Parse units. This is the only place where we actually have to perform string comparisons rather than simply comparing pointers. Calling strncasecmp for such short strings may incur an excessive overhead. If so, we can perform the comparison ourselves, thus avoiding the function call. Making the lexer separate the units from the value probably isn't helpful, particularly thanks to the need to handle "color: 0f0000;" in quirks mode (this will be lexed as a dimension).

svn path=/trunk/libcss/; revision=6044

1 files changed, 86 insertions, 8 deletions

diff --git a/src/parse/properties.c b/src/parse/properties.c
index 02a9ba5..adf3aa7 100644
--- a/src/parse/properties.c
+++ b/src/parse/properties.c
@@ -6502,18 +6502,96 @@ css_error parse_unit_specifier(css_language *c,
 		fixed *length, uint32_t *unit)
 {
 	const css_token *token;
+	fixed num;
+	size_t consumed = 0;
+	css_string tmp;
 
 	UNUSED(c);
-	UNUSED(length);
-	UNUSED(unit);
 
-	/** \todo Parse units */
+	consumeWhitespace(vector, ctx);
 
-	/* For now, consume everything up to the end of the declaration or !, 
- 	 * whichever comes first */
-	while ((token = parserutils_vector_peek(vector, *ctx)) != NULL &&
-			tokenIsChar(token, '!') == false)
-		parserutils_vector_iterate(vector, ctx);
+	token = parserutils_vector_iterate(vector, ctx);
+	if (token == NULL || (token->type != CSS_TOKEN_DIMENSION &&
+			token->type != CSS_TOKEN_PERCENTAGE))
+		return CSS_INVALID;
+
+	tmp.len = token->idata->len;
+	tmp.data = (uint8_t *) token->idata->data;
+	num = number_from_css_string(&tmp, false, &consumed);
+
+	if (token->type == CSS_TOKEN_DIMENSION) {
+		if (consumed == token->idata->len) {
+			/** \todo In quirks mode, non-zero units should be 
+			 * treated as px too */
+			if (num != 0)
+				return CSS_INVALID;
+			*unit = UNIT_PX;
+		} else if (token->idata->len - consumed == 4) {
+			if (strncasecmp((char *) token->idata->data + consumed, 
+					"grad", 4) == 0)
+				*unit = UNIT_GRAD;
+			else
+				return CSS_INVALID;
+		} else if (token->idata->len - consumed == 3) {
+			if (strncasecmp((char *) token->idata->data + consumed,
+					"kHz", 3) == 0)
+				*unit = UNIT_KHZ;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "deg", 3) == 0)
+				*unit = UNIT_DEG;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "rad", 3) == 0)
+				*unit = UNIT_RAD;
+			else
+				return CSS_INVALID;
+		} else if (token->idata->len - consumed == 2) {
+			if (strncasecmp((char *) token->idata->data + consumed,
+					"Hz", 2) == 0)
+				*unit = UNIT_HZ;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "ms", 2) == 0)
+				*unit = UNIT_MS;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "px", 2) == 0)
+				*unit = UNIT_PX;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "ex", 2) == 0)
+				*unit = UNIT_EX;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "em", 2) == 0)
+				*unit = UNIT_EM;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "in", 2) == 0)
+				*unit = UNIT_IN;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "cm", 2) == 0)
+				*unit = UNIT_CM;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "mm", 2) == 0)
+				*unit = UNIT_MM;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "pt", 2) == 0)
+				*unit = UNIT_PT;
+			else if (strncasecmp((char *) token->idata->data + 
+					consumed, "pc", 2) == 0)
+				*unit = UNIT_PC;
+			else
+				return CSS_INVALID;
+		} else if (token->idata->len - consumed == 1) {
+			if (strncasecmp((char *) token->idata->data + consumed,
+					"s", 1) == 0)
+				*unit = UNIT_S;
+			else
+				return CSS_INVALID;
+		} else
+			return CSS_INVALID;
+	} else {
+		if (consumed != token->idata->len)
+			return CSS_INVALID;
+		*unit = UNIT_PCT;
+	}
+
+	*length = num;
 
 	return CSS_OK;
 }