Local paths in common voice (#3736)

* Merge generators for local files and streaming

* add the streaming parameter to _split_generators

* update common_voice

* patrick's comment:
- pass streaming to _generate_examples
- separate in two methods

* add is_streaming attribute to the dl managers

* revert the streaming parameter being passed to _split_generators

Co-authored-by: anton-l <aglozhkov@gmail.com>

Commit from https://github.com/huggingface/datasets/commit/e3c8e2541573b42b8dc23a4a29e197537d309bca

common_voice.py

@@ -15,6 +15,8 @@
 """ Common Voice Dataset"""
 
 
+import os
+
 import datasets
 from datasets.tasks import AutomaticSpeechRecognition
 
@@ -657,63 +659,135 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
 
     def _split_generators(self, dl_manager):
         """Returns SplitGenerators."""
-        archive = dl_manager.download(_DATA_URL.format(self.config.name))
-        path_to_data = "/".join(["cv-corpus-6.1-2020-12-11", self.config.name])
-        path_to_clips = "/".join([path_to_data, "clips"])
+        streaming = dl_manager.is_streaming
+        archive_path = dl_manager.download(_DATA_URL.format(self.config.name))
+        if streaming:
+            # Here we use iter_archive in streaming mode because dl_manager.download_and_extract
+            # doesn't work to stream TAR archives (we have to stream the files in the archive one by one).
+            #
+            # The iter_archive method returns an iterable of (path_within_archive, file_obj) for every
+            # file in the TAR archive.
+            #
+            archive_iterator = dl_manager.iter_archive(archive_path)
+            # we locate the data using the path within the archive
+            path_to_data = "/".join(["cv-corpus-6.1-2020-12-11", self.config.name])
+            path_to_clips = "/".join([path_to_data, "clips"])
+            metadata_filepaths = {
+                split: "/".join([path_to_data, f"{split}.tsv"])
+                for split in ["train", "test", "dev", "other", "validated", "invalidated"]
+            }
+        else:
+            # In non-streaming we can extract the archive locally as usual
+            extracted_dir = dl_manager.extract(archive_path)
+            archive_iterator = None
+            # we locate the data using the local path
+            path_to_data = os.path.join(extracted_dir, "cv-corpus-6.1-2020-12-11", self.config.name)
+            path_to_clips = os.path.join(path_to_data, "clips")
+            metadata_filepaths = {
+                split: os.path.join(path_to_data, f"{split}.tsv")
+                for split in ["train", "test", "dev", "other", "validated", "invalidated"]
+            }
 
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "files": dl_manager.iter_archive(archive),
-                    "filepath": "/".join([path_to_data, "train.tsv"]),
+                    "streaming": streaming,
+                    "archive_iterator": archive_iterator,
+                    "filepath": metadata_filepaths["train"],
                     "path_to_clips": path_to_clips,
                 },
             ),
             datasets.SplitGenerator(
                 name=datasets.Split.TEST,
                 gen_kwargs={
-                    "files": dl_manager.iter_archive(archive),
-                    "filepath": "/".join([path_to_data, "test.tsv"]),
+                    "streaming": streaming,
+                    "archive_iterator": archive_iterator,
+                    "filepath": metadata_filepaths["test"],
                     "path_to_clips": path_to_clips,
                 },
             ),
             datasets.SplitGenerator(
                 name=datasets.Split.VALIDATION,
                 gen_kwargs={
-                    "files": dl_manager.iter_archive(archive),
-                    "filepath": "/".join([path_to_data, "dev.tsv"]),
+                    "streaming": streaming,
+                    "archive_iterator": archive_iterator,
+                    "filepath": metadata_filepaths["dev"],
                     "path_to_clips": path_to_clips,
                 },
             ),
             datasets.SplitGenerator(
                 name="other",
                 gen_kwargs={
-                    "files": dl_manager.iter_archive(archive),
-                    "filepath": "/".join([path_to_data, "other.tsv"]),
+                    "streaming": streaming,
+                    "archive_iterator": archive_iterator,
+                    "filepath": metadata_filepaths["other"],
                     "path_to_clips": path_to_clips,
                 },
             ),
             datasets.SplitGenerator(
                 name="validated",
                 gen_kwargs={
-                    "files": dl_manager.iter_archive(archive),
-                    "filepath": "/".join([path_to_data, "validated.tsv"]),
+                    "streaming": streaming,
+                    "archive_iterator": archive_iterator,
+                    "filepath": metadata_filepaths["validated"],
                     "path_to_clips": path_to_clips,
                 },
             ),
             datasets.SplitGenerator(
                 name="invalidated",
                 gen_kwargs={
-                    "files": dl_manager.iter_archive(archive),
-                    "filepath": "/".join([path_to_data, "invalidated.tsv"]),
+                    "streaming": streaming,
+                    "archive_iterator": archive_iterator,
+                    "filepath": metadata_filepaths["invalidated"],
                     "path_to_clips": path_to_clips,
                 },
             ),
         ]
 
-    def _generate_examples(self, files, filepath, path_to_clips):
+    def _generate_examples(self, streaming, archive_iterator, filepath, path_to_clips):
         """Yields examples."""
+        if streaming:
+            yield from self._generate_examples_streaming(archive_iterator, filepath, path_to_clips)
+        else:
+            yield from self._generate_examples_non_streaming(filepath, path_to_clips)
+
+    def _generate_examples_non_streaming(self, filepath, path_to_clips):
+
+        data_fields = list(self._info().features.keys())
+
+        # audio is not a header of the csv files
+        data_fields.remove("audio")
+        path_idx = data_fields.index("path")
+
+        with open(filepath, encoding="utf-8") as f:
+            lines = f.readlines()
+            headline = lines[0]
+
+            column_names = headline.strip().split("\t")
+            assert (
+                column_names == data_fields
+            ), f"The file should have {data_fields} as column names, but has {column_names}"
+
+            for id_, line in enumerate(lines[1:]):
+                field_values = line.strip().split("\t")
+
+                # set absolute path for mp3 audio file
+                field_values[path_idx] = os.path.join(path_to_clips, field_values[path_idx])
+
+                # if data is incomplete, fill with empty values
+                if len(field_values) < len(data_fields):
+                    field_values += (len(data_fields) - len(field_values)) * ["''"]
+
+                result = {key: value for key, value in zip(data_fields, field_values)}
+
+                # set audio feature
+                result["audio"] = field_values[path_idx]
+
+                yield id_, result
+
+    def _generate_examples_streaming(self, archive_iterator, filepath, path_to_clips):
+        """Yields examples in streaming mode."""
         data_fields = list(self._info().features.keys())
 
         # audio is not a header of the csv files
@@ -722,7 +796,7 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
 
         all_field_values = {}
         metadata_found = False
-        for path, f in files:
+        for path, f in archive_iterator:
             if path == filepath:
                 metadata_found = True
                 lines = f.readlines()
@@ -752,5 +826,7 @@ class CommonVoice(datasets.GeneratorBasedBuilder):
 
                     # set audio feature
                     result["audio"] = {"path": path, "bytes": f.read()}
+                    # set path to None since the path doesn't exist locally in streaming mode
+                    result["path"] = None
 
                     yield path, result