sql: add drop type to the prepared statement generator

[tharun208]

Release justification: fixes for high-priority or high-severity bugs in existing functionality

Previously, prepare statement doesn't have support for drop type
in the optimizer, so it panics during execution as no flags are generated.

This patch fixes this by adding dropType to the optimizer.

Resolves #61226

Release note: none

Signed-off-by: Tharun [email protected]

[blathers-crl]

Thank you for contributing to CockroachDB. Please ensure you have followed the guidelines for creating a PR.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

I have added a few people who may be able to assist in reviewing:

• @rafiss (author of sql: PREPARE q AS DROP TYPE typ causes an internal error #61226)

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is otan.}

[cockroach-teamcity]

This change is 

[tharun208]

@rafiss, can you help by re-triggering the ci ?. thanks in advance.

[rafiss]

thanks for this fix!

this needs a test. can you add one in pkg/sql/logictest/testdata/logic_test/drop_type

the change looks fine, but since it's optimizer-related and we're in stability period, @RaduBerinde can you please sign off?

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @tharun208)

---

pkg/sql/plan_opt.go, line 64 at r1 (raw file):

		*tree.CreateStats,
		*tree.Deallocate, *tree.Discard, *tree.DropDatabase, *tree.DropIndex,
		*tree.DropTable, *tree.DropView, *tree.DropSequence,

could you add DropType here, closer to the other drop statements?

[tharun208]

pkg/sql/logictest/testdata/logic_test/drop_type

@rafiss, I think the test should need to be a part of prepare statement

[blathers-crl]

Thank you for updating your pull request.

My owl senses detect your PR is good for review. Please keep an eye out for any test failures in CI.

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is otan.}

[RaduBerinde]

Can you also add DROP TYPE to TestStatementReuses ? This tests that all those statements work with PREPARE and EXPLAIN.

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @rafiss and @tharun208)

[tharun208]

Can you also add DROP TYPE to TestStatementReuses ? This tests that all those statements work with PREPARE and EXPLAIN.

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @rafiss and @tharun208)

@RaduBerinde done with the changes. To pass the TestStatementReuses , I need to add Explain to the plan_opt. But , now existing test for select explain is failing

[RaduBerinde]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @rafiss and @tharun208)

---

pkg/sql/plan_opt.go, line 65 at r2 (raw file):

		*tree.Deallocate, *tree.Discard, *tree.DropDatabase, *tree.DropIndex,
		*tree.DropTable, *tree.DropView, *tree.DropSequence, *tree.DropType,
		*tree.Execute, *tree.Explain,

Adding Explain here is not correct, Explain has result columns and is supported by the optimizer.

I assume you added it because some PREPARE EXPLAIN case was failing?

[tharun208]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @rafiss and @tharun208)

pkg/sql/plan_opt.go, line 65 at r2 (raw file):

		*tree.Deallocate, *tree.Discard, *tree.DropDatabase, *tree.DropIndex,
		*tree.DropTable, *tree.DropView, *tree.DropSequence, *tree.DropType,
		*tree.Execute, *tree.Explain,

Adding Explain here is not correct, Explain has result columns and is supported by the optimizer.

I assume you added it because some PREPARE EXPLAIN case was failing?

yes, just now checked EXPLAIN will return the result columns and yes, I added *tree.Explain because, when I tried adding DROP TYPE to TestStatementReuses , PREPARE p AS EXPLAIN of my test DROP TYPE type_name failed

[RaduBerinde]

---

pkg/sql/plan_opt.go, line 65 at r2 (raw file):

Previously, tharun208 (Tharun Rajendran) wrote…

yes, just now checked EXPLAIN will return the result columns and yes, I added *tree.Explain because, when I tried adding DROP TYPE to TestStatementReuses , PREPARE p AS EXPLAIN of my test DROP TYPE type_name failed

What was the failure?

[tharun208]

pkg/sql/plan_opt.go, line 65 at r2 (raw file):

Previously, tharun208 (Tharun Rajendran) wrote…
What was the failure?

The same stack trace error we got for DROP TYPE on prepared statement , no Annotations found.

[RaduBerinde]

I think the right fix here is to move the part that uses .Ann() in startExec instead of DropType, similar to how the drop_table code does it.

[tharun208]

I think the right fix here is to move the part that uses .Ann() in startExec instead of DropType, similar to how the drop_table code does it.

yes, I am able to achieve it, without adding dropType to the optimizer.

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @RaduBerinde and @tharun208)

---

pkg/sql/drop_type.go, line 33 at r3 (raw file):

type typeToDrop struct {
	desc   *typedesc.Mutable
	fqName string

i believe fqName should be removed now

---

pkg/sql/drop_type.go, line 111 at r3 (raw file):

		// Record these descriptors for deletion.
		node.td[typeDesc.ID] = typeToDrop{
			desc: typeDesc,

instead of removing fqName here, does it work with:

		fqName, err := getTypeNameFromTypeDescriptor(
			oneAtATimeSchemaResolver{ctx, p},
			typeDesc,
		)
		if err != nil {
			return nil, err
		}
		node.td[typeDesc.ID] = typeToDrop{
			desc:   typeDesc,
			fqName: fqName.FQString(),
		}

---

pkg/sql/drop_type.go, line 151 at r3 (raw file):

func (n *dropTypeNode) startExec(params runParams) error {
	for _, toDrop := range n.td {
		typ, fqName := toDrop.desc, toDrop.fqName

and now this can be changed to

fqName := tree.AsStringWithFQNames(n.n, params.Ann())

---

pkg/sql/drop_type.go, line 152 at r3 (raw file):

	for _, toDrop := range n.td {
		typ, fqName := toDrop.desc, toDrop.fqName
		if err := params.p.dropTypeImpl(params.ctx, typ, tree.AsStringWithFQNames(n.n, params.Ann()), true /* queueJob */); err != nil {

use fqName as the 3rd argument here

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @RaduBerinde and @tharun208)

---

pkg/sql/drop_type.go, line 33 at r3 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

i believe fqName should be removed now

sorry, i didn't mean to publish this comment.

---

pkg/sql/drop_type.go, line 151 at r3 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

and now this can be changed to

fqName := tree.AsStringWithFQNames(n.n, params.Ann())

i didn't mean to publish this comment

[RaduBerinde]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @tharun208)

---

pkg/sql/drop_type.go, line 112 at r3 (raw file):

		node.td[typeDesc.ID] = typeToDrop{
			desc:   typeDesc,
			fqName: tree.AsStringWithFQNames(name, p.Ann()),

I don't get it, now fqName is not set in half the cases, but we are using it in startExec.

Why don't we remove the field altogether and just do getTypeNameFromTypeDescriptor() in all cases instartExec and use FQString()

[RaduBerinde]

---

pkg/sql/drop_type.go, line 152 at r3 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

use fqName as the 3rd argument here

Noete that the third argument is the entire statement, not just the type name. I would move it out of the loop and assign to an aptly named variable, it's silly to reformat for each type.

[tharun208]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @tharun208)

pkg/sql/drop_type.go, line 112 at r3 (raw file):

		node.td[typeDesc.ID] = typeToDrop{
			desc:   typeDesc,
			fqName: tree.AsStringWithFQNames(name, p.Ann()),

I don't get it, now fqName is not set in half the cases, but we are using it in startExec.

Why don't we remove the field altogether and just do getTypeNameFromTypeDescriptor() in all cases instartExec and use FQString()

this will change the TypeName in the event log. Current changes will not alter the event log` and all necessary cases got passed.

[RaduBerinde]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @rafiss and @tharun208)

---

pkg/sql/drop_type.go, line 112 at r3 (raw file):

Previously, tharun208 (Tharun Rajendran) wrote…

this will change the TypeName in the event log. Current changes will not alter the event log` and all necessary cases got passed.

I don't understand.. what I'm suggesting should be equivalent to the latest revision, we would just call getTypeNameFromTypeDescriptor from startExec instead. Here's a diff (it passes the sql package tests and the drop_type logictest):

https://gist.github.com/RaduBerinde/f78a2b2c17f22cc429734b366f53edc5

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @RaduBerinde and @tharun208)

---

pkg/sql/drop_type.go, line 152 at r3 (raw file):

Previously, RaduBerinde wrote…

Noete that the third argument is the entire statement, not just the type name. I would move it out of the loop and assign to an aptly named variable, it's silly to reformat for each type.

oh good call. hmm well also this parameter is the job description, which is used in the db console as a display name. so perhaps it would be better for them to have different names anyway.

maybe change to

params.p.dropTypeImpl(params.ctx, typ, "dropping type " + fqName, true /* queueJob */)

[RaduBerinde]

---

pkg/sql/drop_type.go, line 152 at r3 (raw file):

Previously, rafiss (Rafi Shamim) wrote…

oh good call. hmm well also this parameter is the job description, which is used in the db console as a display name. so perhaps it would be better for them to have different names anyway.

maybe change to

params.p.dropTypeImpl(params.ctx, typ, "dropping type " + fqName, true /* queueJob */)

Yeah that's probably better

[rafiss]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @RaduBerinde, @rafiss, and @tharun208)

---

pkg/sql/drop_type.go, line 135 at r4 (raw file):

	statementStr := tree.AsStringWithFQNames(n.n, params.Ann())
	for _, typeDesc := range n.toDrop {
		err := params.p.dropTypeImpl(params.ctx, typeDesc, "dropping type "+statementStr, true /* queueJob */)

please change "dropping type "+statementStr to dropping type "+typeFQName

and remove the statementStr variable entirely

[tharun208]

Reviewable status: complete! 0 of 0 LGTMs obtained (and 1 stale) (waiting on @RaduBerinde, @rafiss, and @tharun208)

pkg/sql/drop_type.go, line 135 at r4 (raw file):

	statementStr := tree.AsStringWithFQNames(n.n, params.Ann())
	for _, typeDesc := range n.toDrop {
		err := params.p.dropTypeImpl(params.ctx, typeDesc, "dropping type "+statementStr, true /* queueJob */)

please change "dropping type "+statementStr to dropping type "+typeFQName

and remove the statementStr variable entirely

so, this dropping type "+typeFQName will be changed to dropping type "+typeFQName.FQString(). since both are of different types.

[rafiss]

thank you! this was a nice cleanup

[rafiss]

bors r+

[RaduBerinde]

Thanks!

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @rafiss and @tharun208)

[craig]

Build succeeded:

• GitHub CI (Cockroach)