v1.12.1

Patch release with several FabricSpark and Fabric DW improvements, mostly package-compatibility and reliability fixes. No breaking changes.

User-facing changes

New features

• FabricSpark: clustered_by → Delta liquid clustering (#281) — the existing clustered_by model config is now wired to Delta liquid clustering on Lakehouse tables.
• Fabric DW: compare_and_classify_query_results support (#289) — enables the corresponding dbt-audit-helper macro on Fabric DW.
• Fabric DW: configurable lock timeout (#274, #278) — schema-lock waits are now capped via SET LOCK_TIMEOUT, with a configurable value (default 30s) to fail fast instead of hanging on contended DDL.

Fixes

• Fabric DW: release Sch-S locks after synapsesql writes (#272) — Python model writes now await JVM GC so the underlying JDBC sessions release their Sch-S locks on the target table, eliminating multi-minute LCK_M_SCH_M waits on follow-up DDL.
• dbt-date: get_fiscal_periods now dispatched correctly (#277) — closes #243.

Cleanups (user-visible)

• Remove stale dbt-utils overrides for test_not_null_where / test_unique_where — these would have raised NoSuchMacroError against current dbt-utils since the upstream defaults were removed (#286).
• Remove redundant relationships_where override — it now matches the upstream default verbatim (#288).

Dependencies

• Lower requests floor to >=2.32.5 for broader resolver compatibility (#293).
• Merged upstream microsoft/dbt-fabric v1.9.10 + v1.10.0 (review notes in commit; nothing needed porting).

Documentation

• New per-package pages: dbt_artifacts (#270), dbt-project-evaluator (#269).

Internal / dev improvements

Not user-facing — listed for completeness:

• CI: always run python_model tests (#279)
• FabricSpark integration test speedup via HC session pooling (#268)
• Test fixture: close HC Livy sessions before drop_test_schema (#276)
• Comment cleanup in not_empty_string override (#287)

v1.12.0

Highlights

• dbt Core 1.12 support (compatibility)
• FabricSpark (Lakehouse) adapter matured significantly — now a first-class adapter alongside Fabric DW (lakehouse guide)
• Microsoft Purview integration for syncing dbt metadata to your data catalog (docs)
• dbt-external-tables support via OPENROWSET (docs)
• Documentation site rebuilt on Zensical with extensive new content (dbt-fabric.debruyn.dev)

FabricSpark (Lakehouse) adapter

See the Lakehouse guide and FabricSpark configuration.

• New materializations: incremental, snapshot, clone, materialized_view (lake views), and proper view support
• Default materialization changed to view (matches dbt-spark conventions)
• Cross-workspace 4-part naming
• persist_docs support
• High-concurrency Livy sessions for parallel statement execution
• Configurable Fabric API base URLs for non-production tenants (gov clouds, sovereign clouds)
• Fixed source freshness failures on Lakehouse (ported from upstream)
• Fixed snapshots failing with character varying type error
• Fixed TestMetadataWithEmptyFlag parse-time crash
• Fixed escape_single_quotes behavior
• Fixed schema filter push-down in list_relations_without_caching (performance)
• Handle [SCHEMA_NOT_FOUND] gracefully in relation listing
• Force JVM GC after synapsesql writes to release JDBC schema locks

Fabric Data Warehouse

• Snapshots now use a single MERGE statement instead of UPDATE + INSERT (atomic, simpler) — see warehouse snapshots
• apply_label() decoupled from statement termination
• CLUSTER BY support on tables — see cluster by
• New statistics config option — see statistics
• Approximate row count statistics in catalog generation — see catalog stats
• Fixed T-SQL identifier quoting to use brackets ([ ])
• Fixed rows_affected reporting -1 for table materializations
• Automatic retry for transient snapshot isolation errors on metadata queries
• Singleton pollution fix in TestWarehouseSnapshots

Authentication

See the authentication guide for the full walkthrough.

• New workload_identity method for federated credentials (CI/CD-friendly, OIDC)
• New token_credential method (ported from upstream)
• Authentication doc significantly expanded with end-to-end setup walkthroughs

Community package compatibility

Integration tests and macro overrides added for both Fabric and FabricSpark across these packages — see the packages index for required dispatch config and per-package notes:

• dbt-utils
• dbt-date
• dbt-expectations
• dbt-audit-helper
• dbt-codegen
• dbt-profiler
• dbt-external-tables
• dbt-project-evaluator — Lakehouse only; Fabric DW support pending upstream dbt-labs/dbt-project-evaluator#576
• dbt_artifacts — Lakehouse only; Fabric DW (and Synapse) support pending upstream brooklyn-data/dbt_artifacts#529

Documentation

Docs site migrated from mkdocs-material to Zensical: dbt-fabric.debruyn.dev. New and significantly expanded pages:

• Comparison: dbt-fabric vs upstream microsoft/dbt-fabric
• Comparison: dbt-fabricspark vs dbt-synapse-spark
• Lakehouse — full FabricSpark setup, configuration, and feature guide
• External tables — OPENROWSET-based external table workflow
• Purview integration — end-to-end metadata sync setup
• Cluster by
• Statistics
• Catalog stats
• Limitations — explicit unsupported features per adapter
• Compatibility matrix — dbt Core / Python / adapter version table
• Packages index + 7 per-package pages
• Configuration — substantially rewritten with database/lakehouse tabs
• Python models — updated for current Livy behavior

Dependencies

• Bumped dbt-core to 1.11.10
• Bumped dbt-adapters to 1.23.0
• Bumped mssql-python to 1.6.0
• Security fixes via dependency bumps

Tooling and CI

• Devcontainer for cloud development environments (Codespaces, etc.)
• DW and DE integration tests split into separate workflows
• On-demand DE tests via /test-de PR comments
• On-demand DW tests via /test-dw PR comments
• Extensive unit test coverage added for adapter internals (Credentials, Adapter, Column, Relation, ConnectionManager, ApiClient, LivySession, FabricSparkCursor)
• Improved FabricSparkCursor PEP 249 compliance

v1.11.2

• Make sure Python models reuse existing Spark sessions and add concurrency tests

v1.11.1

Add support for scalar functions (new feature in dbt Core 1.11)

v1.11.0

• Add support for dbt Core 1.11 (Functions not supported yet)
• Fix clone support: now deletes existing clones when trying to create a new one
• Incremental models now throw an error when trying to use a strategy that doesn't exist
• Added more integration tests

v1.10.4

• Add support for array_concat macro
• Add support for array_append macro
• Rework incremental materialization logic with more resilient backup strategy and avoid creating views
• Add support for incremental Python models

v1.10.3

What's Changed

• Removed unnamed column warning for date spines
• Fixed validate_sql dbt_utils macro

Full Changelog: v1.10.2...v1.10.3

v1.10.2

What's New

Warehouse Snapshots

Create point-in-time snapshots of your Fabric Data Warehouse directly from dbt using the new create_or_update_fabric_warehouse_snapshot macro. Unlike Microsoft's upstream implementation — which hijacks Python runtime components and bypasses dbt's lifecycle — this adapter gives you a macro you call explicitly from on-run-start, on-run-end, post-hook, or any other Jinja context.

on-run-end:
  - "{{ create_or_update_fabric_warehouse_snapshot('daily_snapshot', 'Snapshot after dbt run') }}"

You get full control: multiple snapshots per run, dynamic names, custom timing — all using standard dbt configuration.

Python Models

Run PySpark transformations against your Fabric Data Warehouse using Python models. Add your Workspace and Lakehouse to profiles.yml and write a model(dbt, spark) function that returns a PySpark DataFrame. The adapter handles Livy session management, code submission, and reading/writing via Fabric's native synapsesql connector.

# profiles.yml
dev:
  type: fabric
  workspace: My Workspace
  database: my_data_warehouse
  schema: dbt
  lakehouse: My Lakehouse
  authentication: CLI

This feature is exclusive to dbt-fabric-samdebruyn.

Documentation

New feature guides have been added to the documentation site:

• Authentication — Covers Azure CLI, DefaultAzureCredential, service principal, environment variables, and Fabric Notebook authentication with ready-to-use profiles.yml examples
• Python Models — How to write, configure, and troubleshoot Python models on Fabric
• Warehouse Snapshots — Using the snapshot macro with hooks, dynamic names, and a comparison with Microsoft's approach