Snowflake Notebooks for Analytics and Sharing: SQL/Python, Snowpark, Git

Snowflake Notebooks is a managed environment inside Snowsight for interactive data analysis where you switch between SQL, Python, and Markdown on a per-cell basis. It brings together Snowpark DataFrame integration, built-in visualization, Git-based version control, and team collaboration, all within Snowflake's security model.

Notebooks Architecture

Snowflake Notebooks runs on the following components:

• Front end: The notebook editor inside Snowsight (web UI). You add, run, and visualize cells right in the browser.
• SQL runtime: SQL cells run on the warehouse you specify, using the same execution engine as regular Snowflake queries.
• Python runtime: Python cells run on Snowpark's server-side runtime, backed by either a warehouse or an SPCS compute pool.
• Storage: The notebook itself is stored as an object in a Snowflake schema. Data never leaves Snowflake.

Creating a Notebook and Basic Operations

You can create a notebook from the Snowsight UI or with SQL.

-- SQLでNotebookを作成
CREATE OR REPLACE NOTEBOOK analytics.notebooks.monthly_report
  FROM '@analytics.notebooks.nb_stage'
  MAIN_FILE = 'monthly_report.ipynb'
  QUERY_WAREHOUSE = analytics_wh
  COMMENT = '月次レポート分析ノートブック';

-- Notebookの一覧確認
SHOW NOTEBOOKS IN SCHEMA analytics.notebooks;

-- Notebookの権限付与
GRANT USAGE ON NOTEBOOK analytics.notebooks.monthly_report
  TO ROLE analyst_role;

Running SQL Cells

SQL cells let you run regular Snowflake SQL as-is. Results are shown as a table and can be switched to a chart view.

-- SQLセルの例: 売上サマリ
SELECT
  DATE_TRUNC('MONTH', order_date) AS order_month,
  product_category,
  COUNT(*) AS order_count,
  SUM(amount) AS total_revenue,
  AVG(amount) AS avg_order_value
FROM sales.orders
WHERE order_date >= '2025-01-01'
GROUP BY order_month, product_category
ORDER BY order_month, total_revenue DESC;

Working with Python Cells

Python cells let you process data with Snowpark DataFrames, and analyze or visualize using libraries like pandas, matplotlib, or plotly. There's also a built-in mechanism to reference SQL cell results from Python cells.

# Pythonセルの例: Snowpark DataFrameによる集計
from snowflake.snowpark.context import get_active_session
from snowflake.snowpark.functions import col, sum as sf_sum, count

session = get_active_session()

orders_df = session.table("sales.orders")

monthly_summary = (
    orders_df
    .filter(col("order_date") >= "2025-01-01")
    .group_by("product_category")
    .agg(
        count("*").alias("order_count"),
        sf_sum("amount").alias("total_revenue")
    )
    .sort(col("total_revenue").desc())
)

monthly_summary.show()

Bridging SQL and Python Cells

SQL cell results are accessible by the cell's name, and you can reference them from a Python cell using that name.

# cell_name_of_sql_cell がSQLセルの名前の場合:
# SQLの結果をSnowpark DataFrameとして取得
df = cell_name_of_sql_cell.to_df()

# SQLの結果をpandas DataFrameとして取得
pdf = cell_name_of_sql_cell.to_pandas()

# pandasで追加分析
print(f"行数: {len(pdf)}")
print(pdf.describe())

# matplotlibで可視化
import matplotlib.pyplot as plt

pdf.plot(kind='bar', x='PRODUCT_CATEGORY', y='TOTAL_REVENUE')
plt.title('カテゴリ別売上')
plt.tight_layout()
plt.show()

Built-in Visualization

SQL cell results can be visualized with Snowsight's built-in chart features in addition to the table view.

• Bar / line charts: Time-series trends and category comparisons
• Scatter plots: Correlation between two variables
• Heatmaps: Distribution of multi-dimensional data
• Area charts: Changes in composition over time

Python cells can also use libraries such as matplotlib, plotly, and Altair for more advanced custom visualizations.

# Pythonセルでのplotlyによるインタラクティブ可視化
import plotly.express as px

pdf = cell_monthly_data.to_pandas()

fig = px.line(
    pdf,
    x='ORDER_MONTH',
    y='TOTAL_REVENUE',
    color='PRODUCT_CATEGORY',
    title='月次売上推移（カテゴリ別）'
)
fig.show()

Git Integration

Snowflake Notebooks supports Git repository integration, enabling notebook version control and team-based development.

-- Git Repository Integrationの作成
CREATE OR REPLACE GIT REPOSITORY analytics.notebooks.ml_repo
  API_INTEGRATION = github_api_integration
  GIT_CREDENTIALS = github_secret
  ORIGIN = 'https://github.com/myorg/ml-notebooks.git';

-- Gitリポジトリからノートブックをフェッチ
ALTER GIT REPOSITORY analytics.notebooks.ml_repo FETCH;

-- リポジトリ内のファイル一覧
SHOW GIT BRANCHES IN analytics.notebooks.ml_repo;
ls @analytics.notebooks.ml_repo/branches/main/;

-- Gitリポジトリからノートブックを作成
CREATE OR REPLACE NOTEBOOK analytics.notebooks.feature_engineering
  FROM '@analytics.notebooks.ml_repo/branches/main/'
  MAIN_FILE = 'feature_engineering.ipynb'
  QUERY_WAREHOUSE = ml_wh;

Collaboration and Access Control

Notebook permissions are managed by Snowflake's RBAC. Multiple users can edit the same notebook, with changes synchronized in real time.

-- アナリストロールにノートブックの使用権限を付与
GRANT USAGE ON NOTEBOOK analytics.notebooks.monthly_report
  TO ROLE analyst_role;

-- データサイエンティストロールにノートブックの所有権を付与
GRANT OWNERSHIP ON NOTEBOOK analytics.notebooks.ml_experiment
  TO ROLE data_scientist_role;

-- ウェアハウスの使用権限（ノートブック実行に必要）
GRANT USAGE ON WAREHOUSE analytics_wh
  TO ROLE analyst_role;

Package Management

Python cells can add packages from Snowflake's Anaconda channel. You can search for and add packages directly from the notebook UI.

# ノートブック内でのパッケージ追加（Pythonセル）
# Snowsightの「Packages」パネルからGUIで追加するか、以下のようにセッションで追加
session.add_packages('pandas', 'scikit-learn==1.4.0', 'plotly')

# requirements.txtからの一括追加
# session.add_requirements('requirements.txt')

Scheduling Notebook Runs

By scheduling notebooks as tasks, you can automate periodic report generation and model retraining.

-- ノートブックをスケジュール実行するタスク
CREATE OR REPLACE TASK run_monthly_report
  WAREHOUSE = analytics_wh
  SCHEDULE = 'USING CRON 0 6 1 * * Asia/Tokyo'
AS
  EXECUTE NOTEBOOK analytics.notebooks.monthly_report;

ALTER TASK run_monthly_report RESUME;

Security and Governance

• No data exfiltration: Analytics data stays inside Snowflake, preventing downloads to the local machine.
• RBAC integration: Notebook execution results are filtered based on the user's role permissions.
• Masking policy enforcement: Tables with Dynamic Data Masking are also displayed with masks inside notebooks.
• Audit logs: Notebook execution history is recorded in ACCOUNT_USAGE.QUERY_HISTORY.

What the Exam Asks About

• Notebooks run inside Snowsight, and data never leaves Snowflake.
• You can switch between SQL / Python / Markdown on a per-cell basis.
• SQL cell results can be received as a DataFrame in a Python cell.
• Git integration supports version control and team-based development.
• Running a notebook requires USAGE permission on the warehouse.
• Snowflake's RBAC, masking policies, and governance apply directly.

Check with a Practice Question

Frequently Asked Questions