Thijs

Results 146 comments of Thijs

Test seems to work: ``` arrow_load_lineitem Old timing: 1.3658527135849 New timing: 1.37[26](https://github.com/duckdb/duckdb/actions/runs/7932419552/job/21658866927?pr=10715#step:11:27)81975364685 arrowtpch Old timing: 5.479[28](https://github.com/duckdb/duckdb/actions/runs/7932419552/job/21658866927?pr=10715#step:11:29)1306266785 New timing: 5.437824249267578 native_load_lineitem Old timing: 44.80615174770355 New timing: 43.87466597557068 pandas_load_lineitem Old timing: 2.689911961555481...

Yea you're right, I noticed that after making the PR I'll add a test there

We're already performing similar logic in s3fs.cpp `S3FileSystem::GetQueryParam` Maybe we should allow the FileSystem implementation to clean the file_path, then we can perform the compression detection Or the compression detection...

I wonder if the autoinstall and autoload settings are possibly disabled ? ``` >>> duckdb.sql("select current_setting('autoinstall_known_extensions')") ┌─────────────────────────────────────────────────┐ │ current_setting('autoinstall_known_extensions') │ │ boolean │ ├─────────────────────────────────────────────────┤ │ false │ └─────────────────────────────────────────────────┘ >>> duckdb.sql("select...

Please check if that's the case or not @adenosinew Maybe something was wrongly configured creating those binaries

Sounds related to this discussion? https://discord.com/channels/909674491309850675/921073327009853451/1202313395466928138 You might need to also purge the metadata for us to be able to reclaim the lost space

> I will open a PR for this. Please don't do that today, we're trying to not put unnecessary pressure on the CI as we're looking to release later today

That sounds like a data ownership issue, heap-use-after-free if I'd have to guess I'd be curious what data you're consuming I also don't think the result collection method is required...

what do you mean by interning? The part it crashes at is constructing the materialized result, which performs a copy But until then it assumes it has (read only) ownership...

@mattaubury that sounds like our dictionary vectortype but I don't think we produce that from an arrow scan