Fix table row inference benchmark using wrong model path

sdks/python/apache_beam/testing/benchmarks/inference/table_row_inference_benchmark.py

@@ -72,32 +72,36 @@ def __init__(self):
         metrics_namespace=self.metrics_namespace,
         is_streaming=False,
         pcollection='RunInference/BeamML_RunInference_Postprocess-0.out0')
-    self.is_streaming = ((self.pipeline.get_option('mode') or
-                          'batch') == 'streaming')
+    self.opts = self.pipeline.get_pipeline_options().view_as(
+        TableRowInferenceOptions)
+    mode = self.opts.mode or 'batch'
+    self.is_streaming = mode == 'streaming'
     if self.is_streaming:
-      self.subscription = self.pipeline.get_option('input_subscription')
+      self.subscription = self.opts.input_subscription
 
   def test(self):
     """Execute the table row inference pipeline for benchmarking."""
-    extra_opts = {}
-
-    mode = self.pipeline.get_option('mode') or 'batch'
-    extra_opts['mode'] = mode
+    mode = self.opts.mode or 'batch'
+    extra_opts = {'mode': mode}
 
     if mode == 'streaming':
-      extra_opts['input_subscription'] = self.pipeline.get_option(
-          'input_subscription')
-      extra_opts['window_size_sec'] = int(
-          self.pipeline.get_option('window_size_sec') or 60)
-      extra_opts['trigger_interval_sec'] = int(
-          self.pipeline.get_option('trigger_interval_sec') or 30)
-    else:
-      extra_opts['input_file'] = self.pipeline.get_option('input_file')
-
-    for opt in ['output_table', 'model_path', 'feature_columns']:
-      val = self.pipeline.get_option(opt)
-      if val:
-        extra_opts[opt] = val
+      if self.opts.input_subscription:
+        extra_opts['input_subscription'] = self.opts.input_subscription
+      extra_opts['window_size_sec'] = (
+          self.opts.window_size_sec
+          if self.opts.window_size_sec is not None else 60)
+      extra_opts['trigger_interval_sec'] = (
+          self.opts.trigger_interval_sec
+          if self.opts.trigger_interval_sec is not None else 30)
+    elif self.opts.input_file:
+      extra_opts['input_file'] = self.opts.input_file
+
+    if self.opts.output_table:
+      extra_opts['output_table'] = self.opts.output_table
+    if self.opts.model_path:
+      extra_opts['model_path'] = self.opts.model_path
+    if self.opts.feature_columns:
+      extra_opts['feature_columns'] = self.opts.feature_columns
 
     self.result = table_row_inference.run(
         self.pipeline.get_full_options_as_args(**extra_opts),

sdks/python/apache_beam/testing/test_pipeline.py

@@ -209,7 +209,9 @@ def get_option(self, opt_name, bool_option=False):
       None if option is not found in existing option list which is generated
       by parsing value of argument `test-pipeline-options`.
     """
-    parser = argparse.ArgumentParser()
+    # Parse one flag at a time; disable prefix matching so e.g. --mode does
+    # not satisfy --model_path when both appear in options_list.
+    parser = argparse.ArgumentParser(allow_abbrev=False)
     opt_name = opt_name[:2] if opt_name[:2] == '--' else opt_name
     # Option name should start with '--' when it's used for parsing.
     if bool_option: