Catalog statistics¶

When you run dbt docs generate, this adapter enriches the catalog with approximate row counts for every table in your Fabric Data Warehouse. These row counts appear in the dbt docs site alongside your column information, giving you a quick sense of table sizes without running any additional queries.

How it works¶

The catalog query uses the T-SQL function OBJECTPROPERTYEX with the 'Cardinality' property to retrieve the optimizer's row count estimate for each table:

cast(objectpropertyex(object_id, 'Cardinality') as int)

This is the same estimate that the query optimizer uses internally, derived from table statistics. It is not an exact count — it is an approximation that avoids the cost of scanning the entire table.

Row counts are included for base tables only. Views do not have cardinality statistics, so the row count stat is excluded from views in the catalog output.

What you see in dbt docs¶

After running dbt docs generate and opening the docs site, each table node shows a Row Count statistic with the approximate number of rows. Views show no row count.

No configuration is needed — the statistics are included automatically whenever you generate the catalog.

Why this adapter?¶

Microsoft's upstream dbt-fabric adapter does not include any statistics in the catalog output. The base catalog query returns null for all stat columns, so the dbt docs site shows no table size information.

This adapter adds row count statistics out of the box, which is especially useful for:

Quickly assessing table sizes during development and code review
Spotting unexpected growth or emptiness in tables after a dbt run
Onboarding new team members who want to understand the data landscape

Limitations¶

Row counts are approximate. They come from the query optimizer's statistics, which may lag behind the actual row count if statistics have not been updated recently.
Row counts are only available for Fabric Data Warehouse (the fabric adapter type). The FabricSpark adapter does not include this feature.