importccl: Set FK and CHECK constraints to unvalidated after IMPORT INTO

[adityamaru27]

This change sets foreign key and CHECK constraints to unvalidated
after the table is imported into, but before the descriptor is
published in a PUBLIC state.

[cockroach-teamcity]

This change is 

[dt]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @adityamaru27 and @dt)

---

pkg/ccl/importccl/import_stmt.go, line 975 at r1 (raw file):

		if !tbl.IsNew {
			// Set FK constraints to unvalidated before publishing the table imported into.

there's a helper for this if you didn't want to find all the index yourself for _, idx := range tbl.AllNonDropIndexes()

---

pkg/ccl/importccl/import_stmt.go, line 990 at r1 (raw file):

			// Set CHECK constraints to unvalidated before publishing the table imported into.
			tableDesc.Checks = make([]*sqlbase.TableDescriptor_CheckConstraint, len(tbl.Desc.Checks))
			for i, c := range tbl.Desc.Checks {

this direct access vs using the accessor is okay because we asserted that Mutations is empty

[adityamaru27]

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

---

pkg/ccl/importccl/import_stmt.go, line 975 at r1 (raw file):

Previously, dt (David Taylor) wrote…

there's a helper for this if you didn't want to find all the index yourself for _, idx := range tbl.AllNonDropIndexes()

Done.

---

pkg/ccl/importccl/import_stmt.go, line 990 at r1 (raw file):

Previously, dt (David Taylor) wrote…

this direct access vs using the accessor is okay because we asserted that Mutations is empty

feels cleaner to use the accessor, done.

[adityamaru27]

bors r+

[dt]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @adityamaru27 and @dt)

---

pkg/ccl/importccl/import_stmt.go, line 975 at r1 (raw file):

Previously, adityamaru27 (Aditya Maru) wrote…

Done.

not quite what I meant: the point of the helper returning pointers to index desc is that you can just mutate via the pointer. I actually think the direct field access was arguably bet

---

pkg/ccl/importccl/import_stmt.go, line 986 at r2 (raw file):

					tableDesc.PrimaryIndex = tableIdx
				} else {
					tableDesc.Indexes = append(tableDesc.Indexes, tableIdx)

Er, I don't think we want to be appending.

[dt]

bors r-

[craig]

Canceled

[dt]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @adityamaru27)

---

pkg/ccl/importccl/import_stmt.go, line 986 at r2 (raw file):

Previously, dt (David Taylor) wrote…

Er, I don't think we want to be appending.

To expand on that: AllNonDropIndexes is returning pointers to things already in desc.Indexes (as well as to the PK and indexes in Mutations) so if we append like this, I think we're actually duplicating the existing indexes, right? Plus, if there were any in mutations (though I think in this case that is not a concern) this would be copying them to the public index list too outside the normal mutation completion path. Oh, I missed the blanking of desc.Indexes above, sorry. 🤦‍♂ that said, I'd think I'd still rather edit directly via the pointer, rather than in these special-case branches for the different descriptor fields, since apparently I'm prone to mis-reading it otherwise 😬.

[adityamaru27]

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

---

pkg/ccl/importccl/import_stmt.go, line 986 at r2 (raw file):

Previously, dt (David Taylor) wrote…

To expand on that: AllNonDropIndexes is returning pointers to things already in desc.Indexes (as well as to the PK and indexes in Mutations) so if we append like this, I think we're actually duplicating the existing indexes, right? Plus, if there were any in mutations (though I think in this case that is not a concern) this would be copying them to the public index list too outside the normal mutation completion path.

Done.

[adityamaru27]

bors r+

[craig]

Build succeeded

• GitHub CI (Cockroach)