Detect breaking changes to column names and data types in state:modified check

.changes/unreleased/Features-20230323-133026.yaml

@@ -0,0 +1,6 @@
+kind: Features
+body: Detect breaking changes to contracts in state:modified check
+time: 2023-03-23T13:30:26.593717-04:00
+custom:
+  Author: gshank
+  Issue: "6869"

core/dbt/contracts/graph/nodes.py

@@ -2,6 +2,7 @@
 import time
 from dataclasses import dataclass, field
 from enum import Enum
+import hashlib
 
 from mashumaro.types import SerializableType
 from typing import (
@@ -39,7 +40,7 @@
 from dbt.contracts.util import Replaceable, AdditionalPropertiesMixin
 from dbt.events.proto_types import NodeInfo
 from dbt.events.functions import warn_or_error
-from dbt.exceptions import ParsingError, InvalidAccessTypeError
+from dbt.exceptions import ParsingError, InvalidAccessTypeError, ModelContractError
 from dbt.events.types import (
     SeedIncreased,
     SeedExceedsLimitSamePath,
@@ -387,6 +388,12 @@ def same_config(self, old) -> bool:
             old.unrendered_config,
         )
 
+    def build_contract_checksum(self):
+        pass
+
+    def same_contract(self, old) -> bool:
+        return True
+
     def patch(self, patch: "ParsedNodePatch"):
         """Given a ParsedNodePatch, add the new information to the node."""
         # explicitly pick out the parts to update so we don't inadvertently
@@ -428,6 +435,7 @@ def same_contents(self, old) -> bool:
             and self.same_persisted_description(old)
             and self.same_fqn(old)
             and self.same_database_representation(old)
+            and self.same_contract(old)
             and True
         )
 
@@ -457,6 +465,7 @@ class CompiledNode(ParsedNode):
     extra_ctes: List[InjectedCTE] = field(default_factory=list)
     _pre_injected_sql: Optional[str] = None
     contract: bool = False
+    contract_checksum: Optional[str] = None
 
     @property
     def empty(self):
@@ -497,6 +506,39 @@ def depends_on_nodes(self):
     def depends_on_macros(self):
         return self.depends_on.macros
 
+    def build_contract_checksum(self):
+        # We don't need to construct the checksum if the model does not
+        # have contract enabled, because it won't be used.
+        # This needs to be executed after contract config is set
+        if self.contract is True:
+            contract_state = ""
+            # We need to sort the columns so that order doesn't matter
+            # columns is a str: ColumnInfo dictionary
+            sorted_columns = sorted(self.columns.values(), key=lambda col: col.name)
+            for column in sorted_columns:
+                contract_state += f"|{column.name}"
+                contract_state += column.data_type
+                contract_state += str(column.constraints)
+            data = contract_state.encode("utf-8")
+            self.contract_checksum = hashlib.new("sha256", data).hexdigest()
+
+    def same_contract(self, old) -> bool:
+        if old.contract is False and self.contract is False:
+            # Not a change
+            return True
+        if old.contract is False and self.contract is True:
+            # A change, but not a breaking change
+            return False
+        if old.contract is True and self.contract is False:
+            # Breaking change: throw an error
+            raise (ModelContractError(node=self))
+        if self.contract_checksum == old.contract_checksum:
+            # Breaking change: throw an error
+            raise (ModelContractError(node=self))
+        else:
+            # No change
+            return False
+
 
 # ====================================
 # CompiledNode subclasses

core/dbt/exceptions.py

@@ -209,6 +209,24 @@ def _fix_dupe_msg(self, path_1: str, path_2: str, name: str, type_name: str) ->
             )
 
 
+class ModelContractError(DbtRuntimeError):
+    CODE = 10016
+    MESSAGE = "Contract Error"
+
+    def __init__(self, node=None):
+        super().__init__(self.message(), node)
+
+    @property
+    def type(self):
+        return "Contract"
+
+    def message(self):
+        return (
+            "There is a breaking change in the model contract; "
+            "you may need to create a new version. See: https://docs.getdbt.com/docs/collaborate/publish/model-versions"
+        )
+
+
 class RecursionError(DbtRuntimeError):
     pass
 

core/dbt/graph/selector_methods.py

@@ -542,6 +542,7 @@ def search(self, included_nodes: Set[UniqueId], selector: str) -> Iterator[Uniqu
             ),
             "modified.relation": self.check_modified_factory("same_database_representation"),
             "modified.macros": self.check_modified_macros,
+            "modified.contract": self.check_modified_factory("same_contract"),
         }
         if selector in state_checks:
             checker = state_checks[selector]

core/dbt/parser/base.py

@@ -314,6 +314,7 @@ def update_parsed_node_config(
         # compatibility with earlier node-only config.
         if config_dict.get("contract", False):
             parsed_node.contract = True
+            parsed_node.build_contract_checksum()
 
         # unrendered_config is used to compare the original database/schema/alias
         # values and to handle 'same_config' and 'same_contents' calls

tests/functional/defer_state/fixtures.py

@@ -50,6 +50,58 @@
       - name: name
 """
 
+contract_schema_yml = """
+version: 2
+models:
+  - name: view_model
+    columns:
+      - name: id
+        tests:
+          - unique:
+              severity: error
+          - not_null
+      - name: name
+  - name: table_model
+    config:
+      contract: true
+    columns:
+      - name: id
+        data_type: integer
+        tests:
+          - unique:
+              severity: error
+          - not_null
+      - name: name
+        data_type: text
+"""
+
+modified_contract_schema_yml = """
+version: 2
+models:
+  - name: view_model
+    columns:
+      - name: id
+        tests:
+          - unique:
+              severity: error
+          - not_null
+      - name: name
+  - name: table_model
+    config:
+      contract: true
+    columns:
+      - name: id
+        data_type: integer
+        tests:
+          - unique:
+              severity: error
+          - not_null
+      - name: name
+        data_type: text
+        constraints:
+          - type: not_null
+"""
+
 exposures_yml = """
 version: 2
 exposures:

tests/functional/defer_state/test_modified_state.py

@@ -5,9 +5,9 @@
 
 import pytest
 
-from dbt.tests.util import run_dbt, update_config_file, write_file
+from dbt.tests.util import run_dbt, update_config_file, write_file, get_manifest
 
-from dbt.exceptions import CompilationError
+from dbt.exceptions import CompilationError, ModelContractError
 
 from tests.functional.defer_state.fixtures import (
     seed_csv,
@@ -18,6 +18,8 @@
     exposures_yml,
     macros_sql,
     infinite_macros_sql,
+    contract_schema_yml,
+    modified_contract_schema_yml,
 )
 
 
@@ -261,3 +263,38 @@ def test_changed_exposure(self, project):
         results = run_dbt(["run", "--models", "+state:modified", "--state", "./state"])
         assert len(results) == 1
         assert results[0].node.name == "view_model"
+
+
+class TestChangedContract(BaseModifiedState):
+    def test_changed_contract(self, project):
+        self.run_and_save_state()
+
+        # update contract for table_model
+        write_file(contract_schema_yml, "models", "schema.yml")
+
+        # This will find the table_model node modified both through a config change
+        # and by a non-breaking change to contract: true
+        results = run_dbt(["run", "--models", "state:modified", "--state", "./state"])
+        assert len(results) == 1
+        assert results[0].node.name == "table_model"
+        manifest = get_manifest(project.project_root)
+        model_unique_id = "model.test.table_model"
+        model = manifest.nodes[model_unique_id]
+        expected_unrendered_config = {"contract": True, "materialized": "table"}
+        assert model.unrendered_config == expected_unrendered_config
+
+        # Run it again with "state:modified:contract", still finds modified due to contract: true
+        results = run_dbt(["run", "--models", "state:modified.contract", "--state", "./state"])
+        assert len(results) == 1
+        # save a new state
+        self.copy_state()
+
+        # This should raise because a column has added a constraint
+        write_file(modified_contract_schema_yml, "models", "schema.yml")
+        with pytest.raises(ModelContractError):
+            results = run_dbt(["run", "--models", "state:modified.contract", "--state", "./state"])
+
+        # Go back to schema file without contract. Should raise an error.
+        write_file(schema_yml, "models", "schema.yml")
+        with pytest.raises(ModelContractError):
+            results = run_dbt(["run", "--models", "state:modified.contract", "--state", "./state"])