Add estimated cardinality support for table functions

[jdaigle]

Adds an optional EstimatedCardinality property to the TableFunction record. When set, the bind callback calls duckdb_bind_set_cardinality to hint the query optimizer. Currently only available through the low-level API where the caller constructs the TableFunction directly. Only estimated (non-exact) cardinality is supported.

Implements #322

[coveralls]

Coverage Report for CI Build 24316052415

Coverage increased (+0.01%) to 89.607%

Details

• Coverage increased (+0.01%) from the base build.
• Patch coverage: 4 of 4 lines across 1 file are fully covered (100%).
• No coverage regressions found.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	2910
Covered Lines:	2661
Line Coverage:	91.44%
Relevant Branches:	1487
Covered Branches:	1279
Branch Coverage:	86.01%
Branches in Coverage %:	Yes
Coverage Strength:	451168.6 hits per line

---

💛 - Coveralls

[Giorgi]

Thanks @jdaigle Why did you decide to limit it to estimate only? Why not allow callers pass exact number too?

Is it ok to set both for a table function at the same time?

[jdaigle]

@Giorgi I personally didn't have a use case for "exact". But I'm happy to improve this to support that use case.

Although I'm not entirely sure how it is supposed to work with "exact". Are we supposed to call duckdb_bind_set_cardinality just once passing true instead of false, or if we have both an "estimate" and "exact" are we supposed to call it twice?

[Giorgi]

Is setting both useful? If we know an exact number, why set an estimate too?

Setting it twice will overwrite the previous value: https://github.com/duckdb/duckdb/blob/main/src%2Fmain%2Fcapi%2Ftable_function-c.cpp#L386-L396

[jdaigle]

@Giorgi oh I see now, I didn't think to check the source code. I've updated to use a record type for both the value and boolean. Thanks!