FS-3692 --resolve

src/include/switch_ivr.h

@@ -155,6 +155,21 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_collect_digits_count(switch_core_sess
 																const char *terminators, char *terminator,
 																uint32_t first_timeout, uint32_t digit_timeout, uint32_t abs_timeout);
 
+/*!
+  \brief play a file to the session while doing speech recognition.
+  \param session the session to play and detect on
+  \param file the path to the file
+  \param mod_name the module name of the ASR library
+  \param grammar the grammar text, URI, or local file name
+  \param result of speech recognition, allocated from the session pool
+  \return SWITCH_STATUS_SUCCESS if all is well
+*/
+SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_session_t *session, 
+														const char *file,
+														const char *mod_name,
+														const char *grammar,
+														char **result);
+
 /*!
   \brief Engage background Speech detection on a session
   \param session the session to attach

src/mod/applications/mod_dptools/mod_dptools.c

@@ -30,6 +30,7 @@
  * Bret McDanel <trixter AT 0xdecafbad dot com>
  * Luke Dashjr <luke@openmethods.com> (OpenMethods, LLC)
  * Cesar Cepeda <cesar@auronix.com>
+ * Chris Rienzo <chris@rienzo.net>
  *
  * mod_dptools.c -- Raw Audio File Streaming Application Module
  *
@@ -437,6 +438,51 @@ SWITCH_STANDARD_APP(detect_speech_function)
 	}
 }
 
+#define PLAY_AND_DETECT_SPEECH_SYNTAX "<file> detect:<engine> {param1=val1,param2=val2}<grammar>"
+SWITCH_STANDARD_APP(play_and_detect_speech_function)
+{
+	switch_channel_t *channel = switch_core_session_get_channel(session);
+	char *argv[2];
+	char *lbuf = NULL;
+	const char *response = "DONE";
+	char *detect = NULL;
+
+	switch_channel_set_variable(channel, "detect_speech_result", "");
+
+	if (zstr(data) || !(lbuf = switch_core_session_strdup(session, data)) || !(detect = strstr(lbuf, "detect:"))) {
+		/* bad input */
+		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Usage: %s\n", PLAY_AND_DETECT_SPEECH_SYNTAX);
+		response = "USAGE ERROR";
+		goto done;
+	}
+
+	/* split input at "detect:" */
+	detect[0] = '\0';
+	detect += 7;
+	if (zstr(detect)) {
+		/* bad input */
+		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Usage: %s\n", PLAY_AND_DETECT_SPEECH_SYNTAX);
+		response = "USAGE ERROR";
+		goto done;
+	}
+
+	/* need to have at 2 parameters for detect */
+	if (switch_separate_string(detect, ' ', argv, (sizeof(argv) / sizeof(argv[0]))) == 2) {
+		char *file = lbuf;
+		char *engine = argv[0];
+		char *grammar = argv[1];
+		char *result = NULL;
+		switch_ivr_play_and_detect_speech(session, file, engine, grammar, &result);
+		switch_channel_set_variable(channel, "detect_speech_result", result);
+	} else {
+		/* bad input */
+		switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_ERROR, "Usage: %s\n", PLAY_AND_DETECT_SPEECH_SYNTAX);
+		response = "USAGE ERROR";
+	}
+
+done:
+	switch_channel_set_variable(channel, SWITCH_CURRENT_APPLICATION_RESPONSE_VARIABLE, response);
+}
 
 #define SCHED_HEARTBEAT_SYNTAX "[0|<seconds>]"
 SWITCH_STANDARD_APP(sched_heartbeat_function)
@@ -4000,6 +4046,7 @@ SWITCH_MODULE_LOAD_FUNCTION(mod_dptools_load)
 	SWITCH_ADD_APP(app_interface, "remove_bugs", "Remove media bugs", "Remove all media bugs from a channel.", remove_bugs_function, "", SAF_NONE);
 	SWITCH_ADD_APP(app_interface, "break", "Break", "Set the break flag.", break_function, "", SAF_SUPPORT_NOMEDIA);
 	SWITCH_ADD_APP(app_interface, "detect_speech", "Detect speech", "Detect speech on a channel.", detect_speech_function, DETECT_SPEECH_SYNTAX, SAF_NONE);
+	SWITCH_ADD_APP(app_interface, "play_and_detect_speech", "Play and do speech recognition", "Play and do speech recognition", play_and_detect_speech_function, PLAY_AND_DETECT_SPEECH_SYNTAX, SAF_NONE);
 	SWITCH_ADD_APP(app_interface, "ivr", "Run an ivr menu", "Run an ivr menu.", ivr_application_function, "<menu_name>", SAF_NONE);
 	SWITCH_ADD_APP(app_interface, "redirect", "Send session redirect", "Send a redirect message to a session.", redirect_function, "<redirect_data>",
 				   SAF_SUPPORT_NOMEDIA);

src/switch_ivr_async.c

@@ -27,6 +27,7 @@
  * Michael Jerris <mike@jerris.com>
  * Bret McDanel <bret AT 0xdecafbad dot com>
  * Luke Dashjr <luke@openmethods.com> (OpenMethods, LLC)
+ * Chris Rienzo <chris@rienzo.net>
  *
  * switch_ivr_async.c -- IVR Library (async operations)
  *
@@ -3143,7 +3144,94 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_bind_dtmf_meta_session(switch_core_se
 	return SWITCH_STATUS_SUCCESS;
 }
 
+typedef struct {
+	int done;
+	char *result;
+} play_and_detect_speech_state_t;
 
+static switch_status_t play_and_detect_input_callback(switch_core_session_t *session, void *input, switch_input_type_t input_type, void *data, unsigned int len)
+{
+	play_and_detect_speech_state_t *state = (play_and_detect_speech_state_t *)data;
+	switch_event_t *event;
+	switch_channel_t *channel = switch_core_session_get_channel(session);
+	if (input_type == SWITCH_INPUT_TYPE_EVENT) {
+		event = (switch_event_t *)input;
+		if (event->event_id == SWITCH_EVENT_DETECTED_SPEECH && !state->done) {
+			const char *speech_type = switch_event_get_header(event, "Speech-Type");
+			if (!zstr(speech_type)) {
+				if (!strcasecmp(speech_type, "detected-speech")) {
+					const char *result;
+					switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) DETECTED SPEECH\n", switch_channel_get_name(channel));
+					result = switch_event_get_body(event);
+					if (!zstr(result)) {
+						state->result = switch_core_session_strdup(session, result);
+					} else {
+						state->result = "";
+					}
+					state->done = 1;
+					return SWITCH_STATUS_BREAK;
+				} else if (!strcasecmp(speech_type, "begin-speaking")) {
+					switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) START OF SPEECH\n", switch_channel_get_name(channel));
+					return SWITCH_STATUS_BREAK;
+				}
+			}
+		}
+	}
+	return SWITCH_STATUS_SUCCESS;
+}
+
+SWITCH_DECLARE(switch_status_t) switch_ivr_play_and_detect_speech(switch_core_session_t *session, const char *file, const char *mod_name, const char *grammar, char **result)
+{
+	switch_status_t status;
+	int recognizing = 0;
+	switch_input_args_t args = { 0 };
+	play_and_detect_speech_state_t state = { 0, "" };
+	switch_channel_t *channel = switch_core_session_get_channel(session);
+
+	if (result == NULL) {
+		goto done;
+	}
+
+	/* start speech detection */
+	if (switch_ivr_detect_speech(session, mod_name, grammar, grammar, NULL, NULL) != SWITCH_STATUS_SUCCESS) {
+		goto done;
+	}
+	recognizing = 1;
+
+	/* play the prompt, looking for detection result */
+	args.input_callback = play_and_detect_input_callback;
+	args.buf = &state;
+	args.buflen = sizeof(state);
+	status = switch_ivr_play_file(session, NULL, file, &args);
+	if (status != SWITCH_STATUS_BREAK && status != SWITCH_STATUS_SUCCESS) {
+		goto done;
+	}
+
+	/* wait for result if not done */
+	if (!state.done) {
+		switch_ivr_detect_speech_start_input_timers(session);
+		switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_INFO, "(%s) WAITING FOR RESULT\n", switch_channel_get_name(channel));
+		while (!state.done && switch_channel_ready(channel)) {
+			status = switch_ivr_sleep(session, 5000, SWITCH_FALSE, &args);
+			if (status != SWITCH_STATUS_BREAK && status != SWITCH_STATUS_SUCCESS) {
+				goto done;
+			}
+		}
+	}
+	recognizing = !state.done;
+
+done:
+	if (recognizing) {
+		switch_ivr_pause_detect_speech(session);
+	}
+
+	*result = state.result;
+
+	if (!state.done) {
+		return SWITCH_STATUS_FALSE;
+	}
+	return SWITCH_STATUS_SUCCESS;
+}
 
 struct speech_thread_handle {
 	switch_core_session_t *session;