sql: a few compatibility features

[jordanlewis]

• add pg_my_temp_schema() builtin
• add array overlaps (&&) operator
• add array contains (@>) and contained_by(<@) operators
• add support for SET SESSION AUTHORIZATION DEFAULT
• add pg_get_function_identity_arguments builtiin

Closes #18442.
Closes #33337.
Closes #39821.
Closes #40211.

[cockroach-teamcity]

This change is 

[knz]

r1 and r4 OK

There's a code mix-up between r2 and r3, so you can either fix the mix-up or squash the commits together.

r5 I really want to trust your understanding of the pg spec but there is currently no guardrail if anyone else tweaks the implementation later - so maybe more tests would help here.

Reviewed 2 of 2 files at r1, 15 of 15 files at r2, 3 of 3 files at r3, 7 of 9 files at r4, 3 of 3 files at r5.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @jordanlewis and @justinj)

---

pkg/sql/logictest/testdata/logic_test/builtin_function, line 2252 at r5 (raw file):

# Sanity check pg_get_function_identity_arguments.
query T
SELECT pg_get_function_identity_arguments('convert_from'::regproc::oid)

I think the complexity of this function mandates a wider diversity of tests.
You probably want a mix of:

• polymorphic functions
• functions with zero arguments
• overloaded functions

and see that the behavior is consistent (and compatible with pg) for all of them.

---

pkg/sql/sem/builtins/pg_builtins.go, line 627 at r1 (raw file):

	),

	"pg_my_temp_schema": makeBuiltin(defProps(),

Maybe add link to pg doc where this is explained in comment above.

---

pkg/sql/sem/tree/eval.go, line 296 at r2 (raw file):

// ArrayContains return true if the haystack contains all needles.
func ArrayContains(haystack *DArray, needles *DArray) (*DBool, error) {

Where is this function used? It seems you got it inlined below.

---

pkg/sql/sem/tree/eval.go, line 303 at r2 (raw file):

		var found bool
		for j := range haystack.Array {
			if needles.Array[i] == haystack.Array[j] {

I think you need Compare here too

---

pkg/sql/sem/tree/eval.go, line 307 at r3 (raw file):

		var found bool
		for _, hay := range haystack.Array {
			if needle.Compare(ctx, hay) == 0 {

ok this change you can pull into the previous commit

---

pkg/sql/sem/tree/eval.go, line 2095 at r3 (raw file):

				haystack := MustBeDArray(left)
				needles := MustBeDArray(right)
				return ArrayContains(ctx, haystack, needles)

ditto

[jordanlewis]

TFTR. I squashed r2 and r3 which I think resolves the issues. I added more tests as well.

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @justinj and @knz)

---

pkg/sql/logictest/testdata/logic_test/builtin_function, line 2252 at r5 (raw file):

Previously, knz (kena) wrote…

I think the complexity of this function mandates a wider diversity of tests.
You probably want a mix of:

• polymorphic functions
• functions with zero arguments
• overloaded functions

and see that the behavior is consistent (and compatible with pg) for all of them.

Done.

---

pkg/sql/sem/builtins/pg_builtins.go, line 627 at r1 (raw file):

Previously, knz (kena) wrote…

Maybe add link to pg doc where this is explained in comment above.

Done.

---

pkg/sql/sem/tree/eval.go, line 296 at r2 (raw file):

Previously, knz (kena) wrote…

Where is this function used? It seems you got it inlined below.

Done.

---

pkg/sql/sem/tree/eval.go, line 303 at r2 (raw file):

Previously, knz (kena) wrote…

I think you need Compare here too

Done.

---

pkg/sql/sem/tree/eval.go, line 307 at r3 (raw file):

Previously, knz (kena) wrote…

ok this change you can pull into the previous commit

Done.

---

pkg/sql/sem/tree/eval.go, line 2095 at r3 (raw file):

Previously, knz (kena) wrote…

ditto

Done.

[knz]

LGTM with two nits

Reviewed 13 of 15 files at r7, 9 of 9 files at r8, 3 of 3 files at r9.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @jordanlewis and @justinj)

---

pkg/sql/parser/sql.y, line 3126 at r9 (raw file):

| SESSION AUTHORIZATION non_reserved_word_or_sconst
  {
    return unimplemented(sqllex, "set session authorization name")

maybe file an issue and link it here? I understand that there are still several errors without issue number throughout the grammar, but I think Andy was asking that any new such errors should be paired with an issue.

---

pkg/sql/sem/builtins/pg_builtins.go, line 615 at r9 (raw file):

	// pg_get_function_identity_arguments returns the argument list necessary to
	// identify a function, in the form it would need to appear in within ALTER
	// FUNCTION, for instance. This form omits default values.

nit: URL to pg doc here too

[jordanlewis]

Fixed nits. TFTRs!

bors r+

[craig]

Build failed

• GitHub CI (Cockroach)

[jordanlewis]

bors r+

[craig]

Build succeeded

• GitHub CI (Cockroach)