code coverage

mlprodict/asv_benchmark/common_asv_skl.py

@@ -93,7 +93,7 @@ def _get_dataset(self, nf, dtype):
         Xt = X_test.astype(xdtype)
         yt = y_test.astype(self.par_ydtype)
         if X_train.shape[0] < X_train.shape[1]:
-            raise RuntimeError(
+            raise RuntimeError(  # pragma: no cover
                 "Unable to train a model with less observations than features "
                 "shape=%r." % (X_train.shape, ))
         return (X_train, y_train), (Xt, yt)

mlprodict/cli/validate.py

@@ -40,9 +40,11 @@ def _save(df, name):
         elif ext == '.csv':
             df.to_csv(name, index=False)
         else:
-            raise ValueError("Unexpected extension in %r." % name)
+            raise ValueError(  # pragma: no cover
+                "Unexpected extension in %r." % name)
         if verbose > 1:
-            fLOG("[mlprodict] wrote '{}'".format(name))
+            fLOG(  # pragma: no cover
+                "[mlprodict] wrote '{}'".format(name))
 
     from pyquickhelper.loghelper import run_cmd
     from pyquickhelper.loghelper.run_cmd import get_interpreter_path
@@ -52,7 +54,7 @@ def _save(df, name):
     register_converters()
     try:
         register_rewritten_operators()
-    except KeyError:
+    except KeyError:  # pragma: no cover
         warnings.warn("converter for HistGradientBoosting* not not exist. "
                       "Upgrade sklearn-onnx")
 
@@ -79,7 +81,8 @@ def _save(df, name):
         if black_list is not None and op in black_list:
             continue
         if verbose > 0:
-            pbar.set_description("[%s]" % (op + " " * (25 - len(op))))
+            pbar.set_description(  # pragma: no cover
+                "[%s]" % (op + " " * (25 - len(op))))
 
         loop_out_raw = os.path.join(
             dump_dir, "bench_raw_%s_%s.csv" % (runtime, op))
@@ -89,9 +92,9 @@ def _save(df, name):
                '--benchmark=1 --dump_folder={3} --runtime={4} --models={5}'.format(
                    get_interpreter_path(), loop_out_raw, loop_out_sum, dump_dir, runtime, op))
         if verbose > 1:
-            fLOG("[mlprodict] cmd '{}'.".format(cmd))
+            fLOG("[mlprodict] cmd '{}'.".format(cmd))  # pragma: no cover
         out, err = run_cmd(cmd, wait=True, fLOG=None)
-        if not os.path.exists(loop_out_sum):
+        if not os.path.exists(loop_out_sum):  # pragma: no cover
             if verbose > 2:
                 fLOG("[mlprodict] unable to find '{}'.".format(loop_out_sum))
             if verbose > 1:
@@ -580,7 +583,7 @@ def latency(model, law='normal', size=1, number=10, repeat=10, max_time=0,
             max_time = None
 
     if law != "normal":
-        raise ValueError(
+        raise ValueError(  # pragma: no cover
             "Only law='normal' is supported, not %r." % law)
 
     if profiling in ('name', 'type') and profile_output in (None, ''):
@@ -599,7 +602,7 @@ def latency(model, law='normal', size=1, number=10, repeat=10, max_time=0,
         gr = gr.reset_index(drop=False)
         if ext == '.csv':
             gr.to_csv(profile_output, index=False)
-        elif ext == '.xlsx':
+        elif ext == '.xlsx':  # pragma: no cover
             gr.to_excel(profile_output, index=False)
         else:
             raise ValueError(  # pragma: no cover

mlprodict/npy/xop.py

@@ -1137,7 +1137,7 @@ def _node_to_graph_preprocess_list(inputs):
     def _node_to_graph_process_input(inputs, set_inputs, node, inp,
                                      new_inputs, new_stack, inputs_dtype):
         if inputs is None and inputs_dtype is None:
-            raise RuntimeError(
+            raise RuntimeError(  # pragma: no cover
                 "Both inputs and inputs_dtype cannot be None at the same time "
                 "for inp=%r." % (inp, ))
         if isinstance(inp, OnnxOperator):

mlprodict/onnxrt/ops_shape/shape_container.py

@@ -248,7 +248,7 @@ def vars_in_values(values):
         for k, v in self.shapes.items():
             try:
                 results[k] = v.resolve(variables)
-            except RuntimeError as e:
+            except RuntimeError as e:  # pragma: no cover
                 raise RuntimeError(
                     "Unable to resolve shapes and constraints:\n%s"
                     "" % pprint.pformat(self.shapes)) from e

mlprodict/onnxrt/ops_shape/shape_result.py

@@ -119,7 +119,7 @@ class ShapeResult:
     def __init__(self, name, shape=None, dtype=None, sparse=False,
                  mtype=OnnxKind.Tensor, constraints=None):
         if not isinstance(name, str):
-            raise TypeError(
+            raise TypeError(  # pragma: no cover
                 "name must be a string not %r." % type(name))
         if not isinstance(sparse, bool):
             raise TypeError(  # pragma: no cover
@@ -184,7 +184,7 @@ def n_dims(self):
         Raises an exception otherwise.
         """
         if self.mtype != OnnxKind.Tensor:
-            raise ShapeInferenceException(
+            raise ShapeInferenceException(  # pragma: no cover
                 "This shape is not a tensor %r." % self)
         return len(self.shape)
 

mlprodict/onnxrt/validate/validate.py

@@ -681,13 +681,13 @@ def fct_batch(se=sess, xo=Xort_test, it=init_types):  # pylint: disable=W0102
                 try:
                     max_rel_diff = measure_relative_difference(
                         ypred, opred[:, 1])
-                except AttributeError:
+                except AttributeError:  # pragma: no cover
                     max_rel_diff = numpy.nan
             else:
                 try:
                     max_rel_diff = measure_relative_difference(
                         ypred, opred)
-                except AttributeError:
+                except AttributeError:  # pragma: no cover
                     max_rel_diff = numpy.nan
 
             if max_rel_diff >= 1e9 and debug:  # pragma: no cover