Querying Azure storage using DuckDB

DuckDB just added support for fsspec, which make Querying Object store Like GCP and Azure storage possible, please notice AWS S3 API was natively supported already.

Previously to Query Azure storage, you had to use pyarrow dataset as a workaround, with the recent update, it is no more needed.

Here is a simple example, Querying a folder of Parquet files partitioned using Hive style, notice DuckDB is smart enough to recognize Date as a partition field

import duckdb
import adlfs ,os
from dotenv import load_dotenv
table_path = os.getenv('table_path') 
fs = adlfs.AzureBlobFileSystem(account_name=AZURE_STORAGE_ACCOUNT_NAME, account_key=AZURE_STORAGE_ACCOUNT_KEY )
con = duckdb.connect()
df = con.execute(f'''
    select *
    from read_parquet('{table_path}/scada/data/*/*.parquet', hive_partitioning=true)
    limit 10

and here is the result

Make sure to have the file”.env” when running the notebook from your computer, here is an example how it looks like

As a PowerBI user, I see a potential for a lightweight ETL process using just python that do complex transformation and output the results as a parquet files which PowerBI can consume.

As I deal with small data( less than 30 GB), Apache Spark does not make much sense to me, Hopefully Synapse will provide us with a cheap single node Notebook experience. I suspect it may be useful for a lot of customers.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: