schemadiff: using MySQL capabilities to analyze a SchemaDiff and whether changes are applicable instantly/immediately.

[shlomi-noach]

Description

See feature description in #14877. schemadiff, together with CapableOf & FlavorCapability, is able to answer:

• Whether a particular ALTER TABLE on a given table schema is eligible for ALGORITHM=INSTANT
• Whether an entire SchemaDiff (the diff of two schemas) result is applicable immediately/instantly, meaning all diffs are either trivially immediate (CREATE TABLE, ALTER VIEW, etc.) or are eligible for ALGORITHM=INSTANT.

This is a first, and isolated, step in a longer change, where:

• Based on AlterTableAlgorithmStrategy hint, we could auto-apply ALGORITHM=INSTANT to alter table statements.
• onlineddl/analysis.go, which has overlapping functionality, can then defer to schemadiff.

To reduce schemadiff's package dependency, this PR also refactors CapableOf & FlavorCapability out of go/vt/mysql and into go/vt/mysql/capabilities.

Related Issue(s)

#14877

Checklist

• "Backport to:" labels have been added if this change should be back-ported to release branches
• If this change is to be back-ported to previous releases, a justification is included in the PR description
• Tests were added or are not required
• Did the new or modified tests pass consistently locally and on CI?
• Documentation was added or is not required

Deployment Notes

[vitess-bot]

Review Checklist

Hello reviewers! 👋 Please follow this checklist when reviewing this Pull Request.

General

• Ensure that the Pull Request has a descriptive title.
• Ensure there is a link to an issue (except for internal cleanup and flaky test fixes), new features should have an RFC that documents use cases and test cases.

Tests

• Bug fixes should have at least one unit or end-to-end test, enhancement and new features should have a sufficient number of tests.

Documentation

• Apply the release notes (needs details) label if users need to know about this change.
• New features should be documented.
• There should be some code comments as to why things are implemented the way they are.
• There should be a comment at the top of each new or modified test to explain what the test does.

New flags

• Is this flag really necessary?
• Flag names must be clear and intuitive, use dashes (-), and have a clear help text.

If a workflow is added or modified:

• Each item in Jobs should be named in order to mark it as required.
• If the workflow needs to be marked as required, the maintainer team must be notified.

Backward compatibility

• Protobuf changes should be wire-compatible.
• Changes to _vt tables and RPCs need to be backward compatible.
• RPC changes should be compatible with vitess-operator
• If a flag is removed, then it should also be removed from vitess-operator and arewefastyet, if used there.
• vtctl command output order should be stable and awk-able.

[mattlord]

Looks good! I had one functionality related question with the rest being minor comments/suggestions/questions.

[mattlord]

We don't need to perform any math if we're checking for equality here, do we? i.e., I don't think we need to do more than:

if len(col.Type.EnumValues) != len(opt.NewColDefinition.Type.EnumValues) {

Maybe that's not what we want to do though either? I'm assuming it's fine if the new value is less than the existing one. I would think we only care if the new value would be larger / require more space:

if (len(col.Type.EnumValues)+7)/8 < (len(opt.NewColDefinition.Type.EnumValues)+7)/8 {

[shlomi-noach]

Unfortunately both assumptions are incorrect. The condition for allowing ALGORITHM=INSTANT is that the storage space for the SET/ENUM value is identical.

Here's how storage for 6 values is the same as the storage for 7 values, and INSTANT is allowed:

> create table t(id int, c1 set('a', 'b', 'c', 'd', 'e', 'f'), primary key(id));
Query OK, 0 rows affected (0.01 sec)

> alter table t modify column c1 set('a', 'b', 'c', 'd', 'e', 'f', 'g'), algorithm=instant;
Query OK, 0 rows affected (0.01 sec)
Records: 0  Duplicates: 0  Warnings: 0

Here's how the storage for 9 values (2 bytes) reduced to 8 values (1 byte) disables INSTANT:

> create table t(id int, c1 set('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i'), primary key(id));
Query OK, 0 rows affected (0.01 sec)

> alter table t modify column c1 set('a', 'b', 'c', 'd', 'e', 'f', 'g', 'h'), algorithm=instant;
ERROR 1846 (0A000): ALGORITHM=INSTANT is not supported. Reason: Need to rebuild the table to change column type. Try ALGORITHM=COPY/INPLACE.

[mattlord]

Just noting that this uses wrapping (which generally means the outer error was caused by the inner). I'm not sure if that's what we want here?

[shlomi-noach]

To the best of my knowledge, errors.Join is not doing wrapping. It implements a slices of err objects, most of which are inaccessible unless you explicitly cast your err to an Unwrap() []error interface, like so:

vitess/go/errors/errors.go

Lines 20 to 33 in bd9e534

	type Wrapped interface {
	Unwrap() []error
	}
	// Unwrap unwraps an error created by errors.Join() in Go 1.20, into its components
	func Unwrap(err error) []error {
	if err == nil {
	return nil
	}
	if u, ok := err.(Wrapped); ok {
	return u.Unwrap()
	}
	return nil
	}

Concatenating errors is a paradigm we use throughout schemadiff, with the explicit intention of returning as much information about diff/problems in one go.

[shlomi-noach]

BTW I realize the above code reads Wrapped and Unwrap, but I think it's not in the same way you meant it (meaning it's not hierarchical parent->child, but flat). Possibly we should rename those function and type.

[mattlord]

Note that if this fixes #14877 -- meaning that it should be closed when this is merged -- then we need to add the Fixes keyword in the Related Issue(s) line. I would have added it myself but not 100% sure that's what you wanted.

[shlomi-noach]

Note that if this fixes #14877

It does not fully fix it. I'll have a followup PR for that!

[shlomi-noach]

Followup PR in #14883