Data Transformation · Topic 10 of 14
Development Workflows, Workspaces and Notebooks
Choose the right Snowflake development tools: Snowsight Workspaces, Snowflake Notebooks, Snowflake CLI and Git, and manage code through dev, test and prod.
- 4 min
- Workspaces replaced Worksheets; they hold SQL, Python and notebooks and can connect to Git.
- Notebooks mix SQL, Python and Markdown cells; warehouse runtime versus container runtime.
- Production code comes from Git through a pipeline, with separate dev, test and prod environments.
What it is
A development workflow is how code gets from an idea to production: write it, version it, test it, promote it. Snowflake gives a data engineer several places to write code and several ways to keep it under control. The exam asks which tool fits a given situation, and what makes the process repeatable.
| Tool | Best for |
|---|---|
| Snowsight Workspaces | Files and folders in the browser: SQL files, Python files, notebooks, Git backed work |
| Snowflake Notebooks | Interactive, cell by cell development mixing SQL, Python (Snowpark) and Markdown |
Snowflake CLI (snow) | Scripting and CI/CD from a terminal or pipeline |
| Local IDE, Python APIs, drivers | Development on your own machine |
| Git repository | Source of truth for all code |
How it works
Code management principles
- Code lives in Git, not in worksheets. Objects are created from versioned scripts, so a deployment can be repeated and audited.
- Separate environments: dev, test and prod, usually separate databases or accounts, each with its own roles.
- Promotion by pipeline, not by hand: the same scripts run in each environment with different parameters.
- Small, reviewable changes on branches, merged through pull requests.
The next pages cover Git, dbt, deployment, testing and environments in detail.
Snowsight Workspaces
Workspaces is the unified editor in Snowsight. It replaced the older Worksheets, and you cannot switch back.
- A workspace holds files and folders. My Workspace is private to you; shared workspaces are for collaboration and have an owner role.
- A SQL file runs with the role and warehouse you pick for it.
- Python files and notebooks can live in the same workspace, so one place holds the whole project.
- A workspace can be connected to a Git repository, so you can work with branches without leaving Snowsight. The current role decides which API integrations and secrets a Git workspace can use.
When a scenario says "a team needs to edit the same project files in the browser and keep them in Git", pick a shared, Git connected workspace, not copying SQL between worksheets.
Snowflake Notebooks
A notebook is an interactive document of SQL, Python and Markdown cells that run one at a time and show results inline.
- Output of a SQL cell can be used in a Python cell, and the other way around, so you can mix SQL transforms with Snowpark and charts.
- Runtimes: the warehouse runtime runs the Python kernel on the notebook warehouse and starts quickly. The container runtime runs on a Snowpark Container Services compute pool and suits heavier, ML style work with CPU or GPU images. In both, SQL and Snowpark queries still run on a warehouse.
- Packages come from the Snowflake Anaconda channel or a stage (warehouse runtime), or from pip and conda (container runtime).
- Visualization libraries such as Streamlit, Altair and Matplotlib are available.
- A notebook can be scheduled to run as a task, which turns exploration into a pipeline step. Notebooks can also integrate with Git. Scheduling and notebook Git integration have been marked as preview features, so do not rely on them as the only answer.
- Notebooks follow role based access control: ownership is needed to edit one.
| Need | Choose |
|---|---|
| Explore data, chart results, document steps | Notebook |
| Reusable production transformation | Version controlled SQL, a stored procedure or a dbt project, run by a task |
| Heavy Python or GPU training | Notebook on container runtime |
A notebook is a good development surface, but a production pipeline should run from version controlled code on a schedule or trigger, not from an open notebook.
Command line and CI/CD
The Snowflake CLI runs SQL files (snow sql -f), deploys Snowpark objects (snow snowpark deploy) and dbt projects (snow dbt deploy), and fits into GitHub Actions or similar tools. It reads connection settings from a configuration file or environment variables, so no credentials sit in the code.
Common mistakes
Keeping the only copy of code in a worksheet or notebook. Without Git there is no history, review or rollback.
Developing against production data. Use a development database, often a zero copy clone, with its own role.
Believing notebook cells replace a scheduler. Schedule the notebook or, better, a task running versioned code.
Mixing up the two notebook runtimes. Warehouse runtime is lightweight and quick; container runtime is for heavier compute and extra packages.
What to remember for the exam
- Workspaces replaced Worksheets; they hold SQL, Python and notebooks and can connect to Git.
- Notebooks mix SQL, Python and Markdown cells; warehouse runtime versus container runtime.
- Production code comes from Git through a pipeline, with separate dev, test and prod environments.
- Snowflake CLI (
snow) is the scripting and CI/CD tool.
Check yourself
Which notebook runtime suits GPU based model work?
The container runtime, which runs on a Snowpark Container Services compute pool.
Why keep pipeline code in Git rather than only in Snowsight?
Git gives history, review, rollback and a repeatable source for deployments to each environment.
Which tool lets a CI pipeline run SQL files and deploy Snowpark or dbt objects?
The Snowflake CLI (snow).
Topic 10 of 14 in Data Transformation
Test yourself on Data Transformation
The full Snowflake DEA-C02 question bank has 709 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by Snowflake. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.