────────────────┤
│ 8094069980479725634 │
│ 8094069980479725634 │

From snowflake docs https://docs.snowflake.com/en/sql-reference/functions/hash#usage-notes :

Any two values of type NUMBER that compare equally will hash to the same hash value, even if the respective types have different precision and/or scale.

Any two values of type FLOAT that can be converted to NUMBER(38, 0) without loss of precision will hash to the same value. For example, the following all return the same hash value:

HASH(10::NUMBER(38,0))

HASH(10::NUMBER(5,3))

HASH(10::FLOAT)

[geooo109]

We should avoid generating raw strings with f-string. We have to create expressions and generate them.

[geooo109]

Did you check if HASH exists in other dialects ?

[fivetran-amrutabhimsenayachit]

The HASH(*) implementation is only handling *and missing other Snowflake variants from the docs:

(* ILIKE 'col1%') - (could be unsupported)
(* EXCLUDE col1)
(mytable.*)

sqlglot.transpile('SELECT HASH(t.*) FROM (SELECT 1 AS a, 2 AS b, 3 AS c) t', read='snowflake', write='duckdb')[0])" | duckdb
Binder Error:
No function matches the given name and argument types 'hash()'. You might need to add explicit type casts.
        Candidate functions:
        hash(ANY, [ANY...]) -> UBIGINT


LINE 1: SELECT HASH(t.*) FROM (SELECT 1 AS a, 2 AS b, 3 AS c) AS t

We could perhaps try this:
HASH(UNPACK(COLUMNS(* EXCLUDE a))) & HASH(UNPACK(COLUMNS(t.*)))
instead of
f"HASH(UNPACK(COLUMNS({table_alias}.*)))"

[geooo109]

I will close this for now, it's complicated and the solution isn't trivial.