feat(postgres): provide translation for hash ops

[deepyaman]

Description of changes

Postgres supports hashing on a per-type basis, with "extended" versions providing 64-bit integer output. To list these, run \df hash*extended:

Schema	Name	Result data type	Argument data types	Type
pg_catalog	hash_aclitem_extended	bigint	aclitem, bigint	func
pg_catalog	hash_array_extended	bigint	anyarray, bigint	func
pg_catalog	hash_multirange_extended	bigint	anymultirange, bigint	func
pg_catalog	hash_numeric_extended	bigint	numeric, bigint	func
pg_catalog	hash_range_extended	bigint	anyrange, bigint	func
pg_catalog	hash_record_extended	bigint	record, bigint	func
pg_catalog	hashbpcharextended	bigint	character, bigint	func
pg_catalog	hashcharextended	bigint	"char", bigint	func
pg_catalog	hashenumextended	bigint	anyenum, bigint	func
pg_catalog	hashfloat4extended	bigint	real, bigint	func
pg_catalog	hashfloat8extended	bigint	double precision, bigint	func
pg_catalog	hashinetextended	bigint	inet, bigint	func
pg_catalog	hashint2extended	bigint	smallint, bigint	func
pg_catalog	hashint4extended	bigint	integer, bigint	func
pg_catalog	hashint8extended	bigint	bigint, bigint	func
pg_catalog	hashmacaddr8extended	bigint	macaddr8, bigint	func
pg_catalog	hashmacaddrextended	bigint	macaddr, bigint	func
pg_catalog	hashnameextended	bigint	name, bigint	func
pg_catalog	hashoidextended	bigint	oid, bigint	func
pg_catalog	hashoidvectorextended	bigint	oidvector, bigint	func
pg_catalog	hashtextextended	bigint	text, bigint	func
pg_catalog	hashtidextended	bigint	tid, bigint	func
pg_catalog	hashvarlenaextended	bigint	internal, bigint	func

This PR implements as many of them as possible, from the set of available Ibis datatypes in ibis/backends/sql/datatypes.py. No attempt to cast is made (e.g. hashing booleans is unsupported, rather than casting to an integer and then using the available hash method).

[deepyaman]

I'm just not testing macaddr here, since alltypes doesn't have that column. Should I just add a macaddr column?

On top of that, I also think something is off with macaddr to begin with, though, since PostgresType.from_ibis(dt.macaddr) returns DataType(this=Type.VARCHAR), and test_macaddr_literal in ibis/backends/tests/test_network.py expects a text type to be returned; I think the change should be bigger, but wanted to get an opinion first. :)

[gforsyth]

adding a column to alltypes will be a headache -- I think we could add a standalone macaddr hash test, but that can also happen in a followup

[deepyaman]

Oh, sorry, what I had in mind was that can just add the column to the alltypes table inside the test, but a standalone test would also work.

In any case, yes, happy to do it in a follow-up, just because I'm not sure about some of the other behaviors mentioned for macaddr (and also because I think that's a much less important use case). 😅

[gforsyth]

Would all of these and also decimal types be covered by hashnumericextended?

[gforsyth]

and you could dispatch on those with is_numeric()

[deepyaman]

Oh, true...

Suggested change

	if arg_dtype.is_int16():
	return self.f.hashint2extended(arg, 0)
	elif arg_dtype.is_int32():
	return self.f.hashint4extended(arg, 0)
	elif arg_dtype.is_int64():
	return self.f.hashint8extended(arg, 0)
	elif arg_dtype.is_float32():
	return self.f.hashfloat4extended(arg, 0)
	elif arg_dtype.is_float64():
	return self.f.hashfloat8extended(arg, 0)
	if arg_dtype.is_numeric():
	return self.f.hash_numeric_extended(arg, 0)

[deepyaman]

@gforsyth Actually, doesn't seem to work. Maybe it would require an explicit cast.

I find these functions pretty poorly documented, so I will go ahead and revert this for now to get it green again.

[gforsyth]

adding a column to alltypes will be a headache -- I think we could add a standalone macaddr hash test, but that can also happen in a followup