in Source/Shared/http_utils.cpp [1359:1547]
bool inner_parse(
const char* encoded,
const char** scheme_begin, const char** scheme_end,
const char** uinfo_begin, const char** uinfo_end,
const char** host_begin, const char** host_end,
_Out_ int* port,
const char** path_begin, const char** path_end,
const char** query_begin, const char** query_end,
const char** fragment_begin, const char** fragment_end)
{
*scheme_begin = nullptr;
*scheme_end = nullptr;
*uinfo_begin = nullptr;
*uinfo_end = nullptr;
*host_begin = nullptr;
*host_end = nullptr;
*port = 0;
*path_begin = nullptr;
*path_end = nullptr;
*query_begin = nullptr;
*query_end = nullptr;
*fragment_begin = nullptr;
*fragment_end = nullptr;
const char* p = encoded;
// IMPORTANT -- A uri may either be an absolute uri, or an relative-reference
// Absolute: 'http://host.com'
// Relative-Reference: '//:host.com', '/path1/path2?query', './path1:path2'
// A Relative-Reference can be disambiguated by parsing for a ':' before the first slash
bool is_relative_reference = true;
const char* p2 = p;
for (; *p2 != '/' && *p2 != '\0'; p2++)
{
if (*p2 == ':')
{
// found a colon, the first portion is a scheme
is_relative_reference = false;
break;
}
}
if (!is_relative_reference)
{
// the first character of a scheme must be a letter
if (!isalpha(*p))
{
return false;
}
// start parsing the scheme, it's always delimited by a colon (must be present)
*scheme_begin = p++;
for (; *p != ':'; p++)
{
if (!is_scheme_character(*p))
{
return false;
}
}
*scheme_end = p;
// skip over the colon
p++;
}
// if we see two slashes next, then we're going to parse the authority portion
// later on we'll break up the authority into the port and host
const char* authority_begin = nullptr;
const char* authority_end = nullptr;
if (*p == '/' && p[1] == '/')
{
// skip over the slashes
p += 2;
authority_begin = p;
// the authority is delimited by a slash (resource), question-mark (query) or octothorpe (fragment)
// or by EOS. The authority could be empty ('file:///C:\file_name.txt')
for (; *p != '/' && *p != '?' && *p != '#' && *p != '\0'; p++)
{
// We're NOT currently supporting IPv6, IPvFuture or username/password in authority
if (!is_authority_character(*p))
{
return false;
}
}
authority_end = p;
// now lets see if we have a port specified -- by working back from the end
if (authority_begin != authority_end)
{
// the port is made up of all digits
const char* port_begin = authority_end - 1;
for (; isdigit(*port_begin) && port_begin != authority_begin; port_begin--)
{
}
if (*port_begin == ':')
{
// has a port
*host_begin = authority_begin;
*host_end = port_begin;
//skip the colon
port_begin++;
*port = convert::scan_string<int>(xsapi_internal_string(port_begin, authority_end), std::locale::classic());
}
else
{
// no port
*host_begin = authority_begin;
*host_end = authority_end;
}
// look for a user_info component
const char* u_end = *host_begin;
for (; is_user_info_character(*u_end) && u_end != *host_end; u_end++)
{
}
if (*u_end == '@')
{
*host_begin = u_end + 1;
*uinfo_begin = authority_begin;
*uinfo_end = u_end;
}
else
{
uinfo_end = uinfo_begin = nullptr;
}
}
}
// if we see a path character or a slash, then the
// if we see a slash, or any other legal path character, parse the path next
if (*p == '/' || is_path_character(*p))
{
*path_begin = p;
// the path is delimited by a question-mark (query) or octothorpe (fragment) or by EOS
for (; *p != '?' && *p != '#' && *p != '\0'; p++)
{
if (!is_path_character(*p))
{
return false;
}
}
*path_end = p;
}
// if we see a ?, then the query is next
if (*p == '?')
{
// skip over the question mark
p++;
*query_begin = p;
// the query is delimited by a '#' (fragment) or EOS
for (; *p != '#' && *p != '\0'; p++)
{
if (!is_query_character(*p))
{
return false;
}
}
*query_end = p;
}
// if we see a #, then the fragment is next
if (*p == '#')
{
// skip over the hash mark
p++;
*fragment_begin = p;
// the fragment is delimited by EOS
for (; *p != '\0'; p++)
{
if (!is_fragment_character(*p))
{
return false;
}
}
*fragment_end = p;
}
return true;
}