Add post processor to collect results into csv

Signed-off-by: Tyler Gu <[email protected]>

acto/post_process/collect_test_result.py

@@ -1,15 +1,44 @@
 import argparse
+import glob
 import json
+import os
+import re
 
 import pandas as pd
 
 from acto.lib.operator_config import OperatorConfig
 from acto.post_process.post_process import PostProcessor
+from acto.result import DifferentialOracleResult
 
 
 class CollectTestResult(PostProcessor):
     """Post processor for diff test"""
 
+    def __init__(self, testrun_dir: str, config: OperatorConfig):
+        super().__init__(testrun_dir, config)
+
+        # Load the post diff test results
+        self.diff_test_results: dict[str, DifferentialOracleResult] = {}
+        post_diff_test_result_files = glob.glob(
+            os.path.join(
+                testrun_dir, "post_diff_test", "compare-results-*.json"
+            )
+        )
+        for result_file in post_diff_test_result_files:
+            if (
+                matches := re.search(r"compare-results-(.*).json", result_file)
+            ) is not None:
+                input_digest = matches.group(1)
+            else:
+                raise ValueError(
+                    f"Could not parse the input digest from the file name: {result_file}"
+                )
+            with open(result_file, "r", encoding="utf-8") as file:
+                results = json.load(file)
+                self.diff_test_results[
+                    input_digest
+                ] = DifferentialOracleResult.model_validate(results[0])
+
     def post_process(self, output_path: str):
         """Post process the results"""
         return self.dump_csv(output_path)
@@ -19,10 +48,15 @@ def dump_csv(self, output_path: str):
         normal_results = []
         for trial in self.trial_to_steps.values():
             for step in trial.steps.values():
+                is_alarm = (
+                    not step.run_result.is_invalid_input()
+                    and step.run_result.oracle_result.is_error()
+                )
                 normal_results.append(
                     {
                         "Trial number": str(step.run_result.step_id),
                         "Testcase": json.dumps(step.run_result.testcase),
+                        "Alarm": is_alarm,
                         "Crash": step.run_result.oracle_result.crash,
                         "Health": step.run_result.oracle_result.health,
                         "Operator log": step.run_result.oracle_result.operator_log,
@@ -34,6 +68,21 @@ def dump_csv(self, output_path: str):
                     }
                 )
 
+        for input_digest, result in self.diff_test_results.items():
+            normal_results.append(
+                {
+                    "Trial number": str(result.to_step),
+                    "Testcase": input_digest,
+                    "Alarm": is_alarm,
+                    "Crash": None,
+                    "Health": None,
+                    "Operator log": None,
+                    "Consistency": None,
+                    "Differential": str(result),
+                    "Custom": None,
+                }
+            )
+
         normal_results_df = pd.DataFrame(normal_results)
         normal_results_df = normal_results_df.sort_values(by=["Trial number"])
         normal_results_df.to_csv(output_path, index=False)

acto/result.py

@@ -1,6 +1,5 @@
 """Model for the result of a run of an Acto"""
 
-
 import enum
 import json
 from typing import Optional, Union
@@ -63,7 +62,7 @@ class DifferentialOracleResult(OracleResult):
 
     model_config = pydantic.ConfigDict(arbitrary_types_allowed=True)
 
-    diff: deepdiff.DeepDiff
+    diff: pydantic.SkipValidation[deepdiff.DeepDiff]
     from_step: StepID
     from_state: dict = pydantic.Field(
         description="The state that the diff was generated from"