kernelci.api.models: create_regression method for Regression model

Add a Regression object builder that will take care of formatting and
returning Regression object for subsequent use, for instance, for
submitting it to the API.

This is meant to be used by client code when detecting and submitting
regressions given two test or build nodes that mark the breaking point.

Signed-off-by: Ricardo Cañuelo <[email protected]>

kernelci/api/models.py

@@ -15,6 +15,8 @@
 from datetime import datetime, timedelta
 from typing import Any, Optional, Dict, List, ClassVar
 import enum
+from operator import attrgetter
+import json
 from bson import ObjectId
 from pydantic import (
     AnyHttpUrl,
@@ -477,6 +479,76 @@ class Regression(Node):
         'data.pass_node.holdoff',
     ]
 
+    @classmethod
+    def create_regression(cls, fail_node, pass_node, as_dict=False):
+        """Builds and returns a Regression object from two source nodes:
+        a failing and a passing node. These two nodes are assumed to be
+        from two different runs of the same test/build with the same
+        configuration. pass_node is the last instance of the test/build
+        that passed and fail_node is the first one that failed.
+
+        The returned Regression will be an in-memory python object not
+        backed up in the DB, which can then be handled and used in a
+        node submit request.
+
+        If 'as_dict' is True, the Regression is returned as a dict
+        rather than as a Regression object.
+
+        May raise a RuntimeError if any of the sanity checks fail.
+        """
+        # Sanity checks:
+        # - fail_node and pass_node must refer to two
+        #   different runs of the same test/build,
+        # - fail node must have run after pass_node
+        # - pass and fail nodes must have correct results (pass and
+        #   fail, respectively)
+        error_msg = ("Error creating regression for nodes "
+                     f"{pass_node.id} (last passed) and "
+                     f"{fail_node.id} (first failed). ")
+        cmp_fields = ['name',
+                      'group',
+                      'path',
+                      'data.kernel_revision',
+                      'data.arch',
+                      'data.defconfig',
+                      'data.compiler',
+                      'data.platform',]
+        for field in cmp_fields:
+            getter = attrgetter(field)
+            if getter(fail_node) != getter(pass_node):
+                raise RuntimeError(error_msg +
+                                   f"{field} is different in the fail node "
+                                   f"({getter(fail_node)} than in the pass "
+                                   f"node ({getter(pass_node)})")
+        if pass_node.created > fail_node.created:
+            raise RuntimeError(error_msg + "The fail node was created before "
+                               "the pass node")
+        if pass_node.result != 'pass':
+            raise RuntimeError(error_msg + "The pass node has a wrong result: "
+                               f"{pass_node.result}")
+        if fail_node.result != 'fail':
+            raise RuntimeError(error_msg + "The fail node has a wrong result: "
+                               f"{fail_node.result}")
+        # End of sanity checks
+        data_field = {
+            'arch': fail_node.data.arch,
+            'defconfig': fail_node.data.defconfig,
+            'compiler': fail_node.data.compiler,
+            'platform': fail_node.data.platform,
+            'failed_kernel_revision': fail_node.data.kernel_revision,
+            'fail_node': fail_node.id,
+            'pass_node': pass_node.id,
+        }
+        regression_obj = cls(name=fail_node.name,
+                             path=fail_node.path,
+                             group=fail_node.group,
+                             data=data_field,
+                             state=StateValues.DONE.value)
+        if as_dict:
+            regression_json = regression_obj.json(exclude_none=True)
+            return json.loads(regression_json)
+        return regression_obj
+
 
 class PublishEvent(BaseModel):
     """API model for the data of a <publish> event"""