add transparent speech handle caching via cache_speech_handles=true variable and an app to clear a cached handle called clear_speech_cache

git-svn-id: http://svn.freeswitch.org/svn/freeswitch/trunk@5876 d0543943-73ff-0310-b7d9-9358b9ac24b2
This commit is contained in:
Anthony Minessale 2007-10-15 16:25:08 +00:00
parent be921bed1a
commit b8457503ea
5 changed files with 145 additions and 42 deletions

View File

@ -313,7 +313,7 @@ SWITCH_DECLARE(switch_status_t) switch_play_and_get_digits(switch_core_session_t
SWITCH_DECLARE(switch_status_t) switch_ivr_speak_text_handle(switch_core_session_t *session,
switch_speech_handle_t *sh,
switch_codec_t *codec, switch_timer_t *timer, char *text, switch_input_args_t *args);
SWITCH_DECLARE(void) switch_ivr_clear_speech_cache(switch_core_session_t *session);
/*!
\brief Speak given text with given tts engine
\param session the session to speak on

View File

@ -97,6 +97,8 @@ SWITCH_BEGIN_EXTERN_C
#define SWITCH_PATH_SEPARATOR "/"
#endif
#define SWITCH_URL_SEPARATOR "://"
#define SWITCH_CACHE_SPEECH_HANDLES_VARIABLE "cache_speech_handles"
#define SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME "__cache_speech_handles_obj__"
#define SWITCH_BYPASS_MEDIA_VARIABLE "bypass_media"
#define SWITCH_ENDPOINT_DISPOSITION_VARIABLE "endpoint_disposition"
#define SWITCH_HOLD_MUSIC_VARIABLE "hold_music"

View File

@ -917,6 +917,10 @@ static switch_status_t on_dtmf(switch_core_session_t *session, void *input, swit
return SWITCH_STATUS_SUCCESS;
}
SWITCH_STANDARD_APP(clear_speech_cache_function)
{
switch_ivr_clear_speech_cache(session);
}
SWITCH_STANDARD_APP(speak_function)
{
@ -939,10 +943,25 @@ SWITCH_STANDARD_APP(speak_function)
mydata = switch_core_session_strdup(session, data);
argc = switch_separate_string(mydata, '|', argv, sizeof(argv) / sizeof(argv[0]));
if (argc == 1) {
text = argv[0];
} else if (argc == 2) {
voice = argv[0];
text = argv[1];
} else {
engine = argv[0];
voice = argv[1];
text = argv[2];
}
engine = argv[0];
voice = argv[1];
text = argv[2];
if (!engine) {
engine = switch_channel_get_variable(channel, "tts_engine");
}
if (!voice) {
voice = switch_channel_get_variable(channel, "tts_voice");
}
if (!(engine && voice && text)) {
if (!engine) {
@ -1322,6 +1341,7 @@ SWITCH_MODULE_LOAD_FUNCTION(mod_dptools_load)
SWITCH_ADD_APP(app_interface, "stop_displace_session", "Stop Displace File", "Stop Displacing to a file", stop_displace_session_function, "<path>", SAF_NONE);
SWITCH_ADD_APP(app_interface, "displace_session", "Displace File", DISPLACE_DESC, displace_session_function, "<path> [+time_limit_ms] [mux]", SAF_NONE);
SWITCH_ADD_APP(app_interface, "speak", "Speak text", SPEAK_DESC, speak_function, "<engine>|<voice>|<text>", SAF_NONE);
SWITCH_ADD_APP(app_interface, "clear_speech_cache", "Clear Speech Handle Cache", "Clear Speech Handle Cache", clear_speech_cache_function, "", SAF_NONE);
SWITCH_ADD_APP(app_interface, "bridge", "Bridge Audio", "Bridge the audio between two sessions", audio_bridge_function, "<channel_url>", SAF_SUPPORT_NOMEDIA);
SWITCH_ADD_APP(app_interface, "system", "Execute a system command", "Execute a system command", system_session_function, "<command>", SAF_SUPPORT_NOMEDIA);

View File

@ -659,7 +659,7 @@ SWITCH_DECLARE(void) switch_core_session_perform_destroy(switch_core_session_t *
switch_core_media_bug_remove_all(*session);
switch_buffer_destroy(&(*session)->raw_read_buffer);
switch_buffer_destroy(&(*session)->raw_write_buffer);
switch_ivr_clear_speech_cache(*session);
switch_channel_uninit((*session)->channel);
pool = (*session)->pool;

View File

@ -1367,6 +1367,36 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_speak_text_handle(switch_core_session
}
struct cached_speech_handle {
char tts_name[80];
char voice_name[80];
switch_speech_handle_t sh;
switch_codec_t codec;
switch_timer_t timer;
};
typedef struct cached_speech_handle cached_speech_handle_t;
SWITCH_DECLARE(void) switch_ivr_clear_speech_cache(switch_core_session_t *session)
{
cached_speech_handle_t *cache_obj = NULL;
switch_channel_t *channel;
channel = switch_core_session_get_channel(session);
assert(channel != NULL);
if ((cache_obj = switch_channel_get_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME))) {
switch_speech_flag_t flags = SWITCH_SPEECH_FLAG_NONE;
if (cache_obj->timer.interval) {
switch_core_timer_destroy(&cache_obj->timer);
}
switch_core_speech_close(&cache_obj->sh, &flags);
switch_core_codec_destroy(&cache_obj->codec);
switch_channel_set_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME, NULL);
}
}
SWITCH_DECLARE(switch_status_t) switch_ivr_speak_text(switch_core_session_t *session,
char *tts_name, char *voice_name, char *text, switch_input_args_t *args)
{
@ -1374,21 +1404,50 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_speak_text(switch_core_session_t *ses
uint32_t rate = 0;
int interval = 0;
switch_frame_t write_frame = { 0 };
switch_timer_t timer;
switch_timer_t ltimer, *timer;
switch_core_thread_session_t thread_session;
switch_codec_t codec;
switch_codec_t lcodec, *codec;
switch_memory_pool_t *pool = switch_core_session_get_pool(session);
char *codec_name;
int stream_id = 0;
switch_status_t status = SWITCH_STATUS_SUCCESS;
switch_speech_handle_t sh;
switch_speech_handle_t lsh, *sh;
switch_speech_flag_t flags = SWITCH_SPEECH_FLAG_NONE;
switch_codec_t *read_codec;
char *timer_name;
char *timer_name, *var;
cached_speech_handle_t *cache_obj = NULL;
int need_create = 1, need_alloc = 1;
channel = switch_core_session_get_channel(session);
assert(channel != NULL);
sh = &lsh;
codec = &lcodec;
timer = &ltimer;
if ((var = switch_channel_get_variable(channel, SWITCH_CACHE_SPEECH_HANDLES_VARIABLE)) && switch_true(var)) {
if ((cache_obj = switch_channel_get_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME))) {
need_create = 0;
if (!strcasecmp(cache_obj->tts_name, tts_name)) {
need_alloc = 0;
} else {
switch_ivr_clear_speech_cache(session);
}
}
if (!cache_obj) {
cache_obj = switch_core_session_alloc(session, sizeof(*cache_obj));
}
if (need_alloc) {
switch_copy_string(cache_obj->tts_name, tts_name, sizeof(cache_obj->tts_name));
switch_copy_string(cache_obj->voice_name, voice_name, sizeof(cache_obj->voice_name));
switch_channel_set_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME, cache_obj);
}
sh = &cache_obj->sh;
codec = &cache_obj->codec;
timer = &cache_obj->timer;
}
timer_name = switch_channel_get_variable(channel, "timer_name");
switch_core_session_reset(session);
@ -1396,13 +1455,21 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_speak_text(switch_core_session_t *ses
rate = read_codec->implementation->samples_per_second;
interval = read_codec->implementation->microseconds_per_frame / 1000;
memset(&sh, 0, sizeof(sh));
if (switch_core_speech_open(&sh, tts_name, voice_name, (uint32_t) rate, interval,
&flags, switch_core_session_get_pool(session)) != SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Invalid TTS module!\n");
switch_core_session_reset(session);
return SWITCH_STATUS_FALSE;
if (need_create) {
memset(sh, 0, sizeof(*sh));
if (switch_core_speech_open(sh, tts_name, voice_name, (uint32_t) rate, interval,
&flags, switch_core_session_get_pool(session)) != SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Invalid TTS module!\n");
switch_core_session_reset(session);
if (cache_obj) {
switch_channel_set_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME, NULL);
}
return SWITCH_STATUS_FALSE;
}
} else if (cache_obj && strcasecmp(cache_obj->voice_name, voice_name)) {
switch_copy_string(cache_obj->voice_name, voice_name, sizeof(cache_obj->voice_name));
switch_core_speech_text_param_tts(sh, "voice", voice_name);
}
switch_channel_answer(channel);
@ -1411,46 +1478,60 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_speak_text(switch_core_session_t *ses
codec_name = "L16";
if (switch_core_codec_init(&codec,
codec_name,
NULL, (int) rate, interval, 1, SWITCH_CODEC_FLAG_ENCODE | SWITCH_CODEC_FLAG_DECODE, NULL, pool) == SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Raw Codec Activated\n");
write_frame.codec = &codec;
} else {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Raw Codec Activation Failed %s@%uhz 1 channel %dms\n", codec_name, rate, interval);
flags = 0;
switch_core_speech_close(&sh, &flags);
switch_core_session_reset(session);
return SWITCH_STATUS_GENERR;
}
if (timer_name) {
if (switch_core_timer_init(&timer, timer_name, interval, (int) sh.samples, pool) != SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "setup timer failed!\n");
switch_core_codec_destroy(write_frame.codec);
if (need_create) {
if (switch_core_codec_init(codec,
codec_name,
NULL, (int) rate, interval, 1, SWITCH_CODEC_FLAG_ENCODE | SWITCH_CODEC_FLAG_DECODE, NULL, pool) == SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Raw Codec Activated\n");
} else {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "Raw Codec Activation Failed %s@%uhz 1 channel %dms\n", codec_name, rate, interval);
flags = 0;
switch_core_speech_close(&sh, &flags);
switch_core_speech_close(sh, &flags);
switch_core_session_reset(session);
if (cache_obj) {
switch_channel_set_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME, NULL);
}
return SWITCH_STATUS_GENERR;
}
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "setup timer success %u bytes per %d ms!\n", sh.samples * 2, interval);
}
write_frame.codec = codec;
if (timer_name) {
if (need_create) {
if (switch_core_timer_init(timer, timer_name, interval, (int) sh->samples, pool) != SWITCH_STATUS_SUCCESS) {
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "setup timer failed!\n");
switch_core_codec_destroy(write_frame.codec);
flags = 0;
switch_core_speech_close(sh, &flags);
switch_core_session_reset(session);
if (cache_obj) {
switch_channel_set_private(channel, SWITCH_CACHE_SPEECH_HANDLES_OBJ_NAME, NULL);
}
return SWITCH_STATUS_GENERR;
}
switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "setup timer success %u bytes per %d ms!\n", sh->samples * 2, interval);
}
/* start a thread to absorb incoming audio */
for (stream_id = 0; stream_id < switch_core_session_get_stream_count(session); stream_id++) {
switch_core_service_session(session, &thread_session, stream_id);
}
}
status = switch_ivr_speak_text_handle(session, &sh, write_frame.codec, timer_name ? &timer : NULL, text, args);
status = switch_ivr_speak_text_handle(session, sh, write_frame.codec, timer_name ? timer : NULL, text, args);
flags = 0;
switch_core_speech_close(&sh, &flags);
switch_core_codec_destroy(&codec);
if (!cache_obj) {
switch_core_speech_close(sh, &flags);
switch_core_codec_destroy(codec);
}
if (timer_name) {
/* End the audio absorbing thread */
switch_core_thread_session_end(&thread_session);
switch_core_timer_destroy(&timer);
if (!cache_obj) {
switch_core_timer_destroy(timer);
}
}
switch_core_session_reset(session);