This is a Visual Studio Code extension that allows you to work with Databricks locally from VSCode in an efficient way, having everything you need integrated into VS Code - see Features. It allows you to manage and execute your notebooks, start/stop clusters, execute jobs and much more!

The extensions can be downloaded from the official Visual Studio Code extension gallery: Databricks VSCode

• Workspace Manager
  • Up-/download of notebooks and whole folders
  • Compare/Diff of local vs online notebook
  • Support for Code Cells if you do not want to use the .ipynb format
• Cluster Manager
  • Start/stop clusters
  • Script cluster definition as JSON
  • Open in OS File Explorer
• Notebook Kernel
  • Execute local code against a running Databricks cluster
  • interactive cell-by-cell execution as in Databricks web UI
  • rich output and visualization of results
• File System Integration
  • integrate DBFS and Databricks Workspace/notebooks next to your local file system
  • easy drag&drop between local, DBFS and also workspace/notebooks!
  • similar, well-known UI; same behavior as if DBFS and workspace/notebooks existed locally
• Connection Manager
  • Easy configuration via standard VS Code settings
  • Manage and switch between multiple Databricks workspaces (e.g. DEV/TEST/PROD)
  • Supports Azure Databricks, Databricks on AWS and Dtabricks on GCP
  • control how notebooks are downloaded (Jupyter notebook, source code, ...)
  • various other settings
  • Load Databricks directly from your Azure Account
• SQL / Data Browser
  • Browse availabl SQL objects from the Databricks metastore
  • databases, tables and views, columns, ...
• Job Manager
  • Start/stop jobs
  • View job-run history + status
  • Script job definition as JSON
  • View job-run output as JSON
• DBFS Browser
  • Supports static mount points (e.g. using Service Principals)
  • Up-/Download of files
• Secret Manager
  • Create/delete secret scopes
  • Create/delete secrets
  • view Azure Key Vault backed secret scopes
• Repo Manager
  • Create/delete linked repositories
  • Update repository to repo or branch
• Integration for CI/CD using DatabricksPS PowerShell module

v1.4.0:

• fixed an issue with the Create cluster link
• rework Azure Connection Manager so it automatically uses existing crednetials without prompt
• added Open Explorer action to locally synced workspace items in Workspace Manager

v1.3.1:

• changed Azure Connection Manager to use VSCode authentication instead of Azure Account Extension
• fixed issue with downloading workspace folders

v1.3.0:

• added new Connection Manager to load Databricks workspaces directly from your Azure Account
• added support for _sqldf in Python/PySpark
• reworked API connection test

v1.2.5:

• fixed parsing of apiRootUrl when reading from the config (only taking schema+authority now)

v1.2.4:

• added Pin and Unpin to Cluster Manager
• minor fixes Cluster Manager
• improved input via textboxes

v1.2.3:

• Databricks Kernels are now moved to the top when opening a notebook from the local sync folder or via dbws:/
• added Insert buttons for Secrets to easily add the code snippet to the current editor/notebook
• fixed an issue with %run

v1.2.2:

• added Restart Kernel to Databricks notebooks
• added details specs of cluster to Notebook Kernel selector
• fixed an issue where the extension did not load when no default connection was specified
• fixed an issue with %run and absolute paths (#93)
• fixed an issue with Files in Repos (#101)
• fixed an issues with CLI Connection Manager (#99)
• fixed an issue when up-/downloading whole folders recursively in Workspace Manager
• prepare to make extension work in the web
  • reworked API connection to use generic Fetch API instead of Axios library
  • added build config for web-extension
  • removed all external dependencies

v1.1.4:

• added support for dbutils.notebook.run in Live-mode (via mounting dbws:/... in VSCode workspace)
• fixed some issues with legacy DBFS and Workspace browser
• fixed an issue with workspace browser file compare
• fixed an issue when opening a job in the browser
• fixed issue where Repo Manager did not use pagination of the API and hence did not show more than 20 entries

v1.1.3:

• added support for %run in notebooks
• added auto-refresh for new notebook kernels

v1.1.2:

• make File System Integration work without non-notebook files (e.g. Files in Repos)

v1.1.1:

• make File System Integration work without having to open the Databricks tab first
• fixed issues with File System Integration
  • folders with a single sub-folder only now work
  • fixed an issue where Repos did not work/could not be browsed properly
• switch focus to VSCode Explorer when adding DBFS/Workspace to the VSCode workspace
• added Pull Button to the Repo Manager to pull the latest changes from Git
• added proper HTML output for notebooks/executions that returned HTML as plain text
• fixed issue with Workspace Manager where files could not be downloaded properly anymore

v1.1.0:

• added File System APIs for dbfs:/ (same as in Databricks) and also dbws:/ for the workspace/notebooks
• added support for Files in Repos
• fixed issue with Execute All Cells in notebook kernel
• fixed issue with multiple notebooks connected to the same kernel
• added packaging of the source code with webpack for improved loading times

v1.0.0:

• added support for Notebook Kernel to allow execution of code on the Databricks cluster from within VSCode notebooks
• major improvements for useability and user-interaction
• performance improvements when refreshing tree-views
• improved logging and output of messages
• use native VSCode icons instead of custom ones

• Support for Drag&Drop of local files directly into the workspace
• Fix remaining issues with Notebook Kernel and code cells
• whatever is requested by the users :)

The extension can be downloaded directly from within VS Code. Simply go to the Extensions tab, search for "Databricks" and select and install the extension "Databricks VSCode" (ID: paiqo.databricks-vscode).

Alternatively it can also be downloaded the .vsix directly from the VS Code marketplace: Databricks VSCode.

Preview-Versions might also be available via github Releases from this repository.

The configuration happens directly via VS Code by simply opening the settings Then either search for "Databricks" or expand Extensions -> Databricks. The settings themselves are very well described and it should be easy for you to populate them. Also, not all of them are mandatory! Some of the optional settings are experimental or still work in progress. To configure multiple Databricks Connections/workspaces, you need to use the JSON editor and add them to databricks.connections:

		...
		"databricks.connectionManager": "VSCode Settings",
		"databricks.connections": [
			{
				"apiRootUrl": "https://adb-98765421987.65.azuredatabricks.net",
				"displayName": "My DEV workspace",
				"localSyncFolder": "c:\\Databricks\\dev",
				"personalAccessToken": "dapi219e30212312311c672bbbbbbbbbb"
			},
			{
				"apiRootUrl": "https://adb-123456789123.45.azuredatabricks.net",
				"displayName": "My TEST workspace",
				"localSyncFolder": "c:\\Databricks\\test",
				"personalAccessToken": "dapi219e30212312311c672aaaaaaaaaa"
			}
		],
		...

The localSyncFolder is the location of a local folder which is used to download/sync files from Databricks and work with them locally (notebooks, DBFS, ...). It also supports environment variables - e.g. %USERPROFILE%\\Databricks or ~\\Databricks. The sensitive values entered like personalAccessToken will be safely stored in the system key chain/credential manager (see databricks.sensitiveValueStore) once the configuration is read the first time. This happens if you open the extension. Existing connections can be updated directly in VSCode settigns or via the JSON editor. To update a personalAccessToken, simply re-enter it and the extension will update it in the system key chain/credential manager. The only important thing to keep in mind is that the displayName should be unique on the whole machine (across all VSCode workspaces) as the displayName is used to identify the personalAccessToken to load from the system key chain/credential manager.

Another important setting which requires modifying the JSON directly are the export formats which can be used to define the format in which notebooks are up-/downloaded. Again, there is a default/current setting databricks.connection.default.exportFormats and it can also configured per Connection under databricks.connections:

		...
		"databricks.connection.default.exportFormats": 
		{
			"Scala": ".scala",
			"Python": ".py.ipynb",
			"SQL": ".sql",
			"R": ".r"
		},
		...

Each filetype can either be exported as raw/source file (.scala, .py, .sql, .r) or, if supported, also as a notebook (.ipynb). This is also very important if you want to upload a local file as these also have to match these extension and will be ignored otherwise! For active development it is recommended to use .ipynb format as this also allows you to execute your local code against a running Databricks cluster - see Notebook Kernel.

If you prefer raw/source files (e.g. for better Git integration), you may also consider using Code Cells be setting "useCodeCells" = true for your corresponding connection. Please be aware that this does currently not work properly with the Notebook Kernel!

All these settings can either be configured on a global/user or on a workspace level. The recommendation is to use workspace configurations and then to include the localSyncFolders into your workspace for easy access to your notebooks and sync to GIT. Using a workspace configuration also allows you to separate different Databricks Connections completely - e.g. for different projects.

NOTE: Changing any of the connection settings only take effect once the connection is activated! If you make changes to your current connection, you need to activate another connection temporary and then the original one again! Alternatively, a restart of VSCode also works.

To use the Databricks CLI Connection Manager, you first need to install and configure the Databricks CLI. Once you have created a connection or profiles, you can proceed here. Basically all you need to do in VSCode for this extension to derive the connections from the Databricks CLI is to change the VSCode setting databricks.connectionManager to Databricks CLI Profiles. This can be done in the regular settings UI or by modifying the settings JSON directly.

In order to work to its full potential, the VSCode extension needs some addional settings which are not maintained by the Databricks CLI. Foremost the localSyncFolder to store files locally (e.g. notebooks, cluster/job definitions, ...). For the Databricks CLI Connection Manager this path defaults to <user home directory>/Databricks-VSCode/<profile name>. If you want to change this you can do so by manually extending your Databricks CLI config file which can usually be found at <user home directory>/.databrickscfg:

[DEV]
host = https://westeurope.azuredatabricks.net/
token = dapi219e30212312311c6721a66ce879e
localSyncFolder = D:\Desktop\sync\dev

[TEST]
host = https://westeurope.azuredatabricks.net/
token = dapi219e30212312311c672aaaaaaaaaa
localSyncFolder = D:\Desktop\sync\test
localSyncSubfolders = {"Workspace": "Workspace","Clusters": "Clusters","DBFS": "DBFS","Jobs": "Jobs"}
exportFormats = {"Scala": ".scala","Python": ".ipynb","SQL": ".sql","R": ".r"}
useCodeCells = true

You can also change the following other settings:

The Azure Connection Manager allows you to use your Azure Active Directory (AAD) account to interact with Databricks. This includes:

• loading Connections directly from your Azure Resources
• Use AAD authentication when using the Databricks API. No Personal Access Token (PAT) is needed!

VSCode will prompt you to use your Microsoft Account two times. The first time is to get a list of all available Azure Databricks workspaces that you have access to and then a second time to establish a connection to the selected/active workspace. Whenever you switch connection/workspace, you may get prompted again!

To activate the Azure Connection Manager, simply set the VSCode setting databricks.connectionManager to Azure and refresh your connections. No additional configurations need to be done. Currently most other connection settings like useCodeCells, exportFormats, etc. cannot currently be controlled and are set to their defaults.

The following Azure-specific settings exist and can be set in the workspace settings:

• databricks.azure.tenantId
• databricks.azure.subscriptionIds
• databricks.azure.workspaces

They are documented via VSCode settings documentation.

The extension supports various connection managers and the list can be easily extended. At the moment these connecton managers exist:

• VSCode Settings
• Databricks CLI
• Azure
• Manual where you are simply prompted to enter connection information at the start of your session.

You can specify the one to use by setting the VSCode setting databricks.connectionManager. Once the extension loads, you will see your list in the Connections view and icons indicating which one is currently active (the green one). To change the Connection, simply click the [Activate] button next to an inactive Connection. All other views will update automatically.

The Workspace Manager connects directly to the Databricks workspace and loads the whole folder strucuture. It displays folders, notebooks and libraries. Notebooks and folders can be up- and downloaded manually by simply clicking the corresponding item next them. If you do an up-/download on a whole folder or on the root, it will up-/download all items recursively. The files are stored in the databricks.connection.default.localSyncFolder (or your Connection) that you configured in your settings/for your Connection. If you doubleclick a file, it will be downloaded locally and opened. Depending on the ExportFormats that you have defined in databricks.connection.default.exportFormats (or your Connection), the item will be downloaded in the corresponding format - basically you can decide between Notebook format and raw/source format. The downloaded files can then be executed directly against the Databricks cluster if Databricks-Connect is setup correctly (Setup Databricks-Connect on AWS, Setup Databricks-Connect on Azure)

The up-/downloaded state of the single items are also reflected in their icons:

If you have set useCodeCells = true in your connection, the Code Cells will be added once you download a raw/source file. They will not be removed again when you upload the raw/source file again!

NOTE: The logic is currently not recursive - if a folder exists online and locally, does not mean that also all sub-folders and files exist online and locally!

• A small red dot at the top right of an item indicates that it only exists online in the Databricks workspace but has not yet been downloaded to the localSyncFolder into the subfolder Workspace.
• A small blue dot at the bottom right of an item indicates that it only exists locally but has not yet been uploaded to the Databricks workspace. Please note that only local files that match the file extensions defined in exportFormats will be considered for an upload. For all other files you will see a warning in VS Code.
• If there is no blue or red dot in the icon then the file/folder exists locally and also in the Databricks workspace. However, this does not mean that the files have to be in sync. It is up to you to know which file is more recent and then sync them accordingly!

This extension also allows you to manage your Databricks clusters directly from within VS Code. So you do not need to open the web UI anymore to start or stop your clusters. It also distinguishes between regular clusters and job clusters which will be displayed in a separate folder. In addition, the Cluster Manager also allows you to script the definition of your cluster and store it locally - e.g. if you want to integrate it as part of your CI/CD. This cluster definition file can for example be used with the DatabricksPS PowerShell Module to automate the cluster deployment. The cluster manager also distinguishes between regular user-created clusters and job-clusters.

Using Databricks Notebook Kernels you can execute local code againt a running Databricks cluster. Simply open a .ipynb notebook and select the Databricks kernel of your choice. A new kernel will be added automatically for each Databricks cluster that is currently running. In case you need to restart the kernel, you can do so by right-clicking the underlying cluster in the Cluster Manager and select Restart Databricks Kernel or use the Restart Kernel button from the toolbar. The execution of the first cell may take a bit longer as the remote execution context has to be set up before any commands can be executed. The notebook kernel supports the magics %sql, %python, %r, %scala, %md, %run and also %pip. Depending on the output of the cell, the results are rendered accordingly - e.g. as text, as table or as image. Due to the limitations of the underlying APIs, the output is limited to one output type so you should avoid mixing print(...), display(df) and df.plot as only one of them will be shown. For better visualization of tabluar results this extension includes a dependency to the extension vscode-data-table which offers much more features than the standard visualizations.

Notebook Kernels also support other features like Files in Repo to build libraries within your repo, _sqldf to expose results of SQL cells to Python/Pyspark, %run to run other notebooks inline with the current notebook and also dbutils.notebook.run().

Whenever a notebook is opened from either the local sync folder or via the Virtual File System using dbws:/ URI, the Databricks notebook kernels are the preferred ones and should appear at the top of the list when you select a kernel.

We distinguish between Live-execution and Offline-execution. In Live-execution mode, files are opened directly from Databricks by mounting the Databricks Workspace into your VSCode Workspace using dbws:/ URI scheme. In this mode there is no intermediate local copy but you work directly against the Databricks Workspace. Everything you run must already exist online in the Databricks Workspace.

This is slightly different in Offline-execution where all files you want to work with need to be synced locally first using the Workspace Manager. This is especially important when it comes %run which behaves slightly differntly compared to Live-execution mode. %run in Offline-execution runs the code from your local file instead of the code that exists in Dtabricks online! Other commands like dbutils.notebook.run() always use the code thats currently online so if you have changed the refernced notebook locally, you have to upload it first. This is simply because we cannot easily replicate the behavior of dbutils.notebook.run() locally!

The File System Integration allows you to mount DBFS and the Databricks workspace/notebooks directly into your VSCode Explorer. You can simply open (or preview) all files and also save them back as if the files were local. So no local copy of your notebooks is necessary anymore but you can work with the files from the Databricks workspace directly. Everything happens automatically - when you open a file, it is downloaded, when you save it, it is uploaded again. This integration also allows easy drag&drop between the different file systems, including the local file system! You want to upload a local notebook to the Databricks workspace? Simply drag&drop it! You want to download a file from DBFS? Simply drag&drop it!

There are two virtual file systems that come with this extension:

• dbws:/ to access your notebook from the DAtabricks workspace
• dbfs:/ to access files on the Databricks File System (DBFS) - similar to the DBFS Browser

The SQL Browser is similar to the Data tab in the Databricks Web UI. It shows you all object that exist in the Hive metastore. To use the SQL Browser you first need to have a running cluster as the Hive metastore can only be browsed via a cluster. By default, the first running cluster is selected for you automatically. However, you can also change the cluster you want to use in the Cluster Manager by right-clicking the cluster of your choice and clicking Use for SQL Browser. For large databases with a lot of tables, it can take quite some time to load all details so please be patient! This is because once you expand a database, the definition of all its tables are queried one by one to get and show all relevant information in the SQL browser then.

The Job Manager allows you to manage all your existing Databricks jobs from within VS Code. It gives you information about currently deployed jobs and their different job-runs/executions. You can also start and stop new job-runs which will then be executed on the cluster. Similar to the Cluster Manager, you can also script the jobs to integrate them in automated CI/CD deployment pipelines using the DatabricksPS PowerShell Module.

The DBFS Browser allows you to browse the whole Databricks File System including mountpoints! You can also download files, view or modify them locally and upload them again. For viewing files I highly recommend installing the extension Data Preview as it supports most formats commonly used with Databricks. Clicking a file will open it as a temporary file in VS Code. If you download it explicitly using the icon next to the item, it will be stored locally in the DBFS folder of your localSyncFolder.

NOTE: The DBFS Browser also supports browsing of mount points if they were configured with a static authentication (e.g. an Azure Service Principal)!

Another tool to help you working with Databricks locally is the Secrets Browser. It allows you to browse, create, update and delete your secret scopes and secrets. This can come in handy if you want to quickly add a new secret as this is otherwise only supported using the plain REST API (or a CLI)!

Using the Repo Manager you can interact with the Repos API which allows you to add new repositories to the Databricks workspace and also update/check-out specific branches and tags. The latest branch/tag can then be found via the Workspace Manager

Q: What can I do if none of the tabs/browsers is showing anything?

A: This is very likely an issue with the connection. Depending on the Connection Manager, please make sure that all relevant fields, especially apiRootUrl and personalAccessToken are set correctly. If you are sure the values are correct, please check the logs in the output window and filter for paiqo.databricks-vscode to see the log outputs for this extension.

Q: My Personal Access Token (PAT) changed, how can I update my connection?

A: Whenever the property personalAccessToken is provided, it will be used and updated in the system key chain/credential manager. Once it is savely stored there, it will be removed again from the VSCode configuration.