htex

commit 7a549bd99d0c278a0fcc199e920e709231ea5d25
parent de3711d4d4300d4b5294ec1a8ea4aac4994febde
Author: Robin <kroekerrobin@gmail.com>
Date:   Fri, 11 Aug 2023 15:49:21 +0200

Fix parsing tag and subtag with same name

e.g.:
<div>
  <div></div>
</div

Diffstat:
M
htex.c
 | 
34
----------------------------------
M
html.c
 | 
94
++++++++++++++++++++++++++++++++++++++++++++-----------------------------------
M
html.h
 | 
5
++++-
M
lib.c
 | 
34
++++++++++++++++++++++++++++++++++

4 files changed, 91 insertions(+), 76 deletions(-)
diff --git a/htex.c b/htex.c
@@ -8,40 +8,6 @@
 #include "lib.c"
 #include "html.c"
 
-// Do not use for reading from a socket fd
-bool tryRead(char *buf, FILE *fp)
-{
-	size_t bytesRead = fread(buf, 1, 1, fp);
-	if (feof(fp) != 0)
-		return false;
-	if (ferror(fp) != 0)
-		tryRead(buf, fp);
-	if (bytesRead != 1)
-		tryRead(buf, fp);
-	return true;
-}
-
-char *readFile(FILE *fp)
-{
-	char *text = NULL;
-	int i = 0;
-	char buf;
-	while (1)
-	{
-		if (tryRead(&buf, fp))
-		{
-			text = realloc(text, (i+1) * sizeof(char));
-			text[i] = buf;
-			i++;
-		}
-		else
-			break;
-	}
-	text = realloc(text, (i+1) * sizeof(char));
-	text[i] = 0;
-	return text;
-}
-
 struct filter_opts *parseFilterOpts(const char *pattern)
 {
 	struct filter_opts *opt = malloc(sizeof(struct filter_opts));
diff --git a/html.c b/html.c
@@ -14,6 +14,9 @@ const char *stateToString(enum state s)
 		case STATE_SCRIPT: return "STATE_SCRIPT";
 		case STATE_SCRIPT_POSSIBLE_END_TAG: return "STATE_SCRIPT_POSSIBLE_END_TAG";
 		case STATE_SCRIPT_END_TAG: return "STATE_SCRIPT_END_TAG";
+		case STATE_STYLE: return "STATE_STYLE";
+		case STATE_STYLE_POSSIBLE_END_TAG: return "STATE_STYLE_POSSIBLE_END_TAG";
+		case STATE_STYLE_END_TAG: return "STATE_STYLE_END_TAG";
 	}
 	return "";
 }
@@ -210,6 +213,7 @@ size_t parseDoctype(const char *text)
 
 struct tag *closeLastUnclosedTag(struct tag_list *tagList, const char *endTag, size_t endOffset)
 {
+	printf("Closing %s\n", endTag);
 	for (int i=tagList->len-1; i>-1; i--)
 	{
 		if (strcmp(tagList->tags[i]->name, endTag) == 0)
@@ -232,18 +236,6 @@ struct tag *getLastOpenTag(struct tag_list *tagList)
 	return tagList->tags[0];
 }
 
-/* void saveOuterHtml(struct tag *tag, char *text)
-{
-	int o = 0;
-	for (int i=tag->_outerHtmlBeginOffset; i<tag->_outerHtmlEndOffset; i++)
-	{
-		tag->outerHtml = realloc(tag->outerHtml, (o+1) * sizeof(char));
-		tag->outerHtml[o] = text[i];
-		o++;
-	}
-	tag->outerHtml = realloc(tag->outerHtml, (o+1) * sizeof(char));
-	tag->outerHtml[o] = 0;
-} */
 char *getOuterHtml(char *text, struct tag *t)
 {
 	char *outerHtml = NULL;
@@ -274,19 +266,6 @@ char *getInnerHtml(char *text, struct tag *t)
 	return innerHtml;
 }
 
-/* void saveInnerHtml(struct tag *tag, char *text)
-{
-	int o = 0;
-	for (int i=tag->_innerHtmlBeginOffset; i<tag->_innerHtmlEndOffset; i++)
-	{
-		tag->innerHtml = realloc(tag->innerHtml, (o+1) * sizeof(char));
-		tag->innerHtml[o] = text[i];
-		o++;
-	}
-	tag->innerHtml = realloc(tag->innerHtml, (o+1) * sizeof(char));
-	tag->innerHtml[o] = 0;
-} */
-
 void setInnerHtmlEndOffset(struct tag *closedTag, char *text, size_t off)
 {
 	int i = off;
@@ -322,9 +301,9 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 		}
 		else
 		{
-			// char *the_codepoint = cpToChars(cp, ret);
-			// printf("cp: %02X, %s, %s\n", cp, the_codepoint, stateToString(state));
-			// free(the_codepoint);
+			char *the_codepoint = cpToChars(cp, ret);
+			printf("cp: %02X, %s, %s\n", cp, the_codepoint, stateToString(state));
+			free(the_codepoint);
 			switch (state)
 			{
 				case STATE_INNER_TEXT:
@@ -333,7 +312,6 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 						state = STATE_TAG;
 						break;
 					}
-					// stillOpenTag->innerText = stringCat(stillOpenTag->innerText, cpToChars(cp, ret));
 					break;
 				case STATE_TAG:
 					if (cp == SOLIDUS)
@@ -346,8 +324,8 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 						state = STATE_COMMENT;
 						break;
 					}
-					struct tag *oneTag = parseTag(text, off, STATE_BEGIN_TAG_NAME, tagList);
 					stillOpenTag = getLastOpenTag(tagList);
+					struct tag *oneTag = parseTag(text, off, STATE_BEGIN_TAG_NAME, tagList);
 					stillOpenTag->children = realloc(
 						stillOpenTag->children,
 						(stillOpenTag->childrenLen+1) * sizeof(struct tag)
@@ -362,12 +340,11 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 						tag->_innerHtmlBeginOffset = off+1;
 						tag->_isVoidElement = isVoidElement(tag->name);
 						if (tag->_isVoidElement)
-						{
 							tag->_outerHtmlEndOffset = off+1;
-							// saveOuterHtml(tag, text);
-						}
 						if (strcmp(tag->name, "script") == 0)
 							state = STATE_SCRIPT;
+						else if (strcmp(tag->name, "style") == 0)
+							state = STATE_STYLE;
 						else
 							state = STATE_INNER_TEXT;
 						break;
@@ -387,8 +364,6 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 					{
 						struct tag *closedTag = closeLastUnclosedTag(tagList, endTag, off+ret);
 						setInnerHtmlEndOffset(closedTag, text, off);
-						// saveOuterHtml(closedTag, text);
-						// saveInnerHtml(closedTag, text);
 						free(endTag);
 						endTag = malloc(sizeof(char));
 						endTag[0] = 0;
@@ -406,9 +381,13 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 						if (tag->_isVoidElement)
 						{
 							tag->_outerHtmlEndOffset = off+1;
-							// saveOuterHtml(tag, text);
 						}
-						state = STATE_INNER_TEXT;
+						if (strcmp(tag->name, "script") == 0)
+							state = STATE_SCRIPT;
+						else if (strcmp(tag->name, "style") == 0)
+							state = STATE_STYLE;
+						else
+							state = STATE_INNER_TEXT;
 						break;
 					}
 					if (isASCIIWhitespace(cp))
@@ -485,9 +464,13 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 						if (tag->_isVoidElement)
 						{
 							tag->_outerHtmlEndOffset = off+1;
-							// saveOuterHtml(tag, text);
 						}
-						state = STATE_INNER_TEXT;
+						if (strcmp(tag->name, "script") == 0)
+							state = STATE_SCRIPT;
+						else if (strcmp(tag->name, "style") == 0)
+							state = STATE_STYLE;
+						else
+							state = STATE_INNER_TEXT;
 						break;
 					}
 					if (
@@ -516,6 +499,33 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 					else
 						hyphenCount = 0;
 					break;
+				case STATE_STYLE:
+					if (cp == LESS_THAN_SIGN)
+					{
+						state = STATE_STYLE_POSSIBLE_END_TAG;
+						break;
+					}
+					break;
+				case STATE_STYLE_POSSIBLE_END_TAG:
+					if (cp == SOLIDUS)
+						state = STATE_STYLE_END_TAG;
+					else
+						state = STATE_STYLE;
+					break;
+				case STATE_STYLE_END_TAG:
+					if (cp == GREATER_THAN_SIGN)
+					{
+						struct tag *closedTag = closeLastUnclosedTag(tagList, endTag, off+ret);
+						setInnerHtmlEndOffset(closedTag, text, off);
+						free(endTag);
+						endTag = malloc(sizeof(char));
+						endTag[0] = 0;
+						state = STATE_INNER_TEXT;
+						break;
+					}
+					if (!isASCIIWhitespace(cp))
+						endTag = stringCat(endTag, cpToChars(cp, ret));
+					break;
 				case STATE_SCRIPT:
 					if (cp == LESS_THAN_SIGN)
 					{
@@ -534,8 +544,6 @@ struct tag *parseTag(char *text, size_t offset, enum state state, struct tag_lis
 					{
 						struct tag *closedTag = closeLastUnclosedTag(tagList, endTag, off+ret);
 						setInnerHtmlEndOffset(closedTag, text, off);
-						// saveOuterHtml(closedTag, text);
-						// saveInnerHtml(closedTag, text);
 						free(endTag);
 						endTag = malloc(sizeof(char));
 						endTag[0] = 0;
@@ -688,6 +696,10 @@ void filterHtml(char *text, struct filter_opts *opts)
 	if (len)
 		text = text + len;
 	struct tag *rootTag = parseTag(text, 0, STATE_INNER_TEXT, tagList);
+	for (int i=0; i<tagList->len; i++)
+	{
+		printf("%s\n", tagList->tags[i]->name);
+	}
 	findTag(rootTag, opts, foundTags);
 	printResult(text, opts, foundTags);
 	// printHtml(rootTag, 0);
diff --git a/html.h b/html.h
@@ -73,7 +73,10 @@ enum state
 	STATE_COMMENT,
 	STATE_SCRIPT,
 	STATE_SCRIPT_POSSIBLE_END_TAG,
-	STATE_SCRIPT_END_TAG
+	STATE_SCRIPT_END_TAG,
+	STATE_STYLE,
+	STATE_STYLE_POSSIBLE_END_TAG,
+	STATE_STYLE_END_TAG
 };
 
 enum attr_value_syntax
diff --git a/lib.c b/lib.c
@@ -77,3 +77,37 @@ char *trim(char *text)
 	free(text);
 	return trimmedText;
 }
+
+// Do not use for reading from a socket fd
+bool tryRead(char *buf, FILE *fp)
+{
+	size_t bytesRead = fread(buf, 1, 1, fp);
+	if (feof(fp) != 0)
+		return false;
+	if (ferror(fp) != 0)
+		tryRead(buf, fp);
+	if (bytesRead != 1)
+		tryRead(buf, fp);
+	return true;
+}
+
+char *readFile(FILE *fp)
+{
+	char *text = NULL;
+	int i = 0;
+	char buf;
+	while (1)
+	{
+		if (tryRead(&buf, fp))
+		{
+			text = realloc(text, (i+1) * sizeof(char));
+			text[i] = buf;
+			i++;
+		}
+		else
+			break;
+	}
+	text = realloc(text, (i+1) * sizeof(char));
+	text[i] = 0;
+	return text;
+}