Allow passing speech text with shared memory

src/calibre/utils/windows/common.h

@@ -62,6 +62,8 @@ static inline void co_task_mem_free(void* m) { CoTaskMemFree(m); }
 typedef generic_raii<wchar_t*, co_task_mem_free, static_cast<wchar_t*>(NULL)> com_wchar_raii;
 static inline void handle_destructor(HANDLE p) { CloseHandle(p); }
 typedef generic_raii<HANDLE, handle_destructor, INVALID_HANDLE_VALUE> handle_raii;
+static inline void mapping_destructor(void *p) { UnmapViewOfFile(p); }
+typedef generic_raii<void*, mapping_destructor, static_cast<void*>(NULL)> mapping_raii;
 
 struct prop_variant : PROPVARIANT {
 	prop_variant(VARTYPE vt=VT_EMPTY) noexcept : PROPVARIANT{} { PropVariantInit(this); this->vt = vt;  }

src/calibre/utils/windows/winspeech.cpp

@@ -865,10 +865,25 @@ handle_speak(id_type cmd_id, std::vector<std::wstring_view> &parts) {
         throw std::string("Not a well formed speak command");
     }
     parts.erase(parts.begin(), parts.begin() + 2);
-    auto address = join(parts);
-    if (address.size() == 0) throw std::string("Address missing");
+    std::wstring address;
+    id_type shm_size = 0;
     if (is_shm) {
-        throw std::string("TODO: Implement support for SHM");
+        shm_size = parse_id(parts.at(0));
+        address = parts.at(1);
+        handle_raii handle(OpenFileMappingW(FILE_MAP_READ, false, address.data()));
+        if (handle.ptr() == INVALID_HANDLE_VALUE) {
+            output_error(cmd_id, "Could not open shared memory at", winrt::to_string(address), __LINE__);
+            return;
+        }
+        mapping_raii mapping(MapViewOfFile(handle.ptr(), FILE_MAP_READ, 0, 0, (SIZE_T)shm_size));
+        if (mapping.ptr() == NULL) {
+            output_error(cmd_id, "Could not map shared memory with error", std::to_string(GetLastError()), __LINE__);
+            return;
+        }
+        address = winrt::to_hstring((const char*)mapping.ptr());
+    } else {
+        address = join(parts);
+        if (address.size() == 0) throw std::string("Address missing");
     }
     sx.speak(cmd_id, address, is_ssml);
 }

src/calibre/utils/windows/winspeech.py

@@ -3,11 +3,13 @@
 
 
 import json
+import struct
 import sys
 from contextlib import closing
 from queue import Queue
 from threading import Thread
 
+from calibre.utils.shm import SharedMemory
 from calibre.utils.ipc.simple_worker import start_pipe_worker
 
 SSML_SAMPLE = '''
@@ -30,7 +32,36 @@ def start_worker():
     return start_pipe_worker('from calibre_extensions.winspeech import run_main_loop; raise SystemExit(run_main_loop())')
 
 
-def develop_speech(text=SSML_SAMPLE):
+def max_buffer_size(text) -> int:
+    if isinstance(text, str):
+        text = [text]
+    ans = 0
+    for x in text:
+        if isinstance(x, int):
+            ans += 5
+        else:
+            ans += 4 * len(x)
+    return ans
+
+
+def encode_to_file_object(text, output) -> int:
+    if isinstance(text, str):
+        text = [text]
+    p = struct.pack
+    sz = 0
+    for x in text:
+        if isinstance(x, int):
+            output.write(b'\0')
+            output.write(p('=I', x))
+            sz += 5
+        else:
+            b = x.encode('utf-8')
+            output.write(b)
+            sz += len(b)
+    return sz
+
+
+def develop_speech(text='Lucca Brazzi sleeps with the fishes.'):
     p = start_worker()
     print('\x1b[32mSpeaking', text, '\x1b[39m]]'[:-2], flush=True)
     q = Queue()
@@ -48,18 +79,20 @@ def develop_speech(text=SSML_SAMPLE):
 
     Thread(name='Echo', target=echo_output, args=(p,), daemon=True).start()
     exit_code = 0
-    with closing(p.stdin), closing(p.stdout):
-        text = text.replace('\n', ' ')
+    with closing(p.stdin), closing(p.stdout), SharedMemory(size=max_buffer_size(text)) as shm:
         st = 'ssml' if '<speak' in text else 'text'
+        sz = encode_to_file_object(text, shm)
         try:
             send('1 echo Synthesizer started')
             send('1 volume 0.1')
-            send(f'2 speak {st} inline', text)
+            send(f'2 speak {st} shm {sz} {shm.name}')
             while True:
                 m = q.get()
+                if m['related_to'] != 2:
+                    continue
                 if m['payload_type'] == 'media_state_changed' and m['state'] == 'ended':
                     break
-                if m['payload_type'] == 'error' and m['related_to'] == 2:
+                if m['payload_type'] == 'error':
                     exit_code = 1
                     break
             send(f'3 echo Synthesizer exiting with exit code: {exit_code}')