CLI option to force refresh results (#616)

* CLI option to force refresh results

* address copilot comments

Author: cosenal

metriq_gym/cli.py

@@ -155,6 +155,11 @@ def build_parser() -> argparse.ArgumentParser:
         default=argparse.SUPPRESS,
         help="Export results to JSON file (optional)",
     )
+    suite_poll.add_argument(
+        "--no-cache",
+        action="store_true",
+        help="Ignore locally cached results and refetch provider job data",
+    )
 
     suite_view = suite_subparsers.add_parser("view", help="View suite jobs")
     suite_view.add_argument(
@@ -210,6 +215,11 @@ def build_parser() -> argparse.ArgumentParser:
         default=argparse.SUPPRESS,
         help="Export results to JSON file (optional)",
     )
+    job_poll.add_argument(
+        "--no-cache",
+        action="store_true",
+        help="Ignore locally cached results and refetch provider job data",
+    )
 
     job_upload.add_argument(
         "--repo",

metriq_gym/run.py

@@ -565,7 +565,8 @@ def fetch_result(
 ) -> Optional["BenchmarkResult"]:
     job_type: JobType = JobType(metriq_job.job_type)
     job_result_type = setup_benchmark_result_class(job_type)
-    if metriq_job.result_data is not None:
+    if metriq_job.result_data is not None and not getattr(args, "no_cache", False):
+        print("[Cached result data]")
         return job_result_type.model_validate(metriq_job.result_data)
 
     job_data: "BenchmarkData" = setup_job_data_class(job_type)(**metriq_job.data)

tests/unit/test_run.py

@@ -1,12 +1,18 @@
+from dataclasses import dataclass
 import logging
 import pytest
 from unittest.mock import MagicMock, patch
 
 from qbraid import QbraidError
+from qbraid.runtime import JobStatus
+from metriq_gym.benchmarks.benchmark import BenchmarkData, BenchmarkResult
 from metriq_gym.run import (
     setup_device,
     dispatch_job,
+    fetch_result,
 )
+from metriq_gym.job_manager import MetriqGymJob, JobManager
+from metriq_gym.constants import JobType
 from metriq_gym.exceptions import QBraidSetupError
 
 
@@ -113,3 +119,100 @@ def test_dispatch_missing_config_file(mock_exists, mock_args, mock_job_manager,
         # Verify output shows file not found errors
         captured = capsys.readouterr()
         assert "Configuration file not found" in captured.out
+
+
+class DummyResult(BenchmarkResult):
+    value: int
+
+
+@dataclass
+class DummyJobData(BenchmarkData):
+    provider_job_ids: list[str]
+
+
+class DummyQuantumJob:
+    def __init__(self, job_id: str, value: int):
+        self.id = job_id
+        self._value = value
+
+    def status(self):
+        return JobStatus.COMPLETED
+
+    def result(self):
+        class _R:
+            data = {"value": self._value}
+
+        return _R()
+
+
+class DummyBenchmark:
+    def poll_handler(self, job_data, result_data, quantum_jobs):
+        return DummyResult(value=result_data[0]["value"])
+
+
+def _make_cached_job(val: int) -> MetriqGymJob:
+    return MetriqGymJob(
+        id="job-1",
+        job_type=JobType.WIT,
+        params={"benchmark_name": JobType.WIT.name},
+        data={"provider_job_ids": ["provider-job-1"]},
+        provider_name="local",
+        device_name="dummy_device",
+        platform={},
+        dispatch_time=None,
+        result_data={"value": val},
+    )
+
+
+def test_fetch_result_uses_cache_when_no_flag(monkeypatch):
+    EXPECTED_CACHED_VALUE = 7
+    job = _make_cached_job(EXPECTED_CACHED_VALUE)
+    jm = JobManager()
+    jm.jobs.append(job)
+    args = MagicMock()
+    args.no_cache = False
+
+    import metriq_gym.run as run_mod
+
+    monkeypatch.setattr(run_mod, "setup_benchmark_result_class", lambda *_: DummyResult)
+    monkeypatch.setattr(run_mod, "setup_job_data_class", lambda *_: DummyJobData)
+    monkeypatch.setattr(run_mod, "setup_benchmark", lambda *_, **__: DummyBenchmark())
+    monkeypatch.setattr(
+        run_mod,
+        "load_job",
+        lambda *_, **__: DummyQuantumJob("provider-job-1", EXPECTED_CACHED_VALUE),
+    )
+    monkeypatch.setattr(run_mod, "validate_and_create_model", lambda params: params)
+
+    result = fetch_result(job, args, jm)
+    assert result.value == EXPECTED_CACHED_VALUE
+
+
+def test_fetch_result_bypasses_cache_with_flag(monkeypatch):
+    EXPECTED_FRESH_VALUE = 42
+    CACHED_VALUE = 7
+    job = _make_cached_job(CACHED_VALUE)
+    jm = JobManager()
+    jm.jobs.append(job)
+    args = MagicMock()
+    args.no_cache = True
+
+    import metriq_gym.run as run_mod
+
+    monkeypatch.setattr(run_mod, "setup_benchmark_result_class", lambda *_: DummyResult)
+    monkeypatch.setattr(run_mod, "setup_job_data_class", lambda *_: DummyJobData)
+    monkeypatch.setattr(run_mod, "setup_benchmark", lambda *_, **__: DummyBenchmark())
+    monkeypatch.setattr(
+        run_mod,
+        "load_job",
+        lambda *_, **__: DummyQuantumJob("provider-job-1", EXPECTED_FRESH_VALUE),
+    )
+    monkeypatch.setattr(run_mod, "validate_and_create_model", lambda params: params)
+
+    result = fetch_result(job, args, jm)
+    assert result.value == EXPECTED_FRESH_VALUE, (
+        "Should fetch fresh value when --no-cache specified"
+    )
+    assert job.result_data == {"value": EXPECTED_FRESH_VALUE}, (
+        "Cached result_data should be updated"
+    )