Get pipe reader working on windows where we cant wait for data to become available on a non-blocking pipe

src/calibre/gui2/tts/piper.py

@@ -566,7 +566,7 @@ class PiperEmbedded:
         lang = get_lang()
         lang = canonicalize_lang(lang) or lang
         self._default_voice = self._voice_for_lang.get(lang) or self._voice_for_lang['eng']
-        self._current_voice = self._process = None
+        self._current_voice = self._process = self._pipe_reader = None
         self._current_audio_rate = 0
 
     def resolve_voice(self, lang: str, voice_name: str) -> Voice:
@@ -610,10 +610,8 @@ class PiperEmbedded:
             self._process.stdin.write(UTTERANCE_SEPARATOR)
             self._process.stdin.flush()
             stderr_data = b''
-            buf, piper_done, errors_from_piper = [], [], []
-
-            def stdout_callback(data: bytes) -> None:
-                buf.append(data)
+            buf = io.BytesIO()
+            piper_done, errors_from_piper = [], []
 
             def stderr_callback(data: bytes) -> bool:
                 nonlocal stderr_data
@@ -621,11 +619,11 @@ class PiperEmbedded:
                 return not piper_done
 
             try:
-                unix_pipe_readers(self._process.stdout, self._process.stderr, stdout_callback, stderr_callback)
+                self._pipe_reader(buf.write, stderr_callback)
             except Exception as e:
                 raise Exception(f'Reading output from piper process failed with error: {e} and STDERR: ' + '\n'.join(errors_from_piper))
 
-            raw_data = b''.join(buf)
+            raw_data = buf.getvalue()
             if needs_conversion:
                 raw_data = resample_raw_audio_16bit(raw_data, self._current_audio_rate, sample_rate)
             yield raw_data, duration_of_raw_audio_data(raw_data, sample_rate)
@@ -637,6 +635,8 @@ class PiperEmbedded:
 
     def shutdown(self):
         if self._process is not None:
+            self._pipe_reader.close()
+            self._pipe_reader = None
             self._process.stdin.close()
             self._process.stdout.close()
             self._process.stderr.close()
@@ -652,25 +652,111 @@ class PiperEmbedded:
         self._current_audio_rate, cmdline = piper_process_metadata(model_path, config_path, self._embedded_settings, self._current_voice)
         import subprocess
         self._process = subprocess.Popen(cmdline, stdout=subprocess.PIPE, stderr=subprocess.PIPE, stdin=subprocess.PIPE)
+        self._pipe_reader = (ThreadedPipeReader if iswindows else PipeReader)(self._process.stdout, self._process.stderr)
 
 
-def unix_pipe_readers(stdout: BinaryIO, stderr: BinaryIO, stdout_callback, stderr_callback, timeout: float = 10.) -> None:
-    from select import select
-    out, err = stdout.fileno(), stderr.fileno()
-    os.set_blocking(out, False)
-    os.set_blocking(err, False)
-    readers = out, err
-    while True:
-        r, _, _ = select(readers, (), (), timeout)
-        if not r:
-            raise TimeoutError('Timed out waiting for output from piper process')
-        if out in r:
-            data = stdout.read(io.DEFAULT_BUFFER_SIZE)
-            stdout_callback(data)
-        if err in r:
-            data = stderr.read(io.DEFAULT_BUFFER_SIZE)
-            if not stderr_callback(data):
+class PipeReader:
+
+    TIMEOUT = 10.  # seconds
+
+    def __init__(self, stdout: BinaryIO, stderr: BinaryIO):
+        self.stdout_fd = stdout.fileno()
+        self.stderr_fd = stderr.fileno()
+        os.set_blocking(self.stdout_fd, False)
+        os.set_blocking(self.stderr_fd, False)
+
+    def close(self):
+        self.stderr_fd = self.stdout_fd = -1
+
+    def __call__(self, stdout_callback, stderr_callback):
+        from select import select
+        out, err = self.stdout_fd, self.stderr_fd
+        readers = out, err
+        buf = memoryview(bytearray(io.DEFAULT_BUFFER_SIZE))
+
+        def readall(fd: int) -> bytes:
+            output = io.BytesIO()
+            while True:
+                try:
+                    num = os.readv(fd, (buf,))
+                except BlockingIOError:
+                    break
+                else:
+                    output.write(buf[:num])
+            return output.getvalue()
+
+        def readinto(fd: int, callback) -> None:
+            while True:
+                try:
+                    num = os.readv(fd, (buf,))
+                except BlockingIOError:
+                    break
+                else:
+                    callback(buf[:num])
+
+        while True:
+            r, _, _ = select(readers, (), (), self.TIMEOUT)
+            if not r:
+                raise TimeoutError('Timed out waiting for output from piper process')
+            if out in r:
+                readinto(out, stdout_callback)
+            if err in r:
+                data = readall(err)
+                if not stderr_callback(data):
+                    # In case there is new data written to stdout
+                    readinto(out, stdout_callback)
+                    break
+
+
+class ThreadedPipeReader(PipeReader):
+
+    def __init__(self, stdout: BinaryIO, stderr: BinaryIO):
+        from queue import Queue
+        from threading import Event, Thread
+        self.shutting_down = Event()
+        self.queue = Queue()
+        self.stdout_thread = t = Thread(target=self._reader, args=(stdout.fileno(), True), daemon=True)
+        t.start()
+        self.stderr_thread = t = Thread(target=self._reader, args=(stderr.fileno(), False), daemon=True)
+        t.start()
+
+    def close(self):
+        self.shutting_down.set()
+
+    def __call__(self, stdout_callback, stderr_callback):
+        from queue import Empty
+        while True:
+            data, is_stdout, err = self.queue.get(True, self.TIMEOUT)
+            if err is not None:
+                raise err
+            if data:
+                if is_stdout:
+                    stdout_callback(data)
+                else:
+                    if not stderr_callback(data):
+                        # in case more data was written to stdout
+                        while True:
+                            try:
+                                data, is_stdout, err = self.queue.get_nowait()
+                            except Empty:
+                                break
+                            if err is not None:
+                                raise err
+                            if is_stdout:
+                                stdout_callback(data)
+                        break
+
+    def _reader(self, pipe_fd: int, is_stdout: bool):
+        while not self.shutting_down.is_set():
+            try:
+                data = os.read(pipe_fd, io.DEFAULT_BUFFER_SIZE)
+            except OSError as e:
+                if not self.shutting_down.is_set():
+                    self.queue.put((b'', is_stdout, e))
                 break
+            else:
+                self.queue.put((data, is_stdout, None))
+
 
 
 def duration_of_raw_audio_data(data: bytes, sample_rate: int = HIGH_QUALITY_SAMPLE_RATE, bytes_per_sample: int = 2, num_channels: int = 1) -> float:
@@ -690,7 +776,7 @@ def develop_embedded():
     for data, duration in p.text_to_raw_audio_data((
         'Hello, good day to you.', 'This is the second sentence.', 'This is the final sentence.'
     )):
-        print(duration, len(data))
+        print(f'{duration=} {len(data)=}')
         all_data.append(data)
         sz += len(data)
     all_data[0] = wav_header_for_pcm_data(sz, HIGH_QUALITY_SAMPLE_RATE)