Load a TileDB DataFrame with Dask Delayed - Example with a timeseries (Slicing on DateTimeIndex)

tildedb_dask_dataframe_example.py

import dask
import numpy as np
import pandas as pd
import dask.dataframe as ddf

dask_chunks = 20

# Partition the dataframe
slices = [
    np.datetime64(d)
    for d in list(pd.date_range(datetm_start, datetm_end, periods=dask_chunks))
]
slices = zip(slices[:-1], slices[1:])

def slice_tiledb(sl):
    tdf = tiledb.open("s3://bucker-timeseries/main", mode="r", ctx=ctx)
    ddd = tdf.query(attrs=columns_to_select, use_arrow=True).df[
        (
            slice(*sl)
        )
    ]
    return ddd

delayed_slices = [dask.delayed(slice_tiledb)(sl) for sl in slices]

df = ddf.from_delayed(delayed_slices)