DiscoverX

Multi-table operations over the lakehouse.

Run a single command to execute operations across many tables.

Operations examples

Operations are applied concurrently across multiple tables

• Maintenance
  • VACUUM all tables (example notebook)
  • OPTIMIZE with z-order on tables having specified columns
  • Detect tables having too many small files (example notebook)
  • Visualise quantity of data written per table per period
• Governance
  • PII detection with Presidio (example notebook)
  • Text Analysis with MosaicML and Databricks MLflow (example notebook)
  • Text Analysis with OpenAI GPT (example notebook)
  • GDPR right of access: extract user data from all tables at once
  • GDPR right of erasure: delete user data from all tables at once
  • Search in any column
  • Update Owner of Data Objects (example notebook)
• Semantic classification
  • Semantic classification of columns by semantic class: email, phone number, IP address, etc.
  • Select data based on semantic classes
  • Delete data based on semantic classes
• Custom
  • Arbitrary SQL template execution across multiple tables
  • Create Mlflow gateway routes for MosaicML and OpenAI (example notebook)
  • Scan using User Specified Data Source Formats (example notebook)

Getting started

Install DiscoverX, in Databricks notebook type

%pip install dbl-discoverx

Get started

from discoverx import DX
dx = DX(locale="US")

You can now run operations across multiple tables.

As an illustration, consider the scenario where you need to retrieve a single row from various tables within a catalog that begins with "dev_" and includes the term "sample" in their names. To achieve this, the following code block utilizes the dx.from_tables function, which applies an SQL query to extract JSON-formatted data:

dx.from_tables("dev_*.*.*sample*")\
  .with_sql("SELECT to_json(struct(*)) AS row FROM {full_table_name} LIMIT 1")\
  .apply()

Available functionality

The available dx functions are

• from_tables("<catalog>.<schema>.<table>") selects tables based on the specified pattern (use * as a wildcard). Returns a DataExplorer object with methods
  • having_columns restricts the selection to tables that have the specified columns
  • with_concurrency defines how many queries are executed concurrently (10 by defailt)
  • with_sql applies a SQL template to all tables. After this command you can apply an action. See in-depth documentation here.
  • unpivot_string_columns returns a melted (unpivoted) dataframe with all string columns from the selected tables. After this command you can apply an action
  • scan (experimental) scans the lakehouse with regex expressions defined by the rules and to power the semantic classification.
• intro gives an introduction to the library
• scan scans the lakehouse with regex expressions defined by the rules and to power the semantic classification. Documentation
• display_rules shows the rules available for semantic classification
• search searches the lakehouse content for by leveraging the semantic classes identified with scan (eg. email, ip address, etc.). Documentation
• select_by_class selects data from the lakehouse content by semantic class. Documentation
• delete_by_class deletes from the lakehouse by semantic class. Documentation

from_tables Actions

After a with_sql or unpivot_string_columns command, you can apply the following actions:

• explain explains the queries that would be executed
• display executes the queries and shows the first 1000 rows of the result in a unioned dataframe
• apply returns a unioned dataframe with the result from the queries

Requirements

• A Databricks workspace
• Unity Catalog

Project Support

Please note that all projects in the /databrickslabs github account are provided for your exploration only, and are not formally supported by Databricks with Service Level Agreements (SLAs). They are provided AS-IS and we do not make any guarantees of any kind. Please do not submit a support ticket relating to any issues arising from the use of these projects.

Any issues discovered through the use of this project should be filed as GitHub Issues on the Repo. They will be reviewed as time permits, but there are no formal SLAs for support.