summaryrefslogtreecommitdiff
path: root/src
diff options
context:
space:
mode:
authorarseny.kapoulkine <arseny.kapoulkine@99668b35-9821-0410-8761-19e4c4f06640>2010-05-10 19:15:44 +0000
committerarseny.kapoulkine <arseny.kapoulkine@99668b35-9821-0410-8761-19e4c4f06640>2010-05-10 19:15:44 +0000
commit7bda2cb529550fae678224102e1c5d0697dacf04 (patch)
treebfd1821299e405ad8f82b8b19df4ccc7188e5836 /src
parent55f3cba20c5fc037e9f5973984ea9d596680a6f8 (diff)
Implemented better DOCTYPE parsing, added more DOCTYPE tests
git-svn-id: http://pugixml.googlecode.com/svn/trunk@409 99668b35-9821-0410-8761-19e4c4f06640
Diffstat (limited to 'src')
-rw-r--r--src/pugixml.cpp152
1 files changed, 114 insertions, 38 deletions
diff --git a/src/pugixml.cpp b/src/pugixml.cpp
index aa10bbc..dcb25f7 100644
--- a/src/pugixml.cpp
+++ b/src/pugixml.cpp
@@ -1739,6 +1739,116 @@ namespace
{
}
+ // DOCTYPE consists of nested sections of the following possible types:
+ // <!-- ... -->, <? ... ?>, "...", '...'
+ // <![...]]>
+ // <!...>
+ // First group can not contain nested groups
+ // Second group can contain nested groups of the same type
+ // Third group can contain all other groups
+ xml_parse_result parse_doctype_primitive(char_t*& s, char_t* buffer_start)
+ {
+ if (*s == '"' || *s == '\'')
+ {
+ // quoted string
+ char_t ch = *s++;
+ SCANFOR(*s == ch);
+ if (!*s) THROW_ERROR(status_bad_doctype, s);
+
+ s++;
+ }
+ else if (s[0] == '<' && s[1] == '?')
+ {
+ // <? ... ?>
+ s += 2;
+ SCANFOR(s[0] == '?' && s[1] == '>'); // no need for ENDSWITH because ?> can't terminate proper doctype
+ if (!*s) THROW_ERROR(status_bad_doctype, s);
+
+ s += 2;
+ }
+ else if (s[0] == '<' && s[1] == '!' && s[2] == '-' && s[3] == '-')
+ {
+ s += 4;
+ SCANFOR(s[0] == '-' && s[1] == '-' && s[2] == '>'); // no need for ENDSWITH because --> can't terminate proper doctype
+ if (!*s) THROW_ERROR(status_bad_doctype, s);
+
+ s += 4;
+ }
+ else THROW_ERROR(status_bad_doctype, s);
+
+ THROW_ERROR(status_ok, s);
+ }
+
+ xml_parse_result parse_doctype_ignore(char_t*& s, char_t* buffer_start)
+ {
+ assert(s[0] == '<' && s[1] == '!' && s[2] == '[');
+ s++;
+
+ while (*s)
+ {
+ if (s[0] == '<' && s[1] == '!' && s[2] == '[')
+ {
+ // nested ignore section
+ xml_parse_result res = parse_doctype_ignore(s, buffer_start);
+
+ if (!res) return res;
+ }
+ else if (s[0] == ']' && s[1] == ']' && s[2] == '>')
+ {
+ // ignore section end
+ s += 3;
+
+ THROW_ERROR(status_ok, s);
+ }
+ else s++;
+ }
+
+ THROW_ERROR(status_bad_doctype, s);
+ }
+
+ xml_parse_result parse_doctype(char_t*& s, char_t* buffer_start, char_t endch, bool toplevel)
+ {
+ assert(s[0] == '<' && s[1] == '!');
+ s++;
+
+ while (*s)
+ {
+ if (s[0] == '<' && s[1] == '!' && s[2] != '-')
+ {
+ if (s[2] == '[')
+ {
+ // ignore
+ xml_parse_result res = parse_doctype_ignore(s, buffer_start);
+
+ if (!res) return res;
+ }
+ else
+ {
+ // some control group
+ xml_parse_result res = parse_doctype(s, buffer_start, endch, false);
+
+ if (!res) return res;
+ }
+ }
+ else if (s[0] == '<' || s[0] == '"' || s[0] == '\'')
+ {
+ // unknown tag (forbidden), or some primitive group
+ xml_parse_result res = parse_doctype_primitive(s, buffer_start);
+
+ if (!res) return res;
+ }
+ else if (*s == '>')
+ {
+ s++;
+
+ THROW_ERROR(status_ok, s);
+ }
+ else s++;
+ }
+
+ THROW_ERROR((toplevel && endch == '>') ? status_ok : status_bad_doctype, s);
+ }
+
xml_parse_result parse_exclamation(char_t*& ref_s, xml_node_struct* cursor, unsigned int optmsk, char_t* buffer_start, char_t endch)
{
// load into registers
@@ -1831,47 +1941,13 @@ namespace
}
else if (s[0] == 'D' && s[1] == 'O' && s[2] == 'C' && s[3] == 'T' && s[4] == 'Y' && s[5] == 'P' && ENDSWITH(s[6], 'E'))
{
- if (s[6] != 'E') THROW_ERROR(status_bad_doctype, s);
+ if (s[6] != 'E') THROW_ERROR(status_bad_doctype, s);
- LOC_DOCTYPE:
- SCANFOR(*s == '\'' || *s == '"' || *s == '[' || *s == '>');
- if (*s == 0 && endch != '>') THROW_ERROR(status_bad_doctype, s);
+ s -= 2;
- if (*s == '\'' || *s == '"') // '...SYSTEM "..."
- {
- ch = *s++;
- SCANFOR(*s == ch);
- if (*s == 0 && endch != '>') THROW_ERROR(status_bad_doctype, s);
+ xml_parse_result res = parse_doctype(s, buffer_start, endch, true);
- s += (*s != 0);
- goto LOC_DOCTYPE;
- }
-
- if(*s == '[') // '...[...'
- {
- ++s;
- unsigned int bd = 1; // Bracket depth counter.
- while (*s!=0) // Loop till we're out of all brackets.
- {
- if (*s == ']') --bd;
- else if (*s == '[') ++bd;
- if (bd == 0) break;
- ++s;
- }
-
- if (bd != 0) THROW_ERROR(status_bad_doctype, s);
- }
-
- SCANFOR(*s == '>');
-
- if (*s == 0)
- {
- if (endch != '>') THROW_ERROR(status_bad_doctype, s);
- }
- else
- {
- ++s;
- }
+ if (!res) return res;
}
else if (*s == 0 && endch == '-') THROW_ERROR(status_bad_comment, s);
else if (*s == 0 && endch == '[') THROW_ERROR(status_bad_cdata, s);