mod_rayo: add some hacky support for SISR in DTMF recognizer

src/mod/event_handlers/mod_rayo/nlsml.c

@@ -397,9 +397,10 @@ static int isdtmf(const char digit)
 /**
  * Construct an NLSML result for digit match
  * @param digits the matching digits
+ * @param interpretation the optional digit interpretation
  * @return the NLSML <result>
  */
-iks *nlsml_create_dtmf_match(const char *digits)
+iks *nlsml_create_dtmf_match(const char *digits, const char *interpretation)
 {
 	iks *result = iks_new("result");
 	iks_insert_attrib(result, "xmlns", NLSML_NS);
@@ -410,10 +411,11 @@ iks *nlsml_create_dtmf_match(const char *digits)
 		int num_digits = strlen(digits);
 		switch_stream_handle_t stream = { 0 };
 
-		iks *interpretation = iks_insert(result, "interpretation");
-		iks *input = iks_insert(interpretation, "input");
-		iks_insert_attrib(input, "mode", "dtmf");
-		iks_insert_attrib(input, "confidence", "100");
+		iks *interpretation_node = iks_insert(result, "interpretation");
+		iks *input_node = iks_insert(interpretation_node, "input");
+		iks *instance_node = iks_insert(interpretation_node, "instance");
+		iks_insert_attrib(input_node, "mode", "dtmf");
+		iks_insert_attrib(input_node, "confidence", "100");
 
 		SWITCH_STANDARD_STREAM(stream);
 		for (i = 0; i < num_digits; i++) {
@@ -426,7 +428,13 @@ iks *nlsml_create_dtmf_match(const char *digits)
 				}
 			}
 		}
-		iks_insert_cdata(input, stream.data, strlen(stream.data));
+		iks_insert_cdata(input_node, stream.data, strlen(stream.data));
+
+		if (zstr(interpretation)) {
+			iks_insert_cdata(instance_node, stream.data, strlen(stream.data));
+		} else {
+			iks_insert_cdata(instance_node, interpretation, strlen(interpretation));
+		}
 		switch_safe_free(stream.data);
 	}
 	return result;

src/mod/event_handlers/mod_rayo/nlsml.h

@@ -42,7 +42,7 @@ enum nlsml_match_type {
 extern int nlsml_init(void);
 enum nlsml_match_type nlsml_parse(const char *result, const char *uuid);
 iks *nlsml_normalize(const char *result);
-extern iks *nlsml_create_dtmf_match(const char *digits);
+extern iks *nlsml_create_dtmf_match(const char *digits, const char *interpretation);
 
 #endif
 

src/mod/event_handlers/mod_rayo/rayo_input_component.c

@@ -157,6 +157,7 @@ static switch_status_t input_component_on_dtmf(switch_core_session_t *session, c
 		int is_term_digit = 0;
 		struct input_component *component;
 		enum srgs_match_type match;
+		const char *interpretation = NULL;
 
 		switch_mutex_lock(handler->mutex);
 
@@ -179,7 +180,7 @@ static switch_status_t input_component_on_dtmf(switch_core_session_t *session, c
 			switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "Collected term digit = \"%c\"\n", dtmf->digit);
 		}
 
-		match = srgs_grammar_match(component->grammar, component->digits);
+		match = srgs_grammar_match(component->grammar, component->digits, &interpretation);
 
 		/* adjust result if terminating digit was pressed */
 		if (is_term_digit) {
@@ -208,7 +209,7 @@ static switch_status_t input_component_on_dtmf(switch_core_session_t *session, c
 				break;
 			}
 			case SMT_MATCH_END: {
-				iks *result = nlsml_create_dtmf_match(component->digits);
+				iks *result = nlsml_create_dtmf_match(component->digits, interpretation);
 				/* notify of match and remove input component */
 				handler->dtmf_component = NULL;
 				switch_core_media_bug_remove(session, &handler->bug);
@@ -248,13 +249,14 @@ static switch_bool_t input_component_bug_callback(switch_media_bug_t *bug, void
 				int elapsed_ms = (switch_micro_time_now() - component->last_digit_time) / 1000;
 				if (component->num_digits && component->inter_digit_timeout > 0 && elapsed_ms > component->inter_digit_timeout) {
 					enum srgs_match_type match;
+					const char *interpretation = NULL;
 					handler->dtmf_component = NULL;
 					switch_core_media_bug_set_flag(bug, SMBF_PRUNE);
 
 					/* we got some input, check for match */
-					match = srgs_grammar_match(component->grammar, component->digits);
+					match = srgs_grammar_match(component->grammar, component->digits, &interpretation);
 					if (match == SMT_MATCH || match == SMT_MATCH_END) {
-						iks *result = nlsml_create_dtmf_match(component->digits);
+						iks *result = nlsml_create_dtmf_match(component->digits, interpretation);
 						/* notify of match */
 						switch_log_printf(SWITCH_CHANNEL_SESSION_LOG(session), SWITCH_LOG_DEBUG, "MATCH = %s\n", component->digits);
 						send_match_event(RAYO_COMPONENT(component), result);

src/mod/event_handlers/mod_rayo/srgs.c

@@ -33,6 +33,7 @@
 #include "srgs.h"
 
 #define MAX_RECURSION 100
+#define MAX_TAGS 30
 
 /** function to handle tag attributes */
 typedef int (* tag_attribs_fn)(struct srgs_grammar *, char **);
@@ -111,6 +112,7 @@ struct item_value {
 	int repeat_min;
 	int repeat_max;
 	const char *weight;
+	char *tag;
 };
 
 /**
@@ -161,6 +163,10 @@ struct srgs_grammar {
 	struct srgs_node *cur;
 	/** rule names mapped to node */
 	switch_hash_t *rules;
+	/** possible matching tags */
+	const char *tags[MAX_TAGS];
+	/** number of tags */
+	int tag_count;
 	/** grammar encoding */
 	char *encoding;
 	/** grammar language */
@@ -708,6 +714,30 @@ static int tag_hook(void *user_data, char *name, char **atts, int type)
 	return result;
 }
 
+/**
+ * Process <tag> CDATA
+ * @param grammar the grammar
+ * @param data the CDATA
+ * @param len the CDATA length
+ * @return IKS_OK
+ */
+static int process_cdata_tag(struct srgs_grammar *grammar, char *data, size_t len)
+{
+	struct srgs_node *item = grammar->cur->parent;
+	if (item && item->type == SNT_ITEM) {
+		item->value.item.tag = switch_core_alloc(grammar->pool, sizeof(char) * (len + 1));
+		item->value.item.tag[len] = '\0';
+		strncpy(item->value.item.tag, data, len);
+		if (grammar->tag_count < MAX_TAGS) {
+			grammar->tags[grammar->tag_count++] = item->value.item.tag;
+		} else {
+			switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_WARNING, "too many <tag>s\n");
+			return IKS_BADXML;
+		}
+	}
+	return IKS_OK;
+}
+
 /**
  * Process CDATA grammar tokens
  * @param grammar the grammar
@@ -956,8 +986,12 @@ static int create_regexes(struct srgs_grammar *grammar, struct srgs_node *node,
 		case SNT_ITEM:
 			if (node->child) {
 				struct srgs_node *item = node->child;
-				if (node->value.item.repeat_min != 1 || node->value.item.repeat_max != 1) {
-					stream->write_function(stream, "%s", "(?:");
+				if (node->value.item.repeat_min != 1 || node->value.item.repeat_max != 1 || !zstr(node->value.item.tag)) {
+					if (zstr(node->value.item.tag)) {
+						stream->write_function(stream, "%s", "(?:");
+					} else {
+						stream->write_function(stream, "(?P<%s>", node->value.item.tag);
+					}
 				}
 				for(; item; item = item->next) {
 					if (!create_regexes(grammar, item, stream)) {
@@ -980,6 +1014,8 @@ static int create_regexes(struct srgs_grammar *grammar, struct srgs_node *node,
 					} else {
 						stream->write_function(stream, "){%i}", node->value.item.repeat_min);
 					}
+				} else if (!zstr(node->value.item.tag)) {
+					stream->write_function(stream, "%s", ")");
 				}
 			}
 			break;
@@ -1171,7 +1207,7 @@ struct srgs_grammar *srgs_parse(struct srgs_parser *parser, const char *document
 }
 
 #define MAX_INPUT_SIZE 128
-#define OVECTOR_SIZE 30
+#define OVECTOR_SIZE MAX_TAGS
 #define WORKSPACE_SIZE 1024
 
 /**
@@ -1214,15 +1250,17 @@ static int is_match_end(pcre *compiled_regex, const char *input)
  * Find a match
  * @param grammar the grammar to match
  * @param input the input to compare
+ * @param interpretation the (optional) interpretation of the input result
  * @return the match result
  */
-enum srgs_match_type srgs_grammar_match(struct srgs_grammar *grammar, const char *input)
+enum srgs_match_type srgs_grammar_match(struct srgs_grammar *grammar, const char *input, const char **interpretation)
 {
 	int result = 0;
 	int ovector[OVECTOR_SIZE];
-	int workspace[WORKSPACE_SIZE];
 	pcre *compiled_regex;
 
+	*interpretation = NULL;
+
 	if (zstr(input)) {
 		return SMT_NO_MATCH;
 	}
@@ -1234,12 +1272,24 @@ enum srgs_match_type srgs_grammar_match(struct srgs_grammar *grammar, const char
 	if (!(compiled_regex = get_compiled_regex(grammar))) {
 		return SMT_NO_MATCH;
 	}
-	result = pcre_dfa_exec(compiled_regex, NULL, input, strlen(input), 0, PCRE_PARTIAL,
-		ovector, sizeof(ovector) / sizeof(ovector[0]),
-		workspace, sizeof(workspace) / sizeof(workspace[0]));
+	result = pcre_exec(compiled_regex, NULL, input, strlen(input), 0, PCRE_PARTIAL,
+		ovector, OVECTOR_SIZE);
 
 	switch_log_printf(SWITCH_CHANNEL_LOG, SWITCH_LOG_DEBUG, "match = %i\n", result);
 	if (result > 0) {
+		int i;
+		char buffer[MAX_INPUT_SIZE + 1];
+		buffer[MAX_INPUT_SIZE] = '\0';
+
+		/* find matching instance... */
+		for (i = 0; i < grammar->tag_count; i++) {
+			buffer[0] = '\0';
+			if (pcre_copy_named_substring(compiled_regex, input, ovector, result, grammar->tags[i], buffer, MAX_INPUT_SIZE) != PCRE_ERROR_NOSUBSTRING && !zstr_buf(buffer)) {
+				*interpretation = grammar->tags[i];
+				break;
+			}
+		}
+
 		if (is_match_end(compiled_regex, input)) {
 			return SMT_MATCH_END;
 		}
@@ -1562,7 +1612,7 @@ int srgs_init(void)
 	add_root_tag_def("grammar", process_grammar, process_cdata_bad, "meta,metadata,lexicon,tag,rule");
 	add_tag_def("ruleref", process_ruleref, process_cdata_bad, "");
 	add_tag_def("token", process_attribs_ignore, process_cdata_ignore, "");
-	add_tag_def("tag", process_attribs_ignore, process_cdata_ignore, "");
+	add_tag_def("tag", process_attribs_ignore, process_cdata_tag, "");
 	add_tag_def("one-of", process_attribs_ignore, process_cdata_tokens, "item");
 	add_tag_def("item", process_item, process_cdata_tokens, "token,ruleref,item,one-of,tag");
 	add_tag_def("rule", process_rule, process_cdata_tokens, "token,ruleref,item,one-of,tag,example");

src/mod/event_handlers/mod_rayo/srgs.h

@@ -51,7 +51,7 @@ extern struct srgs_grammar *srgs_parse(struct srgs_parser *parser, const char *d
 extern const char *srgs_grammar_to_regex(struct srgs_grammar *grammar);
 extern const char *srgs_grammar_to_jsgf(struct srgs_grammar *grammar);
 extern const char *srgs_grammar_to_jsgf_file(struct srgs_grammar *grammar, const char *basedir, const char *ext);
-extern enum srgs_match_type srgs_grammar_match(struct srgs_grammar *grammar, const char *input);
+extern enum srgs_match_type srgs_grammar_match(struct srgs_grammar *grammar, const char *input, const char **interpretation);
 extern void srgs_parser_destroy(struct srgs_parser *parser);
 
 #endif

src/mod/event_handlers/mod_rayo/test_nlsml/main.c

@@ -245,14 +245,15 @@ static const char *nlsml_dtmf_result =
 	"<result xmlns='http://www.ietf.org/xml/ns/mrcpv2' "
 	"xmlns:xf='http://www.w3.org/2000/xforms'><interpretation>"
 	"<input mode='dtmf' confidence='100'>1 2 3 4</input>"
+	"<instance>1 2 3 4</instance>"
 	"</interpretation></result>";
 
 /**
- * Test parsing NLSML example results
+ * Test creating DTMF match result
  */
 static void test_create_dtmf_match(void)
 {
-	iks *result = nlsml_create_dtmf_match("1234");
+	iks *result = nlsml_create_dtmf_match("1234", NULL);
 	char *result_str;
 	ASSERT_NOT_NULL(result);
 	result_str = iks_string(NULL, result);
@@ -260,6 +261,26 @@ static void test_create_dtmf_match(void)
 	iks_free(result_str);
 }
 
+static const char *nlsml_dtmf_instance_result =
+	"<result xmlns='http://www.ietf.org/xml/ns/mrcpv2' "
+	"xmlns:xf='http://www.w3.org/2000/xforms'><interpretation>"
+	"<input mode='dtmf' confidence='100'>1</input>"
+	"<instance>foo</instance>"
+	"</interpretation></result>";
+
+/**
+ * Test creating DTMF match result with instance interpretation
+ */
+static void test_create_dtmf_instance(void)
+{
+	iks *result = nlsml_create_dtmf_match("1", "foo");
+	char *result_str;
+	ASSERT_NOT_NULL(result);
+	result_str = iks_string(NULL, result);
+	ASSERT_STRING_EQUALS(nlsml_dtmf_instance_result, result_str);
+	iks_free(result_str);
+}
+
 static const char *nlsml_good_normalized =
 	"<result x-model='http://theYesNoModel'"
 	" xmlns:xf='http://www.w3.org/2000/xforms'"
@@ -295,6 +316,7 @@ int main(int argc, char **argv)
 	nlsml_init();
 	TEST(test_parse_nlsml_examples);
 	TEST(test_create_dtmf_match);
+	TEST(test_create_dtmf_instance);
 	TEST(test_normalize);
 	return 0;
 }