feat(yxml): support content fragment parsing

deps/yxml.c

@@ -222,7 +222,7 @@ static yxml_ret_t yxml_selfclose(yxml_t *x, unsigned ch) {
 		return YXML_ELEMEND;
 	}
 	x->elem = (char *)x->stack;
-	x->state = YXMLS_misc3;
+	x->state = x->contentmode ? YXMLS_misc2 : YXMLS_misc3;
 	return YXML_ELEMEND;
 }
 
@@ -292,6 +292,7 @@ static yxml_ret_t yxml_refend(yxml_t *x, yxml_ret_t ret) {
 	if(!ch || ch > 0x10FFFF || ch == 0xFFFE || ch == 0xFFFF || (ch-0xDFFF) < 0x7FF)
 		return YXML_EREF;
 	yxml_setutf8(x->data, ch);
+	x->isReference = true;
 	return ret;
 }
 
@@ -308,7 +309,15 @@ void yxml_init(yxml_t *x, void *stack, size_t stacksize) {
 	x->state = YXMLS_init;
 }
 
+void yxml_init_content(yxml_t *x, void *stack, size_t stacksize) {
+	yxml_init(x, stack, stacksize);
+	x->contentmode = true;
+	x->state = YXMLS_misc2;
+}
+
 yxml_ret_t yxml_parse(yxml_t *x, int _ch) {
+	x->isReference = false;
+
 	/* Ensure that characters are in the range of 0..255 rather than -126..125.
 	 * All character comparisons are done with positive integers. */
 	unsigned ch = (unsigned)(_ch+256) & 0xff;
@@ -1026,8 +1035,8 @@ yxml_ret_t yxml_parse(yxml_t *x, int _ch) {
 }
 
 yxml_ret_t yxml_eof(yxml_t *x) {
-	if(x->state != YXMLS_misc3)
+	if(x->state != YXMLS_misc3 &&
+	   !(x->contentmode && x->state == YXMLS_misc2 && x->stacklen == 0))
 		return YXML_EEOF;
 	return YXML_OK;
 }
-

deps/yxml.h

@@ -23,6 +23,7 @@
 #ifndef YXML_H
 #define YXML_H
 
+#include <stdbool.h>
 #include <stdint.h>
 #include <stddef.h>
 
@@ -94,6 +95,10 @@ typedef struct {
 	 */
 	char data[8];
 
+	/* True if the current YXML_CONTENT or YXML_ATTRVAL event was produced from
+	 * a character or entity reference. Valid until the next yxml_parse() call. */
+	bool isReference;
+
 	/* Name of the current attribute. Changed after YXML_ATTRSTART, valid up to
 	 * and including the next YXML_ATTREND. */
 	char *attr;
@@ -119,6 +124,7 @@ typedef struct {
 	int nextstate; /* Used for '@' state remembering and for the "string" consuming state */
 	unsigned ignore;
 	unsigned char *string;
+	bool contentmode;
 } yxml_t;
 
 #ifdef __cplusplus
@@ -127,14 +133,20 @@ extern "C" {
 
 void yxml_init(yxml_t *, void *, size_t);
 
+/* Initialize the parser for an XML content fragment without an enclosing
+ * element. Character data, references, CDATA sections, comments, processing
+ * instructions and balanced child elements are accepted. */
+void yxml_init_content(yxml_t *, void *, size_t);
+
 yxml_ret_t yxml_parse(yxml_t *, int);
 
 /* May be called after the last character has been given to yxml_parse().
- * Returns YXML_OK if the XML document is valid, YXML_EEOF otherwise.  Using
+ * Returns YXML_OK if the XML document or content fragment is valid, YXML_EEOF
- * this function isn't really necessary, but can be used to detect documents
+ * otherwise. Using this function isn't really necessary, but can be used to
- * that don't end correctly. In particular, an error is returned when the XML
+ * detect input that doesn't end correctly. In document mode, an error is
- * document did not contain a (complete) root element, or when the document
+ * returned when there is no complete root element. In content mode, an error
- * ended while in a comment or processing instruction. */
+ * is returned for unbalanced child elements. Both modes reject input ending
+ * inside a reference, CDATA section, comment or processing instruction. */
 yxml_ret_t yxml_eof(yxml_t *);
 
 #ifdef __cplusplus

tests/check_yxml.c

@@ -37,6 +37,31 @@ yxml_parse_full(const char *xml, size_t len, char *buf, size_t buflen,
     return yxml_eof(&x);
 }
 
+static yxml_ret_t
+yxml_parse_content_full(const char *xml, char *buf, size_t buflen,
+                        char *out, size_t outsize, size_t *outlen,
+                        size_t *references) {
+    yxml_t x;
+    yxml_init_content(&x, buf, buflen);
+    *outlen = 0;
+    *references = 0;
+    for(size_t i = 0; i < strlen(xml); i++) {
+        yxml_ret_t ret = yxml_parse(&x, (unsigned char)xml[i]);
+        if(ret < YXML_OK)
+            return ret;
+        if(x.isReference)
+            (*references)++;
+        if(ret != YXML_CONTENT)
+            continue;
+        size_t length = strlen(x.data);
+        if(*outlen > outsize || length > outsize - *outlen)
+            return YXML_ESTACK;
+        memcpy(&out[*outlen], x.data, length);
+        *outlen += length;
+    }
+    return yxml_eof(&x);
+}
+
 /* === xml_tokenize wrapper tests === */
 
 START_TEST(parseElement) {
@@ -344,6 +369,75 @@ START_TEST(yxmlEntityRef_content) {
     ck_assert_int_gt(cnt, 0);
 } END_TEST
 
+START_TEST(yxmlEntityRef_flag) {
+    char buf[512];
+    const char *xml = "<r a=\"plain;&amp;plain\">plain;&#65;plain</r>";
+    yxml_t x;
+    yxml_init(&x, buf, sizeof(buf));
+    size_t contentReferences = 0;
+    size_t attributeReferences = 0;
+    for(size_t i = 0; i < strlen(xml); i++) {
+        yxml_ret_t ret = yxml_parse(&x, (unsigned char)xml[i]);
+        ck_assert_int_ge(ret, YXML_OK);
+        if(x.isReference) {
+            if(ret == YXML_CONTENT)
+                contentReferences++;
+            else if(ret == YXML_ATTRVAL)
+                attributeReferences++;
+            else
+                ck_abort_msg("Reference flag set for event %d", ret);
+        }
+    }
+    ck_assert_uint_eq(contentReferences, 1);
+    ck_assert_uint_eq(attributeReferences, 1);
+} END_TEST
+
+START_TEST(yxmlContentFragment) {
+    char buf[512];
+    char out[128];
+    size_t outlen;
+    size_t references;
+    const char *content =
+        "a&amp;<?pi data?><!--ignored-->&lt;&#65;&#x20AC;"
+        "<![CDATA[&gt;]]>z";
+    yxml_ret_t ret = yxml_parse_content_full(
+        content, buf, sizeof(buf), out, sizeof(out) - 1,
+        &outlen, &references);
+    ck_assert_int_eq(ret, YXML_OK);
+    out[outlen] = 0;
+    ck_assert_str_eq(out, "a&<A\xE2\x82\xAC&gt;z");
+    ck_assert_uint_eq(references, 4);
+
+    yxml_t empty;
+    yxml_init_content(&empty, buf, sizeof(buf));
+    ck_assert_int_eq(yxml_eof(&empty), YXML_OK);
+} END_TEST
+
+START_TEST(yxmlContentFragment_nested) {
+    char buf[512];
+    yxml_t x;
+    yxml_init_content(&x, buf, sizeof(buf));
+    const char *content = "before<child>inside</child>after";
+    for(size_t i = 0; i < strlen(content); i++)
+        ck_assert_int_ge(yxml_parse(&x, (unsigned char)content[i]), YXML_OK);
+    ck_assert_int_eq(yxml_eof(&x), YXML_OK);
+} END_TEST
+
+START_TEST(yxmlContentFragment_incomplete) {
+    const char *invalid[] = {
+        "&amp", "<![CDATA[text", "<!-- comment", "<?pi data", "<child>"
+    };
+    for(size_t i = 0; i < sizeof(invalid) / sizeof(invalid[0]); i++) {
+        char buf[512];
+        yxml_t x;
+        yxml_init_content(&x, buf, sizeof(buf));
+        for(size_t j = 0; j < strlen(invalid[i]); j++)
+            ck_assert_int_ge(yxml_parse(&x, (unsigned char)invalid[i][j]),
+                             YXML_OK);
+        ck_assert_int_eq(yxml_eof(&x), YXML_EEOF);
+    }
+} END_TEST
+
 START_TEST(yxmlEntityRef_unknown) {
     /* Unknown entity -> YXML_EREF */
     char buf[512];
@@ -898,6 +992,10 @@ static Suite *testSuite_yxml(void) {
 
     TCase *tc_ref = tcase_create("yxml_refs");
     tcase_add_test(tc_ref, yxmlEntityRef_content);
+    tcase_add_test(tc_ref, yxmlEntityRef_flag);
+    tcase_add_test(tc_ref, yxmlContentFragment);
+    tcase_add_test(tc_ref, yxmlContentFragment_nested);
+    tcase_add_test(tc_ref, yxmlContentFragment_incomplete);
     tcase_add_test(tc_ref, yxmlEntityRef_unknown);
     tcase_add_test(tc_ref, yxmlEntityRef_tooLong);
     tcase_add_test(tc_ref, yxmlCharRef_decimalASCII);