Simulator: add InsertSelectNested property

[LeMikaelF]

This will catch regressions of #3436, and also opens the door to more complex FROM clauses by allowing the leftmost table in a FROM clause to be nested.

Example of a generated interaction:

 INFO execute_interaction_turso{conn_index=2 interaction=INSERT INTO fortuitous_clique SELECT * FROM (SELECT * FROM (SELECT * FROM (SELECT * FROM (SELECT * FROM (SELECT * FROM (SELECT * FROM fortuitous_clique WHERE TRUE) WHERE TRUE) WHERE TRUE) WHERE TRUE) WHERE TRUE) WHERE TRUE) WHERE TRUE; -- 2}: limbo_sim::runner::execution: 174: 

To test this, I reverted dc231abb2e0fe81cb1e96ab32febe7749c5794c8 locally, and the simulator was able to catch the bug:

Error: failed with error: 'InternalError("Assertion 'table sparkling_darity should have the expected content' failed: expected 16 rows but got 109 for table sparkling_darity")'

While adding this property, I found a bug which Alperen then fixed, so some of the commits on this branch are by him.

[nyrkio]

Commit: 20130df

No performance changes detected. 🚀

[nyrkio]

Commit: db629ea

No performance changes detected. 🚀

[nyrkio]

Commit: 20130df

No performance changes detected. 🚀

[nyrkio]

Commit: db629ea

No performance changes detected. 🚀

[nyrkio]

Commit: 58e5e94

No performance changes detected. 🚀

[nyrkio]

Commit: 58e5e94

No performance changes detected. 🚀

[nyrkio]

Commit: 58e5e94

No performance changes detected. 🚀

[nyrkio]

Commit: f9c15aa

No performance changes detected. 🚀

[nyrkio]

Commit: 3b00099

No performance changes detected. 🚀

[nyrkio]

Commit: 3b00099

No performance changes detected. 🚀

[nyrkio]

Commit: 6cb0d78

No performance changes detected. 🚀

[nyrkio]

Commit: 6cb0d78

No performance changes detected. 🚀

[jussisaurio]

I don't think this needs to be a separate property. Our arbitrary SELECT should sometimes generate subqueries, and since the existing INSERT ... SELECT uses an arbitrary select, it will also sometimes exercise this path. We already have properties like TableHasExpectedContent which should then catch the same issue as this property would. Make sense?

[LeMikaelF]

Yes, absolutely. I realized this just yesterday. I'll make this PR a draft while I modify it.

…

________________________________ De : Jussi Saurio ***@***.***> Envoyé : 13 octobre 2025 07:15 À : tursodatabase/turso ***@***.***> Cc : Mikaël Francoeur ***@***.***>; Author ***@***.***> Objet : Re: [tursodatabase/turso] Simulator: add InsertSelectNested property (PR #3625) [https://avatars.githubusercontent.com/u/22619982?s=20&v=4]jussisaurio left a comment (tursodatabase/turso#3625)<#3625 (comment)> I don't think this needs to be a separate property. Our arbitrary SELECT should sometimes generate subqueries, and since the existing INSERT ... SELECT uses an arbitrary select, it will also sometimes exercise this path. We already have properties like TableHasExpectedContent which should then catch the same issue as this property would. Make sense? — Reply to this email directly, view it on GitHub<#3625 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AFO5REUGVTKB32ARGN477O33XOCWZAVCNFSM6AAAAACIRG2XHSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTGOJXGA3DSOBSGE>. You are receiving this because you authored the thread.Message ID: ***@***.***>