More work on speech backends

src/calibre/gui2/tts/develop.py

@@ -6,11 +6,10 @@
 import re
 from itertools import count
 from PyQt5.Qt import (
-    QDialogButtonBox, QLabel, QMainWindow, Qt, QVBoxLayout, QWidget, pyqtSignal
+    QDialogButtonBox, QLabel, QMainWindow, Qt, QTimer, QVBoxLayout, QWidget,
+    pyqtSignal
 )
 
-from calibre import prepare_string_for_xml
-from calibre.constants import iswindows
 from calibre.gui2 import Application
 
 from .common import EventType
@@ -23,25 +22,22 @@ def add_markup(text):
     counter = count()
     pos_map = {}
     last = None
-    if iswindows:
-        bm = '<bookmark mark="{}"/>'
-    else:
-        bm = '<mark name="{}"/>'
+    bm = Client.mark_template
     for m in re.finditer(r'\w+', text):
         start, end = m.start(), m.end()
         if first:
             first = False
             if start:
-                buf.append(prepare_string_for_xml(text[:start]))
+                buf.append(Client.escape_marked_text(text[:start]))
         num = next(counter)
         buf.append(bm.format(num))
         pos_map[num] = start, end
-        buf.append(prepare_string_for_xml(m.group()))
+        buf.append(Client.escape_marked_text(m.group()))
         last = end
     if last is None:
-        buf.append(prepare_string_for_xml(text))
+        buf.append(Client.escape_marked_text(text))
     else:
-        buf.append(prepare_string_for_xml(text[last:]))
+        buf.append(Client.escape_marked_text(text[last:]))
     return ''.join(buf), pos_map
 
 
@@ -135,5 +131,25 @@ def main():
     tts.tts.shutdown()
 
 
+def headless():
+    app = Application([])
+    c = Client()
+    text = '[[sync 0x123456]]very [[sync 0x80]]good [[sync 0x81]]indeed'
+
+    def callback():
+        for ev in c.get_events():
+            if ev.type is EventType.mark:
+                print('mark:', hex(ev.data))
+            if ev.type in (EventType.end, EventType.cancel):
+                print(ev.type)
+                app.quit()
+
+    def run():
+        c.speak_marked_text(text, callback)
+    QTimer.singleShot(10, run)
+    QTimer.singleShot(5000, app.quit)
+    app.exec_()
+
+
 if __name__ == '__main__':
     main()

src/calibre/gui2/tts/linux.py

@@ -2,15 +2,24 @@
 # vim:fileencoding=utf-8
 # License: GPL v3 Copyright: 2020, Kovid Goyal <kovid at kovidgoyal.net>
 
+from calibre import prepare_string_for_xml
+
 from .common import Event, EventType
 from .errors import TTSSystemUnavailable
 
 
 class Client:
 
+    mark_template = '<mark name="{}"/>'
+
+    @classmethod
+    def escape_marked_text(cls, text):
+        return prepare_string_for_xml(text)
+
     def __init__(self):
         self.create_ssip_client()
         self.pending_events = []
+        self.status = {'synthesizing': False, 'paused': False}
 
     def create_ssip_client(self):
         from speechd.client import SpawnError, SSIPClient
@@ -37,12 +46,27 @@ class Client:
 
     def speak_simple_text(self, text):
         self.set_use_ssml(False)
-        self.ssip_client.speak(text)
+        self.pending_events = []
+        self.ssip_client.speak(text, self.update_status)
+
+    def update_status(self, callback_type, index_mark=None):
+        from speechd.client import CallbackType
+        if callback_type is CallbackType.BEGIN:
+            self.status = {'synthesizing': True, 'paused': False}
+        elif callback_type is CallbackType.END:
+            self.status = {'synthesizing': False, 'paused': False}
+        elif callback_type is CallbackType.CANCEL:
+            self.status = {'synthesizing': False, 'paused': False}
+        elif callback_type is CallbackType.PAUSE:
+            self.status = {'synthesizing': True, 'paused': True}
+        elif callback_type is CallbackType.RESUME:
+            self.status = {'synthesizing': True, 'paused': False}
 
     def speak_marked_text(self, text, callback):
         from speechd.client import CallbackType
 
         def callback_wrapper(callback_type, index_mark=None):
+            self.update_status(callback_type, index_mark)
             if callback_type is CallbackType.INDEX_MARK:
                 event = Event(EventType.mark, index_mark)
             elif callback_type is CallbackType.BEGIN:
@@ -61,6 +85,7 @@ class Client:
             callback()
 
         self.set_use_ssml(True)
+        self.pending_events = []
         self.ssip_client.speak(text, callback=callback_wrapper)
 
     def get_events(self):

src/calibre/gui2/tts/macos.py

@@ -2,15 +2,57 @@
 # vim:fileencoding=utf-8
 # License: GPL v3 Copyright: 2020, Kovid Goyal <kovid at kovidgoyal.net>
 
+from .common import Event, EventType
+
 
 class Client:
 
+    mark_template = '[[sync 0x{:x}]]'
+
+    @classmethod
+    def escape_marked_text(cls, text):
+        return text.replace('[[', ' [ [ ').replace(']]', ' ] ] ')
+
     def __init__(self):
         from calibre_extensions.cocoa import NSSpeechSynthesizer
-        self.nsss = NSSpeechSynthesizer()
+        self.nsss = NSSpeechSynthesizer(self.handle_message)
+        self.current_callback = None
+        self.pending_events = []
 
     def __del__(self):
         self.nsss = None
+    shutdown = __del__
+
+    def handle_message(self, message_type, data):
+        from calibre_extensions.cocoa import MARK, END
+        if self.current_callback is not None:
+            if message_type == MARK:
+                event = Event(EventType.mark, data)
+            elif message_type == END:
+                event = Event(EventType.end if data else EventType.cancel)
+            else:
+                return
+            self.pending_events.append(event)
+            self.current_callback()
 
     def speak_simple_text(self, text):
+        self.current_callback = None
+        self.pending_events = []
+        self.nsss.speak(text.replace('[[', '[').replace(']]', ']'))
+
+    def speak_marked_text(self, text, callback):
+        self.current_callback = callback
+        self.pending_events = []
         self.nsss.speak(text)
+
+    def get_events(self):
+        events = self.pending_events
+        self.pending_events = []
+        return events
+
+    @property
+    def status(self):
+        ans = self.nsss.status()
+        ans['synthesizing'] = ans.get('synthesizing', False)
+        ans['paused'] = ans.get('paused', False)
+        return ans

src/calibre/gui2/tts/nsss.m

@@ -11,30 +11,77 @@
 typedef struct {
     PyObject_HEAD
     NSSpeechSynthesizer *nsss;
+	PyObject *callback;
 } NSSS;
 
+typedef enum { MARK, END } MessageType;
 
 static PyTypeObject NSSSType = {
     PyVarObject_HEAD_INIT(NULL, 0)
 };
+
+static void
+dispatch_message(NSSS *self, MessageType which, unsigned long val) {
+	PyGILState_STATE state = PyGILState_Ensure();
+	PyObject *ret = PyObject_CallFunction(self->callback, "ik", which, val);
+	if (ret) Py_DECREF(ret);
+	else PyErr_Print();
+	PyGILState_Release(state);
+}
+
+@interface SynthesizerDelegate : NSObject <NSSpeechSynthesizerDelegate> {
+	NSSS *parent;
+}
+- (id)initWithNSSS:(NSSS *)x;
+@end
+
+@implementation SynthesizerDelegate
+
+- (id)initWithNSSS:(NSSS *)x {
+    self = [super init];
+    if (self) parent = x;
+    return self;
+}
+
+- (void)speechSynthesizer:(NSSpeechSynthesizer *)sender didFinishSpeaking:(BOOL)success {
+	dispatch_message(parent, END, success);
+}
+
+- (void)speechSynthesizer:(NSSpeechSynthesizer *)sender didEncounterSyncMessage:(NSString *)message {
+	NSError *err = nil;
+	NSNumber *syncProp = (NSNumber*) [sender objectForProperty: NSSpeechRecentSyncProperty error: &err];
+	if (syncProp && !err) dispatch_message(parent, MARK, syncProp.unsignedLongValue);
+}
+
+@end
 // }}}
 
 static PyObject *
 NSSS_new(PyTypeObject *type, PyObject *args, PyObject *kwds) {
+	PyObject *callback;
+	if (!PyArg_ParseTuple(args, "O", &callback)) return NULL;
+	if (!PyCallable_Check(callback)) { PyErr_SetString(PyExc_TypeError, "callback must be a callable"); return NULL; }
 	NSSS *self = (NSSS *) type->tp_alloc(type, 0);
 	if (self) {
+		self->callback = callback;
+		Py_INCREF(callback);
 		self->nsss = [[NSSpeechSynthesizer alloc] initWithVoice:nil];
 		if (self->nsss) {
-
-		} else PyErr_NoMemory();
+			self->nsss.delegate = [[SynthesizerDelegate alloc] initWithNSSS:self];
+		} else return PyErr_NoMemory();
 	}
 	return (PyObject*)self;
 }
 
 static void
 NSSS_dealloc(NSSS *self) {
-	if (self->nsss) [self->nsss release];
+	if (self->nsss) {
+		if (self->nsss.delegate) [self->nsss.delegate release];
+		self->nsss.delegate = nil;
+		[self->nsss release];
+	}
 	self->nsss = nil;
+	Py_CLEAR(self->callback);
 }
 
 static PyObject*
@@ -73,6 +120,20 @@ NSSS_get_all_voices(NSSS *self, PyObject *args) {
 	return ans;
 }
 
+static PyObject*
+NSSS_set_command_delimiters(NSSS *self, PyObject *args) {
+	// this function doesn't actually work
+	// https://openradar.appspot.com/6524554
+	const char *left, *right;
+	if (!PyArg_ParseTuple(args, "ss", &left, &right)) return NULL;
+	NSError *err = nil;
+	[self->nsss setObject:@{NSSpeechCommandPrefix:@(left), NSSpeechCommandSuffix:@(right)} forProperty:NSSpeechCommandDelimiterProperty error:&err];
+	if (err) {
+		PyErr_SetString(PyExc_OSError, [[NSString stringWithFormat:@"Failed to set delimiters: %@", err] UTF8String]);
+		return NULL;
+	}
+	Py_RETURN_NONE;
+}
 
 static PyObject*
 NSSS_get_current_voice(NSSS *self, PyObject *args) {
@@ -144,10 +205,33 @@ NSSS_start_saving_to_path(NSSS *self, PyObject *args) {
 	Py_RETURN_FALSE;
 }
 
+static PyObject*
+NSSS_status(NSSS *self, PyObject *args) {
+	NSError *err = nil;
+	NSDictionary *status = [self->nsss objectForProperty:NSSpeechStatusProperty error:&err];
+	if (err) {
+		PyErr_SetString(PyExc_OSError, [[err localizedDescription] UTF8String]);
+		return NULL;
+	}
+	PyObject *ans = PyDict_New();
+	if (ans) {
+		NSNumber *result = [status objectForKey:NSSpeechStatusOutputBusy];
+		if (result) {
+			if (PyDict_SetItemString(ans, "synthesizing", [result boolValue] ? Py_True : Py_False) != 0) { Py_CLEAR(ans); return NULL; }
+		}
+		result = [status objectForKey:NSSpeechStatusOutputPaused];
+		if (result) {
+			if (PyDict_SetItemString(ans, "paused", [result boolValue] ? Py_True : Py_False) != 0) { Py_CLEAR(ans); return NULL; }
+		}
+	}
+	return ans;
+}
+
 // Boilerplate {{{
 #define M(name, args) { #name, (PyCFunction)NSSS_##name, args, ""}
 static PyMethodDef NSSS_methods[] = {
     M(get_all_voices, METH_NOARGS),
+    M(status, METH_NOARGS),
     M(speak, METH_VARARGS),
     M(start_saving_to_path, METH_VARARGS),
     M(speaking, METH_NOARGS),
@@ -159,6 +243,7 @@ static PyMethodDef NSSS_methods[] = {
     M(set_current_volume, METH_VARARGS),
     M(get_current_rate, METH_NOARGS),
     M(set_current_rate, METH_VARARGS),
+	M(set_command_delimiters, METH_VARARGS),
     {NULL, NULL, 0, NULL}
 };
 #undef M
@@ -180,6 +265,8 @@ nsss_init_module(PyObject *module) {
         Py_DECREF(&NSSSType);
         return -1;
     }
+	PyModule_AddIntMacro(module, MARK);
+	PyModule_AddIntMacro(module, END);
 
 	return 0;
 }

src/calibre/gui2/tts/windows.py

@@ -4,11 +4,20 @@
 
 
 from threading import Thread
+
+from calibre import prepare_string_for_xml
+
 from .common import Event, EventType
 
 
 class Client:
 
+    mark_template = '<bookmark mark="{}"/>'
+
+    @classmethod
+    def escape_marked_text(cls, text):
+        return prepare_string_for_xml(text)
+
     def __init__(self):
         from calibre.utils.windows.winsapi import ISpVoice
         self.sp_voice = ISpVoice()
@@ -33,7 +42,9 @@ class Client:
                 c()
 
     def get_events(self):
-        from calibre_extensions.winsapi import SPEI_TTS_BOOKMARK, SPEI_START_INPUT_STREAM, SPEI_END_INPUT_STREAM
+        from calibre_extensions.winsapi import (
+            SPEI_END_INPUT_STREAM, SPEI_START_INPUT_STREAM, SPEI_TTS_BOOKMARK
+        )
         ans = []
         for (stream_number, event_type, event_data) in self.sp_voice.get_events():
             if stream_number == self.current_stream_number:
@@ -49,11 +60,15 @@ class Client:
         return ans
 
     def speak_simple_text(self, text):
-        from calibre_extensions.winsapi import SPF_ASYNC, SPF_PURGEBEFORESPEAK, SPF_IS_NOT_XML
+        from calibre_extensions.winsapi import (
+            SPF_ASYNC, SPF_IS_NOT_XML, SPF_PURGEBEFORESPEAK
+        )
         self.current_callback = None
         self.current_stream_number = self.sp_voice.speak(text, SPF_ASYNC | SPF_PURGEBEFORESPEAK | SPF_IS_NOT_XML)
 
     def speak_marked_text(self, text, callback):
-        from calibre_extensions.winsapi import SPF_ASYNC, SPF_PURGEBEFORESPEAK, SPF_IS_XML
+        from calibre_extensions.winsapi import (
+            SPF_ASYNC, SPF_IS_XML, SPF_PURGEBEFORESPEAK
+        )
         self.current_callback = callback
         self.current_stream_number = self.sp_voice.speak(text, SPF_ASYNC | SPF_PURGEBEFORESPEAK | SPF_IS_XML, True)