fix: multiple small improvements to the wizard

 - Move the wavs question after fully processing the filelist, including
   permission.
 - Catch the case where we don't have permission to write the directory with a
   friendly error message.
 - Make the ValidateWavsStep question use the same type of prompt as the other
   similar questions, for better UX consistency
 - If we accept a wavs dir / filelist combo with missing wav files, give a
   bright error message to make it clear future problems will happen.
 - Don't save the config if no dataset was created.

The ideas implemented here were refined in consultation with Samuel Larkin.

everyvoice/tests/test_wizard.py

@@ -284,6 +284,13 @@ def test_output_path_step(self):
                 self.assertFalse(step.validate(tmpdirname))
             os.unlink(dataset_file)
 
+            # Bad case 3: file under read-only directory
+            ro_dir = Path(tmpdirname) / "read-only"
+            ro_dir.mkdir(mode=0x555)
+            with capture_stdout() as out:
+                self.assertFalse(step.validate(str(ro_dir)))
+            self.assertIn("could not create", out.getvalue())
+
             # Good case
             with capture_stdout() as stdout:
                 with monkeypatch(step, "prompt", Say(tmpdirname)):
@@ -295,11 +302,16 @@ def test_output_path_step(self):
 
     def test_more_data_step(self):
         """Exercise giving an invalid response and a yes response to more data."""
-        tour = Tour("testing", [basic.MoreDatasetsStep()])
-        step = tour.steps[0]
+        tour = Tour(
+            "testing",
+            [dataset.FilelistStep(state_subset="dataset_0"), basic.MoreDatasetsStep()],
+        )
+
+        step = tour.steps[1]
         self.assertFalse(step.validate("foo"))
         self.assertTrue(step.validate("yes"))
         self.assertEqual(len(step.children), 0)
+
         with patch_menu_prompt(0):  # answer 0 is "no"
             step.run()
         self.assertEqual(len(step.children), 1)
@@ -309,6 +321,15 @@ def test_more_data_step(self):
             step.run()
         self.assertGreater(len(step.descendants), 10)
 
+    def test_no_data_to_save(self):
+        """When the tour created no datasets at all, saving the config is skipped."""
+        tour = Tour("testing", [basic.MoreDatasetsStep()])
+        step = tour.steps[0]
+        with patch_menu_prompt(0), capture_stdout() as out:  # answer 0 is "no"
+            step.run()
+        self.assertEqual(len(step.children), 0)
+        self.assertIn("No dataset to save", out.getvalue())
+
     def test_dataset_name(self):
         step = dataset.DatasetNameStep()
         with monkeypatch(builtins, "input", Say(("", "bad/name", "good-name"), True)):
@@ -368,10 +389,6 @@ def test_sample_rate_config(self):
     def test_dataset_subtour(self):
         tour = Tour("unit testing", steps=dataset.get_dataset_steps())
 
-        wavs_dir_step = find_step(SN.wavs_dir_step, tour.steps)
-        with monkeypatch(wavs_dir_step, "prompt", Say(str(self.data_dir))):
-            wavs_dir_step.run()
-
         filelist = str(self.data_dir / "unit-test-case1.psv")
         filelist_step = find_step(SN.filelist_step, tour.steps)
         monkey = monkeypatch(filelist_step, "prompt", Say(filelist))
@@ -413,10 +430,14 @@ def test_dataset_subtour(self):
             step.run()
         self.assertEqual(step.state["filelist_headers"][2], "text")
 
+        wavs_dir_step = find_step(SN.wavs_dir_step, tour.steps)
+        with monkeypatch(wavs_dir_step, "prompt", Say(str(self.data_dir))):
+            wavs_dir_step.run()
+
         validate_wavs_step = find_step(SN.validate_wavs_step, tour.steps)
-        with monkeypatch(builtins, "input", Say("no")), capture_stdout() as out:
+        with patch_menu_prompt(1), capture_stdout() as out:
             validate_wavs_step.run()
-        self.assertEqual(step.state[SN.validate_wavs_step], "no")
+        self.assertEqual(step.state[SN.validate_wavs_step][:2], "No")
         self.assertIn("Warning: 3 wav files were not found", out.getvalue())
 
         text_representation_step = find_step(
@@ -605,7 +626,9 @@ def test_prompt(self):
             )
             self.assertEqual(answer, 1)
 
-    def monkey_run_tour(self, name: str, steps_and_answers: list[StepAndAnswer]):
+    def monkey_run_tour(
+        self, name: str, steps_and_answers: list[StepAndAnswer]
+    ) -> tuple[Tour, str]:
         """Create and run a tour with the monkey answers given
 
         Args:
@@ -677,10 +700,9 @@ def test_monkey_tour_1(self):
 
     def test_monkey_tour_2(self):
         data_dir = Path(__file__).parent / "data"
-        tour, log = self.monkey_run_tour(
+        tour, out = self.monkey_run_tour(
             "monkey tour 2",
             [
-                StepAndAnswer(dataset.WavsDirStep(), Say(str(data_dir))),
                 StepAndAnswer(
                     dataset.FilelistStep(),
                     Say(str(data_dir / "metadata.psv")),
@@ -699,9 +721,10 @@ def test_monkey_tour_2(self):
                     Say("no"),
                     children_answers=[RecursiveAnswers(Say("eng"))],
                 ),
+                StepAndAnswer(dataset.WavsDirStep(), Say(str(data_dir))),
                 StepAndAnswer(
                     dataset.ValidateWavsStep(),
-                    monkeypatch(builtins, "input", Say("Yes")),
+                    patch_menu_prompt(0),  # 0 is Yes
                     children_answers=[
                         RecursiveAnswers(Say(str(data_dir / "lj/wavs"))),
                         RecursiveAnswers(null_patch()),
@@ -717,8 +740,10 @@ def test_monkey_tour_2(self):
             ],
         )
 
-        self.assertIn("Warning: 5 wav files were not found", log)
-        self.assertIn("Great! All audio files found in directory", log)
+        tree = str(RenderTree(tour.root))
+        self.assertIn("├── Validate Wavs Step", tree)
+        self.assertIn("│   └── Validate Wavs Step", tree)
+        self.assertIn("Great! All audio files found in directory", out)
 
         # print(tour.state)
         self.assertEqual(len(tour.state["filelist_data"]), 5)

everyvoice/wizard/basic.py

@@ -148,20 +148,26 @@ def validate(self, response) -> bool:
             return False
         assert self.state is not None, "OutputPathStep requires NameStep"
         output_path = path / self.state.get(StepNames.name_step, "DEFAULT_NAME")
+        TRY_AGAIN = "Please choose another output directory or start again and choose a different project name."
         if output_path.exists():
-            print(
-                f"Sorry, '{output_path}' already exists. Please choose another output directory or start again and choose a different project name."
-            )
+            print(f"Sorry, '{output_path}' already exists.", TRY_AGAIN)
+            return False
+
+        # We create the output directory in validate() instead of effect() so that
+        # failure can be reported to the user and the question asked again if necessary.
+        try:
+            output_path.mkdir(parents=True, exist_ok=True)
+        except OSError as e:
+            print(f"Sorry, could not create '{output_path}': {e}.", TRY_AGAIN)
             return False
+
+        self.output_path = output_path
         return True
 
     def effect(self):
-        assert self.state is not None, "OutputPathStep requires NameStep"
-        output_path = Path(self.response) / self.state.get(
-            StepNames.name_step, "DEFAULT_NAME"
+        print(
+            f"The Configuration Wizard 🧙 will put your files here: '{self.output_path}'"
         )
-        output_path.mkdir(parents=True, exist_ok=True)
-        print(f"The Configuration Wizard 🧙 will put your files here: '{output_path}'")
 
 
 class ConfigFormatStep(Step):
@@ -450,6 +456,8 @@ def effect(self):
                 + [MoreDatasetsStep()],
                 self,
             )
+        elif len([key for key in self.state.keys() if key.startswith("dataset_")]) == 0:
+            print("No dataset to save, exiting.")
         else:
             self.tour.add_step(
                 ConfigFormatStep(name=StepNames.config_format_step), self

everyvoice/wizard/dataset.py

@@ -5,6 +5,9 @@
 from unicodedata import normalize
 
 import questionary
+import rich
+from rich.panel import Panel
+from rich.style import Style
 from tqdm import tqdm
 
 from everyvoice.config.type_definitions import DatasetTextRepresentation
@@ -295,21 +298,36 @@ def wav_file_early_validation(self) -> int:
     def prompt(self):
         error_count = self.wav_file_early_validation()
         if error_count:
-            return input("Do you want to pick a different wavs directory? ").lower()
-        return "ok"
+            return get_response_from_menu_prompt(
+                title="Do you want to pick a different wavs directory?",
+                choices=(
+                    "Yes",
+                    "No, I will fix my audio basenames or add missing audio files later.",
+                ),
+            )
+        else:
+            return "OK"
 
     def validate(self, response):
-        return response in ("y", "yes", "n", "no", "ok")
+        return response[:3] in ("OK", "Yes", "No,")
 
     def effect(self):
-        if self.response[:1] == "y":
+        if self.response == "Yes":
             self.tour.add_steps(
                 [
                     WavsDirStep(state_subset=self.state_subset),
                     ValidateWavsStep(state_subset=self.state_subset),
                 ],
                 self,
             )
+        elif self.response.startswith("No"):
+            rich.print(
+                Panel(
+                    "Continuing despite missing audio files. Make sure you fix your filelist later or add missing audio files, otherwise entries in your filelist with missing audio files will be skipped during preprocessing and therefore be ignored during training.",
+                    title="Missing audio files",
+                    border_style=Style(color="#EF1010"),
+                )
+            )
 
 
 class FilelistTextRepresentationStep(Step):
@@ -711,11 +729,11 @@ def effect(self):
 
 def get_dataset_steps(dataset_index=0):
     return [
-        WavsDirStep(state_subset=f"dataset_{dataset_index}"),
         FilelistStep(state_subset=f"dataset_{dataset_index}"),
         [
             DatasetPermissionStep(state_subset=f"dataset_{dataset_index}"),
             FilelistFormatStep(state_subset=f"dataset_{dataset_index}"),
+            WavsDirStep(state_subset=f"dataset_{dataset_index}"),
             ValidateWavsStep(state_subset=f"dataset_{dataset_index}"),
             FilelistTextRepresentationStep(state_subset=f"dataset_{dataset_index}"),
             TextProcessingStep(state_subset=f"dataset_{dataset_index}"),

everyvoice/wizard/prompts.py

@@ -1,5 +1,5 @@
 import sys
-from typing import List, Tuple, Union
+from typing import Union
 
 import simple_term_menu
 from questionary import Style
@@ -27,18 +27,18 @@
 
 def get_response_from_menu_prompt(
     prompt_text: str = "",
-    choices: Tuple[str, ...] = (),
+    choices: tuple[str, ...] = (),
     title: str = "",
     multi=False,
     search=False,
     return_indices=False,
-) -> Union[int, str, List[int], List[str]]:
+) -> Union[int, str, list[int], list[str]]:
     """Given some prompt text and a list of choices, create a simple terminal window
        and return the index of the choice
 
     Args:
         prompt_text (str): rich prompt text to print before menu
-        choices (List[str]): choices to display
+        choices (list[str]): choices to display
 
     Returns:
         int: index of choice