Short description and motivation for the proposed feature
Allow users to add custom COMMENT descriptions to views (and optionally columns) during CREATE VIEW. These comments should be stored as metadata and surfaced through the Data Catalog, enabling Agents to better understand the semantic meaning of tables and columns when querying data.
Video or screenshots
Currently, the Data Catalog auto-generates metadata descriptions for tables and columns. However, auto-generated descriptions may not capture domain-specific context, business logic, or disambiguate similarly named fields. This can lead to:
-Ambiguity: Agents may misinterpret column meanings (e.g., status could refer to order status, material status, or approval status).
-Lack of business context: Auto-generated descriptions miss nuances like units, valid value ranges, or business rules.
-Reduced accuracy: Without clear semantic understanding, Agent-generated SQL queries may target wrong columns or produce incorrect results.
Describe some possible solutions
1、COMMENT support in CREATE VIEW syntax:
CREATE VIEW my_view
COMMENT 'This view contains aircraft material status tracking data, including delivery and installation timelines'
AS (
SELECT * FROM datasource.my_table
);
2、Column-level comments:
CREATE VIEW my_view (
material_code COMMENT 'Unique identifier for the material part (MPart)',
status COMMENT 'Material status: 已发料 = Delivered, 已安装 = Installed, 已冻结 = Frozen',
delivery_date COMMENT 'Date when material was delivered to the assembly line'
)
AS (
SELECT * FROM datasource.my_table
);
3、Integration with Data Catalog:
User-defined comments should be stored in INFORMATION_SCHEMA.META_TABLES (TABLE_DESCRIPTION) and INFORMATION_SCHEMA.META_COLUMNS (COLUMN_DESCRIPTION).
User-defined comments should take priority over auto-generated descriptions.
Anything else?
No response
Short description and motivation for the proposed feature
Allow users to add custom COMMENT descriptions to views (and optionally columns) during CREATE VIEW. These comments should be stored as metadata and surfaced through the Data Catalog, enabling Agents to better understand the semantic meaning of tables and columns when querying data.
Video or screenshots
Currently, the Data Catalog auto-generates metadata descriptions for tables and columns. However, auto-generated descriptions may not capture domain-specific context, business logic, or disambiguate similarly named fields. This can lead to:
-Ambiguity: Agents may misinterpret column meanings (e.g., status could refer to order status, material status, or approval status).
-Lack of business context: Auto-generated descriptions miss nuances like units, valid value ranges, or business rules.
-Reduced accuracy: Without clear semantic understanding, Agent-generated SQL queries may target wrong columns or produce incorrect results.
Describe some possible solutions
1、COMMENT support in CREATE VIEW syntax:
CREATE VIEW my_view
COMMENT 'This view contains aircraft material status tracking data, including delivery and installation timelines'
AS (
SELECT * FROM datasource.my_table
);
2、Column-level comments:
CREATE VIEW my_view (
material_codeCOMMENT 'Unique identifier for the material part (MPart)',statusCOMMENT 'Material status: 已发料 = Delivered, 已安装 = Installed, 已冻结 = Frozen',delivery_dateCOMMENT 'Date when material was delivered to the assembly line')
AS (
SELECT * FROM datasource.my_table
);
3、Integration with Data Catalog:
User-defined comments should be stored in INFORMATION_SCHEMA.META_TABLES (TABLE_DESCRIPTION) and INFORMATION_SCHEMA.META_COLUMNS (COLUMN_DESCRIPTION).
User-defined comments should take priority over auto-generated descriptions.
Anything else?
No response