persist_docs.columns triggers a full-plan probe query on views even when the model has no column docs

[PhilippLange]

Describe the bug

With persist_docs: {columns: true}, databricks__create_view_as always calls get_columns_in_query(sql) before creating a view, which executes select * from (<full model sql>) as __dbt_sbq where false limit 0 against the warehouse. It does this even when the model has no columns: entries in any schema YAML, in which case the resulting column list contains no comment clauses and is semantically identical to emitting no column list at all.

On a wide view the probe hangs for a long time. where false is only folded after full analysis, so Spark still plans the whole query and does Delta metadata/file listing on every source.

Expected behavior

No probe query when the model has no column descriptions. dbt's own macros/adapters/persist_docs.sql already uses this guard:

{% if for_columns and config.persist_column_docs() and model.columns %}

The Databricks view path omits the and model.columns short-circuit.

Suggested fix

In dbt/include/databricks/macros/relations/view/create.sql:

{%- if config.persist_column_docs() and model.columns -%}

Steps to reproduce

1. Set persist_docs: {relation: true, columns: true} in dbt_project.yml
2. Create a materialized='view' model with no corresponding schema.yml entry
3. Run dbt run -s that_model and observe the where false limit 0 query in the Databricks query history

Notes

• Workaround: persist_docs={'relation': True, 'columns': False} per model.

System information

• dbt-databricks: 1.11.6
• dbt-core: 1.11.6
• databricks-sql-connector: 4.1.3

[PhilippLange]

Sketch of a deeper fix, as an alternative to the guard.

The guard only covers models with no documented columns. Views with column docs still pay for the check, since the column list has to be known before the DDL is emitted.

databricks__alter_view already reapplies comments after the fact, taking column names from the catalog rather than the query:

{%- set existing_columns = adapter.get_columns_in_relation(target_relation) -%}
{%- set columns_to_persist = adapter.get_persist_doc_columns(existing_columns, model.columns) -%}
{{ alter_column_comment(target_relation, columns_to_persist) }}

Creation could do the same: drop the inline column list from databricks__create_view_as, and call that block after the view exists.