FS-6535 --resolve mod_rayo: allow multiple grammars (including dtmf) when using unimrcp

src/include/switch_ivr.h

@@ -238,7 +238,7 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_resume_detect_speech(switch_core_sess
   \param name the grammar name
   \return SWITCH_STATUS_SUCCESS if all is well
 */
-SWITCH_DECLARE(switch_status_t) switch_ivr_detect_speech_load_grammar(switch_core_session_t *session, char *grammar, char *name);
+SWITCH_DECLARE(switch_status_t) switch_ivr_detect_speech_load_grammar(switch_core_session_t *session, const char *grammar, const char *name);
 
 /*!
   \brief Unload a grammar on a background speech detection handle

src/mod/event_handlers/mod_rayo/rayo_input_component.c

@@ -413,14 +413,141 @@ static int validate_call_input(iks *input, const char **error)
 	return 1;
 }
 
+static char *setup_grammars_pocketsphinx(struct input_component *component, switch_core_session_t *session, iks *input, const struct xmpp_error **stanza_error, const char **error_detail)
+{
+	const char *jsgf_path;
+	switch_stream_handle_t grammar = { 0 };
+	SWITCH_STANDARD_STREAM(grammar);
+
+	/* transform SRGS grammar to JSGF */
+	if (!(component->grammar = srgs_parse(globals.parser, iks_find_cdata(input, "grammar")))) {
+		*stanza_error = STANZA_ERROR_BAD_REQUEST;
+		*error_detail = "Failed to parse grammar body";
+		return NULL;
+	}
+
+	jsgf_path = srgs_grammar_to_jsgf_file(component->grammar, SWITCH_GLOBAL_dirs.grammar_dir, "gram");
+	if (!jsgf_path) {
+		*stanza_error = STANZA_ERROR_BAD_REQUEST;
+		*error_detail = "Grammar conversion to JSGF error";
+		return NULL;
+	}
+
+	/* build pocketsphinx grammar string */
+	grammar.write_function(&grammar,
+		"{start-input-timers=%s,no-input-timeout=%d,speech-timeout=%d,confidence-threshold=%d}%s",
+		component->start_timers ? "true" : "false",
+		component->initial_timeout,
+		component->max_silence,
+		(int)ceil(component->min_confidence * 100.0),
+		jsgf_path);
+
+	return (char *)grammar.data;
+}
+
+static char *setup_grammars_unimrcp(struct input_component *component, switch_core_session_t *session, iks *input, const struct xmpp_error **stanza_error, const char **error_detail)
+{
+	iks *grammar_tag;
+	switch_stream_handle_t grammar_uri_list = { 0 };
+	SWITCH_STANDARD_STREAM(grammar_uri_list);
+
+	/* unlock handler mutex, otherwise deadlock will happen when switch_ivr_detect_speech_init adds a new media bug */
+	switch_mutex_unlock(component->handler->mutex);
+	if (switch_ivr_detect_speech_init(session, component->recognizer, "", NULL) != SWITCH_STATUS_SUCCESS) {
+		switch_mutex_lock(component->handler->mutex);
+		*stanza_error = STANZA_ERROR_INTERNAL_SERVER_ERROR;
+		*error_detail = "Failed to initialize recognizer";
+		return NULL;
+	}
+	switch_mutex_lock(component->handler->mutex);
+
+	/* load unimrcp grammars and return uri-list */
+	grammar_uri_list.write_function(&grammar_uri_list, "{start-recognize=true,start-input-timers=%s,confidence-threshold=%f,sensitivity-level=%f",
+							component->start_timers ? "true" : "false",
+							component->min_confidence,
+							component->sensitivity);
+	if (component->initial_timeout > 0) {
+		grammar_uri_list.write_function(&grammar_uri_list, ",no-input-timeout=%d",
+			component->initial_timeout);
+	}
+
+	if (component->max_silence > 0) {
+		grammar_uri_list.write_function(&grammar_uri_list, ",speech-complete-timeout=%d,speech-incomplete-timeout=%d",
+			component->max_silence,
+			component->max_silence);
+	}
+
+	if (!zstr(component->language)) {
+		grammar_uri_list.write_function(&grammar_uri_list, ",speech-language=%s", component->language);
+	}
+
+	if (!strcmp(iks_find_attrib_soft(input, "mode"), "any") || !strcmp(iks_find_attrib_soft(input, "mode"), "dtmf")) {
+		/* set dtmf params */
+		if (component->inter_digit_timeout > 0) {
+			grammar_uri_list.write_function(&grammar_uri_list, ",dtmf-interdigit-timeout=%d", component->inter_digit_timeout);
+		}
+		if (component->term_digit) {
+			grammar_uri_list.write_function(&grammar_uri_list, ",dtmf-term-char=%c", component->term_digit);
+		}
+	}
+	grammar_uri_list.write_function(&grammar_uri_list, "}");
+
+	for (grammar_tag = iks_find(input, "grammar"); grammar_tag; grammar_tag = iks_next_tag(grammar_tag)) {
+		const char *grammar_name;
+		iks *grammar_cdata;
+		const char *grammar;
+
+		/* is this a grammar? */
+		if (strcmp("grammar", iks_name(grammar_tag))) {
+			continue;
+		}
+
+		/* get the srgs contained in this grammar */
+		if (!(grammar_cdata = iks_child(grammar_tag)) || iks_type(grammar_cdata) != IKS_CDATA) {
+			*stanza_error = STANZA_ERROR_BAD_REQUEST;
+			*error_detail = "Missing grammar";
+			switch_safe_free(grammar_uri_list.data);
+			return NULL;
+		}
+
+		/* load the grammar */
+		grammar = switch_core_sprintf(RAYO_POOL(component), "{start-recognize=false}inline:%s", iks_cdata(grammar_cdata));
+		grammar_name = switch_core_sprintf(RAYO_POOL(component), "grammar-%d", rayo_actor_seq_next(RAYO_ACTOR(component)));
+		/* unlock handler mutex, otherwise deadlock will happen if switch_ivr_detect_speech_load_grammar removes the media bug */
+		switch_mutex_unlock(component->handler->mutex);
+		if (switch_ivr_detect_speech_load_grammar(session, grammar, grammar_name) != SWITCH_STATUS_SUCCESS) {
+			switch_mutex_lock(component->handler->mutex);
+			*stanza_error = STANZA_ERROR_INTERNAL_SERVER_ERROR;
+			*error_detail = "Failed to load grammar";
+			switch_safe_free(grammar_uri_list.data);
+			return NULL;
+		}
+		switch_mutex_lock(component->handler->mutex);
+
+		/* add grammar to uri-list */
+		grammar_uri_list.write_function(&grammar_uri_list, "session:%s\r\n", grammar_name);
+	}
+
+	return (char *)grammar_uri_list.data;
+}
+
+static char *setup_grammars_unknown(struct input_component *component, switch_core_session_t *session, iks *input, const struct xmpp_error **stanza_error, const char **error_detail)
+{
+	switch_stream_handle_t grammar = { 0 };
+	SWITCH_STANDARD_STREAM(grammar);
+	grammar.write_function(&grammar, "%s", iks_find_cdata(input, "grammar"));
+	return (char *)grammar.data;
+}
+
 /**
  * Start call input on voice resource
  */
-static iks *start_call_voice_input(struct input_component *component, switch_core_session_t *session, iks *input, iks *iq, const char *output_file, int barge_in)
+static iks *start_call_voice_input(struct input_component *component, switch_core_session_t *session, iks *input, iks *iq, int barge_in)
 {
 	struct input_handler *handler = component->handler;
-	switch_stream_handle_t grammar = { 0 };
-	SWITCH_STANDARD_STREAM(grammar);
+	char *grammar = NULL;
+	const struct xmpp_error *stanza_error = NULL;
+	const char *error_detail = NULL;
 
 	if (component->speech_mode && handler->voice_component) {
 		/* don't allow multi voice input */
@@ -441,72 +568,23 @@ static iks *start_call_voice_input(struct input_component *component, switch_cor
 		RAYO_UNLOCK(component);
 		RAYO_DESTROY(component);
 		return iks_new_error_detailed(iq, STANZA_ERROR_BAD_REQUEST, "Must use the same recognizer for the entire call");
+	} else if (zstr(handler->last_recognizer)) {
+		handler->last_recognizer = switch_core_session_strdup(session, component->recognizer);
 	}
-	handler->last_recognizer = switch_core_session_strdup(session, component->recognizer);
 
 	if (!strcmp(component->recognizer, "pocketsphinx")) {
-		const char *jsgf_path;
-
-		/* transform SRGS grammar to JSGF */
-		if (!(component->grammar = srgs_parse(globals.parser, iks_find_cdata(input, "grammar")))) {
-			switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Failed to parse grammar body\n");
-			handler->voice_component = NULL;
-			RAYO_UNLOCK(component);
-			RAYO_DESTROY(component);
-			return iks_new_error_detailed(iq, STANZA_ERROR_BAD_REQUEST, "Failed to parse grammar body");
-		}
-		jsgf_path = srgs_grammar_to_jsgf_file(component->grammar, SWITCH_GLOBAL_dirs.grammar_dir, "gram");
-		if (!jsgf_path) {
-			handler->voice_component = NULL;
-			RAYO_UNLOCK(component);
-			RAYO_DESTROY(component);
-			return iks_new_error_detailed(iq, STANZA_ERROR_INTERNAL_SERVER_ERROR, "Grammar conversion to JSGF error");
-		}
-
-		/* build pocketsphinx grammar string */
-		grammar.write_function(&grammar,
-			"{start-input-timers=%s,no-input-timeout=%d,speech-timeout=%d,confidence-threshold=%d}%s",
-			component->start_timers ? "true" : "false",
-			component->initial_timeout,
-			component->max_silence,
-			(int)ceil(component->min_confidence * 100.0),
-			jsgf_path);
-	} else if (!strncmp(component->recognizer, "unimrcp", strlen("unimrcp"))) {
-		/* send inline grammar to unimrcp */
-		grammar.write_function(&grammar, "{start-input-timers=%s,confidence-threshold=%f,sensitivity-level=%f",
-								component->start_timers ? "true" : "false",
-								component->min_confidence,
-								component->sensitivity);
-
-		if (component->initial_timeout > 0) {
-			grammar.write_function(&grammar, ",no-input-timeout=%d",
-				component->initial_timeout);
-		}
-
-		if (component->max_silence > 0) {
-			grammar.write_function(&grammar, ",speech-complete-timeout=%d,speech-incomplete-timeout=%d",
-				component->max_silence,
-				component->max_silence);
-		}
-
-		if (!zstr(component->language)) {
-			grammar.write_function(&grammar, ",speech-language=%s", component->language);
-		}
-
-		if (!strcmp(iks_find_attrib_soft(input, "mode"), "any")) {
-			/* set dtmf params */
-			if (component->inter_digit_timeout > 0) {
-				grammar.write_function(&grammar, ",dtmf-interdigit-timeout=%d", component->inter_digit_timeout);
-			}
-			if (component->term_digit) {
-				grammar.write_function(&grammar, ",dtmf-term-char=%c", component->term_digit);
-			}
-		}
-
-		grammar.write_function(&grammar, "}inline:%s", iks_find_cdata(input, "grammar"));
+		grammar = setup_grammars_pocketsphinx(component, session, input, &stanza_error, &error_detail);
+	} else if (!strcmp(component->recognizer, "unimrcp")) {
+		grammar = setup_grammars_unimrcp(component, session, input, &stanza_error, &error_detail);
 	} else {
-		/* passthrough to unknown ASR module */
-		grammar.write_function(&grammar, "%s", iks_find_cdata(input, "grammar"));
+		grammar = setup_grammars_unknown(component, session, input, &stanza_error, &error_detail);
+	}
+
+	if (!grammar) {
+		handler->voice_component = NULL;
+		RAYO_UNLOCK(component);
+		RAYO_DESTROY(component);
+		return iks_new_error_detailed(iq, stanza_error, error_detail);
 	}
 
 	/* acknowledge command */
@@ -514,15 +592,15 @@ static iks *start_call_voice_input(struct input_component *component, switch_cor
 
 	/* start speech detection */
 	switch_channel_set_variable(switch_core_session_get_channel(session), "fire_asr_events", "true");
-	switch_mutex_unlock(handler->mutex); /* unlock handler mutex, otherwise deadlock will happen when switch_ivr_detect_speech adds a new media bug */
-	if (switch_ivr_detect_speech(session, component->recognizer, grammar.data, "mod_rayo_grammar", "", NULL) != SWITCH_STATUS_SUCCESS) {
+	/* unlock handler mutex, otherwise deadlock will happen if switch_ivr_detect_speech adds a media bug */
+	switch_mutex_unlock(handler->mutex);
+	if (switch_ivr_detect_speech(session, component->recognizer, grammar, "mod_rayo_grammar", "", NULL) != SWITCH_STATUS_SUCCESS) {
 		switch_mutex_lock(handler->mutex);
 		handler->voice_component = NULL;
 		rayo_component_send_complete(RAYO_COMPONENT(component), COMPONENT_COMPLETE_ERROR);
-	} else {
-		switch_mutex_lock(handler->mutex);
 	}
-	switch_safe_free(grammar.data);
+	switch_mutex_lock(handler->mutex);
+	switch_safe_free(grammar);
 
 	return NULL;
 }
@@ -530,7 +608,7 @@ static iks *start_call_voice_input(struct input_component *component, switch_cor
 /**
  * Start call input on DTMF resource
  */
-static iks *start_call_dtmf_input(struct input_component *component, switch_core_session_t *session, iks *input, iks *iq, const char *output_file, int barge_in)
+static iks *start_call_dtmf_input(struct input_component *component, switch_core_session_t *session, iks *input, iks *iq, int barge_in)
 {
 	/* parse the grammar */
 	if (!(component->grammar = srgs_parse(globals.parser, iks_find_cdata(input, "grammar")))) {
@@ -558,7 +636,7 @@ static iks *start_call_dtmf_input(struct input_component *component, switch_core
  * @param input the input request
  * @param iq the original input/prompt request
  */
-static iks *start_call_input(struct input_component *component, switch_core_session_t *session, iks *input, iks *iq, const char *output_file, int barge_in)
+static iks *start_call_input(struct input_component *component, switch_core_session_t *session, iks *input, iks *iq, int barge_in)
 {
 	iks *result = NULL;
 
@@ -610,9 +688,9 @@ static iks *start_call_input(struct input_component *component, switch_core_sess
 	component->speech_mode = strcmp(iks_find_attrib_soft(input, "mode"), "dtmf");
 
 	if (component->speech_mode) {
-		result = start_call_voice_input(component, session, input, iq, output_file, barge_in);
+		result = start_call_voice_input(component, session, input, iq, barge_in);
 	} else {
-		result = start_call_dtmf_input(component, session, input, iq, output_file, barge_in);
+		result = start_call_dtmf_input(component, session, input, iq, barge_in);
 	}
 
 	switch_mutex_unlock(handler->mutex);
@@ -671,7 +749,7 @@ static iks *start_call_input_component(struct rayo_actor *call, struct rayo_mess
 		switch_core_destroy_memory_pool(&pool);
 		return iks_new_error_detailed(iq, STANZA_ERROR_INTERNAL_SERVER_ERROR, "Failed to create input entity");
 	}
-	return start_call_input(input_component, session, input, iq, NULL, 0);
+	return start_call_input(input_component, session, input, iq, 0);
 }
 
 /**

src/switch_ivr_async.c

@@ -4042,7 +4042,7 @@ SWITCH_DECLARE(switch_status_t) switch_ivr_resume_detect_speech(switch_core_sess
 	return SWITCH_STATUS_FALSE;
 }
 
-SWITCH_DECLARE(switch_status_t) switch_ivr_detect_speech_load_grammar(switch_core_session_t *session, char *grammar, char *name)
+SWITCH_DECLARE(switch_status_t) switch_ivr_detect_speech_load_grammar(switch_core_session_t *session, const char *grammar, const char *name)
 {
 	switch_channel_t *channel = switch_core_session_get_channel(session);
 	struct speech_thread_handle *sth = switch_channel_get_private(channel, SWITCH_SPEECH_KEY);