radaroptions 0.1.0

DuckDB-based reader for an options-chain data lake stored in Azure Data Lake Storage

radaroptions

DuckDB-based reader for an options-chain data lake stored in Azure Data Lake Storage Gen2. No JVM, no Spark cluster required - reads partitioned parquet directly via DuckDB's azure extension.

Auth is picked automatically based on which environment variable is set:

• A read-only SAS URL (RADAR_LAKE_SAS_URL) - the normal way to get access. One single value someone gives you (bundles account + container + token) - no Azure account needed at all.
• credential_chain - an Azure CLI session (az login) or a service principal (AZURE_TENANT_ID / AZURE_CLIENT_ID / AZURE_CLIENT_SECRET), for anyone with their own Azure AD identity that has read access to the storage account (set RADAR_LAKE_ACCOUNT / RADAR_LAKE_CONTAINER too in this case, since there's no URL to derive them from).

Install

pip install radaroptions

Configure

Installing the package doesn't grant you access. No access yet? Email [e-mail].

RADAR_LAKE_SAS_URL=<the URL they give you>

That's the only variable most people need. Running a read without it set fails with a clear message pointing back to this contact.

Usage

Python

from radaroptions import read_day, read_range

df = read_day("2026-06-11")

df = read_range(
    "2026-01-01", "2026-06-30",
    columns=["strike", "option_type", "open_interest", "gamma", "current_price", "time"],
)

CLI

radaroptions read-day 2026-06-11 --out day.parquet
radaroptions read-range 2026-01-01 2026-06-30 -c strike,option_type,open_interest
radaroptions query "SELECT count(*) FROM read_parquet('abfss://<container>@<account>.dfs.core.windows.net/2026/06/*/*.parquet')"

read_range globs one file pattern per (year, month) covering the requested dates and filters on the data's own time column, using union_by_name=true so columns added or renamed over the years don't break the read.

Troubleshooting

• TLS / proxy errors on Windows: set RADAR_LAKE_AZURE_TRANSPORT=curl to switch DuckDB's azure extension from WinHTTP to curl.
• Schema errors reading a date range: a genuine type change in the underlying data (e.g. a column that used to be decimal and is now double) can still error even with union_by_name=true - pass columns=[...] to select only the fields you need.

License

Proprietary - see LICENSE. Installable for authorized use only; no rights are granted to copy, modify, or redistribute this software.