sqlmesh 0.220.1.dev2__tar.gz → 0.220.2.dev2__tar.gz

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sqlmesh
-Version: 0.220.1.dev2
+Version: 0.220.2.dev2
 Summary: Next-generation data transformation framework
 Author-email: "TobikoData Inc." <engineering@tobikodata.com>
 License:                                  Apache License
@@ -344,7 +344,6 @@ Requires-Dist: pyarrow; extra == "web"
 Provides-Extra: lsp
 Requires-Dist: fastapi==0.115.5; extra == "lsp"
 Requires-Dist: watchfiles>=0.19.0; extra == "lsp"
-Requires-Dist: uvicorn[standard]==0.22.0; extra == "lsp"
 Requires-Dist: sse-starlette>=0.2.2; extra == "lsp"
 Requires-Dist: pyarrow; extra == "lsp"
 Requires-Dist: pygls<2.0.0,>=1.2.0; extra == "lsp"

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/pyproject.toml

@@ -135,7 +135,7 @@ lsp = [
     # Duplicate of web
     "fastapi==0.115.5",
     "watchfiles>=0.19.0",
-    "uvicorn[standard]==0.22.0",
+    # "uvicorn[standard]==0.22.0",
     "sse-starlette>=0.2.2",
     "pyarrow",
     # For lsp

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/sqlmesh/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.220.1.dev2'
-__version_tuple__ = version_tuple = (0, 220, 1, 'dev2')
+__version__ = version = '0.220.2.dev2'
+__version_tuple__ = version_tuple = (0, 220, 2, 'dev2')
 
-__commit_id__ = commit_id = 'gdf7d61be3'
+__commit_id__ = commit_id = 'g862ef50b0'

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/sqlmesh/core/plan/common.py

@@ -16,8 +16,9 @@ def should_force_rebuild(old: Snapshot, new: Snapshot) -> bool:
     if new.is_view and new.is_indirect_non_breaking and not new.is_forward_only:
         # View models always need to be rebuilt to reflect updated upstream dependencies
         return True
-    if new.is_seed:
+    if new.is_seed and not new.is_metadata:
         # Seed models always need to be rebuilt to reflect changes in the seed file
+        # Unless only their metadata has been updated (eg description added) and the seed file has not been touched
         return True
     return is_breaking_kind_change(old, new)
 

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/sqlmesh.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sqlmesh
-Version: 0.220.1.dev2
+Version: 0.220.2.dev2
 Summary: Next-generation data transformation framework
 Author-email: "TobikoData Inc." <engineering@tobikodata.com>
 License:                                  Apache License
@@ -344,7 +344,6 @@ Requires-Dist: pyarrow; extra == "web"
 Provides-Extra: lsp
 Requires-Dist: fastapi==0.115.5; extra == "lsp"
 Requires-Dist: watchfiles>=0.19.0; extra == "lsp"
-Requires-Dist: uvicorn[standard]==0.22.0; extra == "lsp"
 Requires-Dist: sse-starlette>=0.2.2; extra == "lsp"
 Requires-Dist: pyarrow; extra == "lsp"
 Requires-Dist: pygls<2.0.0,>=1.2.0; extra == "lsp"

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/sqlmesh.egg-info/requires.txt

@@ -113,7 +113,6 @@ PyGithub>=2.6.0
 [lsp]
 fastapi==0.115.5
 watchfiles>=0.19.0
-uvicorn[standard]==0.22.0
 sse-starlette>=0.2.2
 pyarrow
 pygls<2.0.0,>=1.2.0

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/tests/core/test_integration.py

@@ -10662,3 +10662,121 @@ def entrypoint(evaluator: MacroEvaluator) -> str:
     assert len(model_a.intervals)
 
     set_console(orig_console)
+
+
+def test_seed_model_metadata_update_does_not_trigger_backfill(tmp_path: Path):
+    """
+    Scenario:
+        - Create a seed model; perform initial population
+        - Modify the model with a metadata-only change and trigger a plan
+
+    Outcome:
+        - The seed model is modified (metadata-only) but this should NOT trigger backfill
+        - There should be no missing_intervals on the plan to backfill
+    """
+
+    models_path = tmp_path / "models"
+    seeds_path = tmp_path / "seeds"
+    models_path.mkdir()
+    seeds_path.mkdir()
+
+    seed_model_path = models_path / "seed.sql"
+    seed_path = seeds_path / "seed_data.csv"
+
+    seed_path.write_text("\n".join(["id,name", "1,test"]))
+
+    seed_model_path.write_text("""
+    MODEL (
+        name test.source_data,
+        kind SEED (
+            path '../seeds/seed_data.csv'
+        )
+    );    
+    """)
+
+    config = Config(
+        gateways={"": GatewayConfig(connection=DuckDBConnectionConfig())},
+        model_defaults=ModelDefaultsConfig(dialect="duckdb", start="2024-01-01"),
+    )
+    ctx = Context(paths=tmp_path, config=config)
+
+    plan = ctx.plan(auto_apply=True)
+
+    original_seed_snapshot = ctx.snapshots['"memory"."test"."source_data"']
+    assert plan.directly_modified == {original_seed_snapshot.snapshot_id}
+    assert plan.metadata_updated == set()
+    assert plan.missing_intervals
+
+    # prove data loaded
+    assert ctx.engine_adapter.fetchall("select id, name from memory.test.source_data") == [
+        (1, "test")
+    ]
+
+    # prove no diff
+    ctx.load()
+    plan = ctx.plan(auto_apply=True)
+    assert not plan.has_changes
+    assert not plan.missing_intervals
+
+    # make metadata-only change
+    seed_model_path.write_text("""
+    MODEL (
+        name test.source_data,
+        kind SEED (
+            path '../seeds/seed_data.csv'
+        ),
+        description 'updated by test'
+    );    
+    """)
+
+    ctx.load()
+    plan = ctx.plan(auto_apply=True)
+    assert plan.has_changes
+
+    new_seed_snapshot = ctx.snapshots['"memory"."test"."source_data"']
+    assert (
+        new_seed_snapshot.version == original_seed_snapshot.version
+    )  # should be using the same physical table
+    assert (
+        new_seed_snapshot.snapshot_id != original_seed_snapshot.snapshot_id
+    )  # but still be different due to the metadata change
+    assert plan.directly_modified == set()
+    assert plan.metadata_updated == {new_seed_snapshot.snapshot_id}
+
+    # there should be no missing intervals to backfill since all we did is update a description
+    assert not plan.missing_intervals
+
+    # there should still be no diff or missing intervals in 3 days time
+    assert new_seed_snapshot.model.interval_unit.is_day
+    with time_machine.travel(timedelta(days=3)):
+        ctx.clear_caches()
+        ctx.load()
+        plan = ctx.plan(auto_apply=True)
+        assert not plan.has_changes
+        assert not plan.missing_intervals
+
+    # change seed data
+    seed_path.write_text("\n".join(["id,name", "1,test", "2,updated"]))
+
+    # new plan - NOW we should backfill because data changed
+    ctx.load()
+    plan = ctx.plan(auto_apply=True)
+    assert plan.has_changes
+
+    updated_seed_snapshot = ctx.snapshots['"memory"."test"."source_data"']
+
+    assert (
+        updated_seed_snapshot.snapshot_id
+        != new_seed_snapshot.snapshot_id
+        != original_seed_snapshot.snapshot_id
+    )
+    assert not updated_seed_snapshot.forward_only
+    assert plan.directly_modified == {updated_seed_snapshot.snapshot_id}
+    assert plan.metadata_updated == set()
+    assert plan.missing_intervals
+
+    # prove backfilled data loaded
+    assert ctx.engine_adapter.fetchall("select id, name from memory.test.source_data") == [
+        (1, "test"),
+        (2, "updated"),
+    ]

{sqlmesh-0.220.1.dev2 → sqlmesh-0.220.2.dev2}/tests/core/test_plan.py

@@ -1189,6 +1189,65 @@ def test_forward_only_plan_seed_models(make_snapshot, mocker: MockerFixture):
     assert not snapshot_a_updated.is_forward_only
 
 
+def test_seed_model_metadata_change_no_missing_intervals(
+    make_snapshot: t.Callable[..., Snapshot],
+):
+    snapshot_a = make_snapshot(
+        SeedModel(
+            name="a",
+            kind=SeedKind(path="./path/to/seed"),
+            seed=Seed(content="content"),
+            column_hashes={"col": "hash1"},
+            depends_on=set(),
+        )
+    )
+    snapshot_a.categorize_as(SnapshotChangeCategory.BREAKING)
+    snapshot_a.add_interval("2022-01-01", now())
+
+    snapshot_a_metadata_updated = make_snapshot(
+        SeedModel(
+            name="a",
+            kind=SeedKind(path="./path/to/seed"),
+            seed=Seed(content="content"),
+            column_hashes={"col": "hash1"},
+            depends_on=set(),
+            description="foo",
+        )
+    )
+    assert snapshot_a_metadata_updated.version is None
+    assert snapshot_a_metadata_updated.change_category is None
+
+    context_diff = ContextDiff(
+        environment="prod",
+        is_new_environment=True,
+        is_unfinalized_environment=False,
+        normalize_environment_name=True,
+        create_from="prod",
+        create_from_env_exists=True,
+        added=set(),
+        removed_snapshots={},
+        modified_snapshots={
+            snapshot_a_metadata_updated.name: (snapshot_a_metadata_updated, snapshot_a)
+        },
+        snapshots={snapshot_a_metadata_updated.snapshot_id: snapshot_a_metadata_updated},
+        new_snapshots={snapshot_a_metadata_updated.snapshot_id: snapshot_a},
+        previous_plan_id=None,
+        previously_promoted_snapshot_ids=set(),
+        previous_finalized_snapshots=None,
+        previous_gateway_managed_virtual_layer=False,
+        gateway_managed_virtual_layer=False,
+        environment_statements=[],
+    )
+
+    plan = PlanBuilder(context_diff).build()
+    assert snapshot_a_metadata_updated.change_category == SnapshotChangeCategory.METADATA
+    assert not snapshot_a_metadata_updated.is_forward_only
+    assert not plan.missing_intervals  # plan should have no missing intervals
+    assert (
+        snapshot_a_metadata_updated.intervals == snapshot_a.intervals
+    )  # intervals should have been copied
+
+
 def test_start_inference(make_snapshot, mocker: MockerFixture):
     snapshot_a = make_snapshot(
         SqlModel(name="a", query=parse_one("select 1, ds"), start="2022-01-01")