ygg 0.6.4

Type-friendly utilities for moving data between Python objects, Arrow, Polars, Pandas, Spark, and Databricks

Yggdrasil (Python)

Type-friendly utilities for moving data between Python objects, Arrow, Polars, pandas, Spark, and Databricks. The package bundles enhanced dataclasses, casting utilities, and lightweight wrappers around Databricks and HTTP clients so Python/data engineers can focus on schemas instead of plumbing.

When to use this package

Use Yggdrasil when you need to:

• Convert payloads across dataframe engines without rewriting type logic for each backend.
• Define dataclasses that auto-coerce inputs, expose defaults, and surface Arrow schemas.
• Run Databricks SQL jobs or manage clusters with minimal boilerplate.
• Add resilient retries, concurrency helpers, and dependency guards to data pipelines.

Prerequisites

• Python 3.10+
• uv for virtualenv and dependency management.

Optional extras:

• polars, pandas, pyarrow, and pyspark for engine-specific conversions.
• databricks-sdk for workspace, SQL, jobs, and compute helpers.
• msal for Azure AD authentication when using MSALSession.

Installation

From the python/ directory:

uv venv .venv
source .venv/bin/activate
uv pip install -e .[dev]

Extras are grouped by engine:

• .[polars], .[pandas], .[spark], .[databricks] – install only the integrations you need.
• .[dev] – adds testing, linting, and typing tools (pytest, ruff, black, mypy).

Databricks example

Install the databricks extra and run SQL with typed results:

from yggdrasil.databricks.workspaces import Workspace
from yggdrasil.databricks.sql import SQLEngine

ws = Workspace(host="https://<workspace-url>", token="<token>")
engine = SQLEngine(workspace=ws)

stmt = engine.execute("SELECT 1 AS value")
result = stmt.wait(engine)
tbl = result.arrow_table()
print(tbl.to_pandas())

Parallel processing and retries

from yggdrasil.pyutils import parallelize, retry

@parallelize(max_workers=4)
def square(x):
    return x * x

@retry(tries=5, delay=0.2, backoff=2)
def sometimes_fails(value: int) -> int:
    ...

print(list(square(range(5))))

Project layout

• yggdrasil/dataclasses – yggdataclass decorator plus Arrow schema helpers.
• yggdrasil/types – casting registry (convert, register_converter), Arrow inference, and default generators.
• yggdrasil/libs – optional bridges to Polars, pandas, Spark, and Databricks SDK types.
• yggdrasil/databricks – workspace, SQL, jobs, and compute helpers built on the Databricks SDK.
• yggdrasil/requests – retry-capable HTTP sessions and Azure MSAL auth helpers.
• yggdrasil/pyutils – concurrency and retry decorators.
• yggdrasil/ser – serialization helpers and dependency inspection utilities.
• tests/ – pytest-based coverage for conversions, dataclasses, requests, and platform helpers.

Testing

From python/:

pytest

Optional checks when developing:

ruff check
black .
mypy

Troubleshooting and common pitfalls

• Missing optional dependency: Install the matching extra (e.g., uv pip install -e .[polars]) or wrap calls with require_polars/require_pyspark from yggdrasil.libs.
• Schema mismatches: Use arrow_field_from_hint and CastOptions to enforce expected Arrow metadata when casting.
• Databricks auth: Provide host and token to Workspace. For Azure, ensure environment variables align with your workspace deployment.

Contributing

1. Fork and branch.
2. Install with uv pip install -e .[dev].
3. Run tests and linters.
4. Submit a PR describing the change and any new examples added to the docs.

Experimental native acceleration

For hot paths, an optional Rust extension scaffold lives in rust/. Use yggdrasil.rs to access the fast path with a built-in Python fallback so base installs remain pure-Python compatible.