Issue #18 fix

sherpa/core.py

@@ -385,7 +385,6 @@ def keras_callback(self, trial, objective_name, context_names=[]):
         return keras.callbacks.LambdaCallback(on_epoch_end=send_call)
 
 
-
 class _Runner(object):
     """
     Encapsulates all functionality needed to run a Study in parallel.
@@ -405,8 +404,8 @@ class _Runner(object):
         scheduler (sherpa.schedulers.Scheduler): a scheduler object.
         database (sherpa.database._Database): the database.
         max_concurrent (int): how many trials to run in parallel.
-        command (str): the command that runs a trial script e.g.
-            "python train_nn.py".
+        command (list[str]): components of the command that runs a trial script
+            e.g. ["python", "train_nn.py"].
         resubmit_failed_trials (bool): whether a failed trial should be
             resubmitted.
         
@@ -617,7 +616,7 @@ def optimize(parameters, algorithm, lower_is_better,
                          scheduler=scheduler,
                          database=db,
                          max_concurrent=max_concurrent,
-                         command=' '.join(['python', filename]),
+                         command=['python', filename],
                          resubmit_failed_trials=resubmit_failed_trials)
         runner.run_loop()
     return study.get_best_result()

sherpa/schedulers.py

@@ -54,8 +54,8 @@ def submit_job(self, command, env={}, job_name=''):
         Submits a job to the scheduler.
 
         Args:
-            command (str): the command to run by the scheduler e.g.
-                ``python train.py``
+            command (list[str]): components to the command to run by the
+                scheduler e.g. ``["python", "train.py"]``
             env (dict): environment variables to pass to the job.
             job_name (str): this specifies a name for the job and its output
                 directory.
@@ -105,7 +105,7 @@ def __init__(self, submit_options='', output_dir=''):
     def submit_job(self, command, env={}, job_name=''):
         env.update(os.environ.copy())
         optns = self.submit_options.split(' ') if self.submit_options else []
-        process = subprocess.Popen(optns + command.split(' '), env=env)
+        process = subprocess.Popen(optns + command, env=env)
         self.jobs[process.pid] = process
         return process.pid
 
@@ -181,7 +181,7 @@ def submit_job(self, command, env={}, job_name=''):
         job_script += 'echo "Running from" ${HOSTNAME}\n'
         for var_name, var_value in env.items():
             job_script += 'export {}={}\n'.format(var_name, var_value)
-        job_script += command  # 'python file.py args...'
+        job_script += " ".join(command)  # 'python file.py args...'
 
         # Submit command to SGE.
         # Note: submitting job using drmaa didn't work because we weren't able

tests/test_schedulers.py

@@ -92,7 +92,7 @@ def test_local_scheduler(test_dir):
 
     s = sherpa.schedulers.LocalScheduler()
 
-    job_id = s.submit_job("python {}/test.py".format(test_dir),
+    job_id = s.submit_job(["python", "{}/test.py".format(test_dir)],
                           env={'SHERPA_TRIAL_ID': '3'})
 
     assert s.get_status(job_id) == sherpa.schedulers._JobStatus.running
@@ -101,7 +101,7 @@ def test_local_scheduler(test_dir):
     testlogger.debug(s.get_status(job_id))
     assert s.get_status(job_id) == sherpa.schedulers._JobStatus.finished
 
-    job_id = s.submit_job("python {}/test.py".format(test_dir))
+    job_id = s.submit_job(["python", "{}/test.py".format(test_dir)])
     time.sleep(1)
     s.kill_job(job_id)
     time.sleep(1)

tests/test_sherpa.py

@@ -170,7 +170,7 @@ def get_test_study():
     mock_algorithm.get_suggestion.return_value = {'a': 1, 'b': 2}
     mock_stopping_rule = mock.MagicMock()
 
-    s = sherpa.Study(parameters=get_test_parameters(),
+    s = sherpa.Study(parameters=list(get_test_parameters()),
                      algorithm=mock_algorithm,
                      stopping_rule=mock_stopping_rule,
                      lower_is_better=True)
@@ -194,7 +194,7 @@ def test_runner_update_results():
 
     r = sherpa.core._Runner(study=get_test_study(), scheduler=mock.MagicMock(),
                       database=mock_db, max_concurrent=1,
-                      command="python test.py")
+                      command=["python", "test.py"])
 
     # new trial
     t = get_test_trial()
@@ -220,7 +220,7 @@ def test_runner_update_active_trials():
 
     r = sherpa.core._Runner(study=mock_study, scheduler=mock_scheduler,
                       database=mock.MagicMock(), max_concurrent=1,
-                      command="python test.py")
+                      command=["python", "test.py"])
 
     t = get_test_trial()
     r._all_trials[t.id] = {'trial': t, 'job_id': None}
@@ -242,7 +242,7 @@ def test_runner_stop_bad_performers():
                       scheduler=mock.MagicMock(),
                       database=mock.MagicMock(),
                       max_concurrent=1,
-                      command="python test.py")
+                      command=["python", "test.py"])
 
     # setup
     t = get_test_trial()
@@ -262,7 +262,7 @@ def test_runner_stop_bad_performers():
 
 def test_runner_submit_new_trials():
     mock_scheduler = mock.MagicMock()
-    mock_scheduler.submit.side_effect = ['job1', 'job2', 'job3']
+    mock_scheduler.submit_job.side_effect = ['job1', 'job2', 'job3']
     mock_study = mock.MagicMock()
     mock_study.get_suggestion.side_effect = [get_test_trial(1),
                                              get_test_trial(2),
@@ -272,12 +272,12 @@ def test_runner_submit_new_trials():
                       scheduler=mock_scheduler,
                       database=mock.MagicMock(),
                       max_concurrent=3,
-                      command="python test.py")
+                      command=["python", "test.py"])
 
     r.submit_new_trials()
 
-    mock_scheduler.submit.has_calls([mock.call("python test.py"),
-                                     mock.call("python test.py"),
-                                     mock.call("python test.py")])
+    mock_scheduler.submit_job.has_calls([mock.call(["python", "test.py"]),
+                                         mock.call(["python", "test.py"]),
+                                         mock.call(["python", "test.py"])])
     assert len(r._active_trials) == 3
     assert len(r._all_trials) == 3