query hangs forever using views, completes immediately with equivalent tables about duckdb HOT 4 CLOSED

So, the script eventually finishes, so perhaps "never finishes" isn't quite correct.

I think it's strange though that the parquet version is so much slower than the native version:

• Native: 0.3s
• Parquet: 34s

100x difference seems pretty rough :)

Bumping up the scale factor to 0.1 things start to go haywire (and the massive .tmp directory gets created):

• Native: 0.3s
• Parquet: ? (still waiting on this to finish, up to 445GB of .tmp now)

from duckdb.

Well with sf=0.1, I got this error message eventually:

duckdb.duckdb.OutOfMemoryException: Out of Memory Error: failed to offload data block of size 256.0 KiB (791.4 GiB/791.4 GiB used).

from duckdb.

Hi Phillip, thanks for opening this. Did you try it on the latest nightly build (pip install duckdb --pre --upgrade)?

$ time python my.py

starting query
python3.11 my.py  0.97s user 0.22s system 245% cpu 0.484 total

from duckdb.

Excellent, seems to be fixed on nightly. 🎉

Any idea what the fix was?

from duckdb.