tokenizer.c: Support question mark tokens.

src/libespeak-ng/tokenizer.c

@@ -93,9 +93,10 @@ typedef enum {
 	ESPEAKNG_CTYPE_LOWERCASE,
 	ESPEAKNG_CTYPE_UPPERCASE,
 	ESPEAKNG_CTYPE_FULL_STOP,
+	ESPEAKNG_CTYPE_QUESTION_MARK,
 } espeakng_CTYPE;
 
-#define ESPEAKNG_CTYPE_PROPERTY_MASK 0x800000000000C001ull
+#define ESPEAKNG_CTYPE_PROPERTY_MASK 0xC00000000000C001ull
 
 // Reference: http://www.unicode.org/reports/tr14/tr14-32.html -- Unicode Line Breaking Algorithm
 static espeakng_CTYPE codepoint_type(uint32_t c)
@@ -136,6 +137,8 @@ static espeakng_CTYPE codepoint_type(uint32_t c)
 		return ESPEAKNG_CTYPE_UPPERCASE;
 	case ESPEAKNG_PROPERTY_FULL_STOP:
 		return ESPEAKNG_CTYPE_FULL_STOP;
+	case ESPEAKNG_PROPERTY_QUESTION_MARK:
+		return ESPEAKNG_CTYPE_QUESTION_MARK;
 	}
 
 	// 4. Classify the remaining codepoints.
@@ -271,6 +274,10 @@ tokenizer_state_default(espeak_ng_TOKENIZER *tokenizer)
 		current += utf8_out(c, current);
 		*current = '\0';
 		return ESPEAKNG_TOKEN_FULL_STOP;
+	case ESPEAKNG_CTYPE_QUESTION_MARK:
+		current += utf8_out(c, current);
+		*current = '\0';
+		return ESPEAKNG_TOKEN_QUESTION_MARK;
 	default:
 		current += utf8_out(c, current);
 		*current = '\0';

src/libespeak-ng/tokenizer.h

@@ -52,6 +52,7 @@ typedef enum
 	ESPEAKNG_TOKEN_WORD_MIXEDCASE,
 	ESPEAKNG_TOKEN_WORD_CAPITALIZED,
 	ESPEAKNG_TOKEN_FULL_STOP,
+	ESPEAKNG_TOKEN_QUESTION_MARK,
 } espeak_ng_TOKEN_TYPE;
 
 espeak_ng_TOKEN_TYPE

tests/tokenizer.c

@@ -465,13 +465,21 @@ test_Latn_punctuation_tokens()
 	espeak_ng_TOKENIZER *tokenizer = create_tokenizer();
 	espeak_ng_TEXT_DECODER *decoder = create_text_decoder();
 
-	assert(text_decoder_decode_string(decoder, ".", -1, ESPEAKNG_ENCODING_US_ASCII) == ENS_OK);
+	assert(text_decoder_decode_string(decoder, ". ?", -1, ESPEAKNG_ENCODING_US_ASCII) == ENS_OK);
 	assert(tokenizer_reset(tokenizer, decoder, ESPEAKNG_TOKENIZER_OPTION_TEXT) == 1);
 
 	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_FULL_STOP);
 	assert(tokenizer_get_token_text(tokenizer) != NULL);
 	assert(strcmp(tokenizer_get_token_text(tokenizer), ".") == 0);
 
+	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_WHITESPACE);
+	assert(tokenizer_get_token_text(tokenizer) != NULL);
+	assert(strcmp(tokenizer_get_token_text(tokenizer), " ") == 0);
+
+	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_QUESTION_MARK);
+	assert(tokenizer_get_token_text(tokenizer) != NULL);
+	assert(strcmp(tokenizer_get_token_text(tokenizer), "?") == 0);
+
 	assert(tokenizer_read_next_token(tokenizer) == ESPEAKNG_TOKEN_END_OF_BUFFER);
 	assert(tokenizer_get_token_text(tokenizer) != NULL);
 	assert(*tokenizer_get_token_text(tokenizer) == '\0');
@@ -564,6 +572,9 @@ print_tokens(espeak_ng_TEXT_DECODER *decoder)
 	case ESPEAKNG_TOKEN_FULL_STOP:
 		printf("full stop          : %s\n", tokenizer_get_token_text(tokenizer));
 		break;
+	case ESPEAKNG_TOKEN_QUESTION_MARK:
+		printf("question mark      : %s\n", tokenizer_get_token_text(tokenizer));
+		break;
 	}
 }