技能 数据科学 Foundry本地开发与测试循环

Foundry本地开发与测试循环

v20260423
palantir-local-dev-loop
本指南指导用户搭建Palantir Foundry的本地开发环境。它允许开发者使用PySpark在本地、使用样本数据运行数据转换和单元测试,并通过Mock Foundry API进行集成测试。这确保了代码的快速、独立迭代,适用于构建复杂的ETL数据管道逻辑。
获取技能
93 次下载
概览

Palantir Local Dev Loop

Overview

Set up local development for Palantir Foundry integrations. Covers running transforms locally against sample data, mocking the Foundry API for fast iteration, and testing with pytest before pushing to Foundry.

Prerequisites

  • Completed palantir-install-auth setup
  • Python 3.9+ with pip
  • A Foundry Code Repository cloned locally (or a standalone project)

Instructions

Step 1: Project Structure

my-foundry-project/
├── src/myproject/
│   ├── __init__.py
│   ├── pipeline.py          # @transform functions
│   └── utils.py             # Shared logic
├── tests/
│   ├── conftest.py           # Fixtures with sample DataFrames
│   ├── test_pipeline.py      # Transform unit tests
│   └── sample_data/          # CSV/Parquet test fixtures
├── .env                      # FOUNDRY_HOSTNAME, FOUNDRY_TOKEN
├── requirements.txt          # foundry-platform-sdk, pytest, pyspark
└── pyproject.toml

Step 2: Install Local Dependencies

set -euo pipefail
pip install foundry-platform-sdk pyspark pytest pandas
python -c "import foundry; import pyspark; print('Dependencies ready')"

Step 3: Test Transforms Locally with PySpark

# tests/conftest.py
import pytest
from pyspark.sql import SparkSession

@pytest.fixture(scope="session")
def spark():
    return SparkSession.builder.master("local[2]").appName("test").getOrCreate()

@pytest.fixture
def sample_orders(spark):
    data = [
        ("ORD-001", "alice@company.com", "2026-03-01", 99.99),
        ("ORD-002", "bob@test.com", "2026-03-02", 49.99),      # test email
        (None, "carol@company.com", "2026-03-03", 149.99),       # null ID
    ]
    return spark.createDataFrame(data, ["order_id", "email", "order_date_str", "total"])
# tests/test_pipeline.py
from myproject.pipeline import clean_orders

def test_clean_orders_removes_nulls_and_test_emails(sample_orders):
    result = clean_orders(sample_orders)
    assert result.count() == 1  # Only alice remains
    assert result.columns == ["order_id", "email", "order_date", "total_cents"]
    row = result.first()
    assert row.total_cents == 9999  # 99.99 * 100

Step 4: Mock Foundry API for Integration Tests

# tests/test_api.py
import pytest
from unittest.mock import MagicMock, patch

def test_list_ontology_objects():
    mock_client = MagicMock()
    mock_client.ontologies.OntologyObject.list.return_value.data = [
        MagicMock(properties={"fullName": "Alice", "department": "Engineering"}),
    ]

    result = mock_client.ontologies.OntologyObject.list(
        ontology="test", object_type="Employee", page_size=10
    )
    assert len(result.data) == 1
    assert result.data[0].properties["fullName"] == "Alice"

Step 5: Run Tests

set -euo pipefail
pytest tests/ -v --tb=short
# Expected: all tests pass against local Spark + mocked API

Step 6: Live API Smoke Test (Optional)

# scripts/smoke_test.py — runs against real Foundry (needs credentials)
import os, foundry, sys

client = foundry.FoundryClient(
    auth=foundry.UserTokenAuth(
        hostname=os.environ["FOUNDRY_HOSTNAME"],
        token=os.environ["FOUNDRY_TOKEN"],
    ),
    hostname=os.environ["FOUNDRY_HOSTNAME"],
)

try:
    ontologies = list(client.ontologies.Ontology.list())
    print(f"Smoke test passed: {len(ontologies)} ontologies accessible")
except foundry.ApiError as e:
    print(f"Smoke test failed: {e.status_code} {e.message}", file=sys.stderr)
    sys.exit(1)

Output

  • Local PySpark environment for testing transforms without Foundry
  • Mocked Foundry API client for integration tests
  • pytest suite validating pipeline logic
  • Optional live smoke test for credential verification

Error Handling

Error Cause Solution
Java not found (PySpark) JDK not installed Install JDK 11+: apt install openjdk-11-jdk
ModuleNotFoundError: pyspark Missing dependency pip install pyspark
Import error on transform functions Circular imports Keep transforms in separate modules
Spark AnalysisException Column name mismatch Print df.columns in test to debug

Examples

Watch Mode with pytest-watch

pip install pytest-watch
ptw tests/ -- -v --tb=short
# Re-runs tests on every file save

Resources

Next Steps

  • Apply SDK patterns: palantir-sdk-patterns
  • Build data pipelines: palantir-core-workflow-a
信息
Category 数据科学
Name palantir-local-dev-loop
版本 v20260423
大小 5.38KB
更新时间 2026-04-28
语言