adding methods to ShellOutputSpec and TaskBase that allows to predic the generated output names before running the task

Author: djarecka

pydra/engine/core.py

@@ -311,9 +311,27 @@ def set_state(self, splitter, combiner=None):
 
     @property
     def output_names(self):
-        """Get the names of the outputs generated by the task."""
+        """Get the names of the outputs from the task's output_spec
+            (not everything has to be generated, see generated_output_names).
+        """
         return [f.name for f in attr.fields(make_klass(self.output_spec))]
 
+    @property
+    def generated_output_names(self):
+        """ Get the names of the outputs generated by the task.
+            If the spec doesn't have generated_output_names method,
+            it uses output_names.
+            The results depends on the input provided to the task
+        """
+        output_klass = make_klass(self.output_spec)
+        if hasattr(output_klass, "generated_output_names"):
+            output = output_klass(**{f.name: None for f in attr.fields(output_klass)})
+            return output.generated_output_names(
+                inputs=self.inputs, output_dir=self.output_dir
+            )
+        else:
+            return self.output_names
+
     @property
     def can_resume(self):
         """Whether the task accepts checkpoint-restart."""

pydra/engine/specs.py

@@ -408,7 +408,7 @@ def collect_additional_outputs(self, inputs, output_dir):
                         raise AttributeError(
                             "File has to have default value or metadata"
                         )
-                    elif not fld.default == attr.NOTHING:
+                    elif fld.default != attr.NOTHING:
                         additional_out[fld.name] = self._field_defaultvalue(
                             fld, output_dir
                         )
@@ -420,6 +420,34 @@ def collect_additional_outputs(self, inputs, output_dir):
                     raise Exception("not implemented (collect_additional_output)")
         return additional_out
 
+    def generated_output_names(self, inputs, output_dir):
+        """ Returns a list of all outputs that will be generated by the task.
+            Takes into account the task input and the requires list for the output fields.
+            TODO: should be in all Output specs?
+        """
+        output_names = ["return_code", "stdout", "stderr"]
+        for fld in attr_fields(self):
+            if fld.name not in ["return_code", "stdout", "stderr"]:
+                if fld.type is File:
+                    # assuming that field should have either default or metadata, but not both
+                    if (
+                        fld.default is None or fld.default == attr.NOTHING
+                    ) and not fld.metadata:  # TODO: is it right?
+                        raise AttributeError(
+                            "File has to have default value or metadata"
+                        )
+                    elif fld.default != attr.NOTHING:
+                        output_names.append(fld.name)
+                    elif (
+                        fld.metadata
+                        and self._field_metadata(fld, inputs, output_dir)
+                        != attr.NOTHING
+                    ):
+                        output_names.append(fld.name)
+                else:
+                    raise Exception("not implemented (collect_additional_output)")
+        return output_names
+
     def _field_defaultvalue(self, fld, output_dir):
         """Collect output file if the default value specified."""
         if not isinstance(fld.default, (str, Path)):

pydra/engine/tests/test_shelltask.py

@@ -2654,6 +2654,12 @@ def test_shell_cmd_inputspec_outputspec_2():
     )
     shelly.inputs.file1 = "new_file_1.txt"
     shelly.inputs.file2 = "new_file_2.txt"
+    # all fileds from output_spec should be in output_names and generated_output_names
+    assert (
+        shelly.output_names
+        == shelly.generated_output_names
+        == ["return_code", "stdout", "stderr", "newfile1", "newfile2"]
+    )
 
     res = shelly()
     assert res.output.stdout == ""
@@ -2714,6 +2720,20 @@ def test_shell_cmd_inputspec_outputspec_2a():
         output_spec=my_output_spec,
     )
     shelly.inputs.file1 = "new_file_1.txt"
+    # generated_output_names shoule know that newfile2 will not be generated
+    assert shelly.output_names == [
+        "return_code",
+        "stdout",
+        "stderr",
+        "newfile1",
+        "newfile2",
+    ]
+    assert shelly.generated_output_names == [
+        "return_code",
+        "stdout",
+        "stderr",
+        "newfile1",
+    ]
 
     res = shelly()
     assert res.output.stdout == ""
@@ -2834,6 +2854,20 @@ def test_shell_cmd_inputspec_outputspec_3a():
     )
     shelly.inputs.file1 = "new_file_1.txt"
     shelly.inputs.file2 = "new_file_2.txt"
+    # generated_output_names shoule know that newfile2 will not be generated
+    assert shelly.output_names == [
+        "return_code",
+        "stdout",
+        "stderr",
+        "newfile1",
+        "newfile2",
+    ]
+    assert shelly.generated_output_names == [
+        "return_code",
+        "stdout",
+        "stderr",
+        "newfile1",
+    ]
 
     res = shelly()
     assert res.output.stdout == ""
@@ -2884,6 +2918,12 @@ def test_shell_cmd_inputspec_outputspec_4():
     )
     shelly.inputs.file1 = "new_file_1.txt"
     shelly.inputs.additional_inp = 2
+    # generated_output_names should be the same as output_names
+    assert (
+        shelly.output_names
+        == shelly.generated_output_names
+        == ["return_code", "stdout", "stderr", "newfile1"]
+    )
 
     res = shelly()
     assert res.output.stdout == ""