schemadiff: analyze and report foreign key loops/cycles

[shlomi-noach]

Description

See #15061 for details.

schemadiff does not support foreign key cycles (with the exception of self referencing tables). In this PR, schemadiff identifies cycles and reports those cycles as part of the schema validation process. Taking the same example from #15061:

create table t11 (id int primary key, i int, constraint f11 foreign key (i) references t12 (id));
create table t12 (id int primary key, i int, constraint f12 foreign key (i) references t11 (id));

As of this PR, schemadiff will now report:

table `t11` participates in foreign key loop: `t11`, `t12`, `t11`
table `t12` participates in foreign key loop: `t11`, `t12`, `t11`

schemadiff will further identify tables that reference foreign key loops, i.e. their immediate parent or on of their indirect parents participates in a loop, and will report something like:

table `t13` references foreign key loop: `t11`, `t12`, `t11`

Computing foreign key loops uses DFS, and for every loop found it caches the tables participating in that loop, so as to never recompute the same table twice. This limits the number of loops found per table to 1, which is sufficient, as we already only report one foreign key error per table anyway (otherwise this can blow up with excessive error messages).

Related Issue(s)

• Fixes Feature Request: foreign key loop analysis in schemadiff #15061
• schemadiff: tracking issue #10203

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.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (f751c83) 47.49% compared to head (1a4ed10) 40.88%.
Report is 41 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main   #15062       +/-   ##
===========================================
- Coverage   47.49%   40.88%    -6.61%     
===========================================
  Files        1149     1454      +305     
  Lines      239475   368246   +128771     
===========================================
+ Hits       113730   150567    +36837     
- Misses     117138   201488    +84350     
- Partials     8607    16191     +7584     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dbussink]

Is it useful to have for each step in the loop also the column name? To more exactly identify the specific loop?

[shlomi-noach]

I'm afraid of clutter. The foreign key might reference multiple columns. It is indeed possible that same table child will reference table parent with two different foreign keys, one of which has a cycle, the other does not. In such case it would indeed be beneficial to identify the referenced columns both on parent and child; but again this is an amount of information that is likely to create much background noise. Alternatively, we could use the foreign key name participating in the loop.
Either solution would complicate the loop detection logic only slightly; question is how useful vs. confusing would the information be.

[shlomi-noach]

Just took a look at the code. The current approach (only report table name) is consistent with other schemadiff error reports, that suffice with table name or column name without drilling into specific details. I think this should be fine as it is for now.

[mattlord]

Nitty, but I think this is slightly more performant:

	escaped := make([]string, len(e.Loop))
	for i, t := range e.Loop {
		escaped[i] = sqlescape.EscapeID(t)

[shlomi-noach]

Applied.

[mattlord]

I think "infinite loop" or "circular foreign key references" might be more descriptive for user facing error messages and docs, etc.

[shlomi-noach]

fixed

[mattlord]

I think this comment is a little off, no? Maybe ~ // We only return the tail end of seen which begins with this table.?

[shlomi-noach]

Reworded as:

// We only return the suffix of `seen` that starts (and now ends) with this table.