Tried filtering on a timestamp column using duckdb against lance, and it did not work.
The same query worked against a parquet file.
import pyarrow as pa
import lancedb
import duckdb
from datetime import datetime, timedelta
data = pa.table({
"c1": ["A", "B"],
"c2": ["a", "b"],
"c3": [1, 2],
"c4": [datetime(2021,6,16),datetime(2021,6,17)]
})
db = lancedb.connect("lancedb_files")
db.create_table("lance_table", data)
table = db.open_table("lance_table")
arrow_dataset = table.to_lance()
duckdb.query("select * from arrow_dataset where c4='2021-06-16 00:00:00.000000'").df()
pa.dataset.write_dataset(data, "parquet_files", format='parquet')
duckdb.query("select * from 'parquet_files/*.parquet' where c4='2021-06-16 00:00:00.000000'").df()
| index |
c1 |
c2 |
c3 |
c4 |
| 0 |
A |
a |
1 |
2021-06-16 00:00:00 |
c1: string
c2: string
c3: int64
c4: timestamp[us]
Not sure if this is a bug that is to be reported here, or if it should be filed with duckdb .
Tried filtering on a timestamp column using duckdb against lance, and it did not work.
The same query worked against a parquet file.
c1: string
c2: string
c3: int64
c4: timestamp[us]
Not sure if this is a bug that is to be reported here, or if it should be filed with duckdb .