lance-format / lance-format/lance
[BUG](python): The order of the table was changed after executing the `compact_files` operation
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 7.1k
- Forks
- 852
- Avg merge
- 3d 18h
- Merged PRs (30d)
- 272
Description
Lance Version
pylance 0.20.0
What happened
The order of the dataset was changed after executing the compact_files operation.
My code is as follows:
import lance
from lance.dataset import DatasetOptimizer
DB_PATH = "<path-to-my-dataset>"
def main():
dataset = lance.dataset(DB_PATH)
print(dataset.take([0,1,2,3,4,5,6,7,8,9]) # show the first ten elements
# compact the dataset
optim = DatasetOptimizer(dataset)
optim.compact_files(num_threads=8)
print(dataset.take([0,1,2,3,4,5,6,7,8,9]) # the first ten elements was changed
return
if __name__ == "__main__":
main()
My enviroment
I am using a ubuntu server with 64 cores and 512G memory.
The dataset has 5 columns: title(str), section(str), text(str), id(str), and vector(list[float]).
How to reproduce
This dataset has 38 Million records of 768 dim vector and payload. I'm not sure if its feasible to share the dataset.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start with the Python DatasetOptimizer.compact_files operation and compare dataset.take results before and after compaction. Investigate how the operation represents record order, then reproduce the reported change with a smaller dataset if possible. Done means the ordering behavior is understood and a regression test covers the intended result.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- databases
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 30/100