Will ducklake v1.0 ship with full inlining support?

[ccrouch]

I saw a recent comment to the affirmative, but I wanted to double check...

#618 (reply in thread)

I've done some basic experimentation with ducklake (ext version 3f1b372) and duckdb v1.4.4 as the catalog and inlining appears to work for inserts and deletes but not updates. Also when using MERGE INTO with INSERT that also does not seem to inline (at least according to the test codex said it ran :-) ). I'm 🤞 v1.0 and duckdb v1.5 is going fix everything 😄 .

Thanks very much
Charles

[pdet]

Yes, the plan is for Ducklake V1.0 to fully support inline (i.e., insertions, deletions, and updates); it will also default to using inlining.

[ccrouch]

Awesome, thanks @pdet! Quick follow up, will the inline support include both DuckDB and Postgres catalogs?

[pdet]

Absolutely!

[jpdenford]

Hi @pdet, I can't say how excited I am for v1.0 - love what you guys are doing.

Do you know whether V1.0 will support inlining of records that include a variant type/column in Postgres?
If not, do you have plans to support that at some stage?

Thanks

[pdet]

Hi @jpdenford, indeed, variant inlining won't be shipped for non-duckdb catalogs. We do plan on implementing it, but I don't have a specific deadline in mind yet :-)

[jpdenford]

Cool, thank you 👍
Just to clarify, is the current behaviour (for non-duckdb catalogs) to write records immediately to object storage if the table contains any variant columns (as though the data_inlining_row_limit had been set to 0 for that table)?

[pdet]

Precisely, if the table has a Variant type, the system should detect that and not inline it.