private void ResolveReferences()
{
int gid = 1;
Hashtable dict = new Hashtable();
ArrayList explicit_numeric_groups = null;
// number unnamed groups
foreach (CapturingGroup group in caps)
{
if (group.Name != null)
{
continue;
}
dict.Add(gid.ToString(), group);
group.Index = gid++;
++num_groups;
}
// number named groups
foreach (CapturingGroup group in caps)
{
if (group.Name == null)
{
continue;
}
if (dict.Contains(group.Name))
{
CapturingGroup prev = (CapturingGroup)dict [group.Name];
group.Index = prev.Index;
if (group.Index == gid)
{
gid++;
}
else if (group.Index > gid)
{
explicit_numeric_groups.Add(group);
}
continue;
}
if (Char.IsDigit(group.Name [0]))
{
int ptr = 0;
int group_gid = ParseDecimal(group.Name, ref ptr);
if (ptr == group.Name.Length)
{
group.Index = group_gid;
dict.Add(group.Name, group);
++num_groups;
if (group_gid == gid)
{
gid++;
}
else
{
// all numbers before 'gid' are already in the dictionary. So, we know group_gid > gid
if (explicit_numeric_groups == null)
{
explicit_numeric_groups = new ArrayList(4);
}
explicit_numeric_groups.Add(group);
}
continue;
}
}
string gid_s = gid.ToString();
while (dict.Contains(gid_s))
{
gid_s = (++gid).ToString();
}
dict.Add(gid_s, group);
dict.Add(group.Name, group);
group.Index = gid++;
++num_groups;
}
gap = gid; // == 1 + num_groups, if explicit_numeric_groups == null
if (explicit_numeric_groups != null)
{
HandleExplicitNumericGroups(explicit_numeric_groups);
}
// resolve references
foreach (Expression expr in refs.Keys)
{
string name = (string)refs [expr];
if (!dict.Contains(name))
{
if (expr is CaptureAssertion && !Char.IsDigit(name [0]))
{
continue;
}
BackslashNumber bn = expr as BackslashNumber;
if (bn != null && bn.ResolveReference(name, dict))
{
continue;
}
throw NewParseException("Reference to undefined group " +
(Char.IsDigit(name[0]) ? "number " : "name ") +
name);
}
CapturingGroup group = (CapturingGroup)dict[name];
if (expr is Reference)
{
((Reference)expr).CapturingGroup = group;
}
else if (expr is CaptureAssertion)
{
((CaptureAssertion)expr).CapturingGroup = group;
}
else if (expr is BalancingGroup)
{
((BalancingGroup)expr).Balance = group;
}
}
}
private Expression ParseSpecial(RegexOptions options)
{
int p = ptr;
bool ecma = IsECMAScript(options);
Expression expr = null;
switch (pattern[ptr++])
{
// categories
case 'd':
expr = new CharacterClass(ecma ? Category.EcmaDigit : Category.Digit, false);
break;
case 'w':
expr = new CharacterClass(ecma ? Category.EcmaWord : Category.Word, false);
break;
case 's':
expr = new CharacterClass(ecma ? Category.EcmaWhiteSpace : Category.WhiteSpace, false);
break;
case 'p':
// this is odd - ECMAScript isn't supposed to support Unicode,
// yet \p{..} compiles and runs under the MS implementation
// identically to canonical mode. That's why I'm ignoring the
// value of ecma here.
expr = new CharacterClass(ParseUnicodeCategory(), false);
break;
case 'D':
expr = new CharacterClass(ecma ? Category.EcmaDigit : Category.Digit, true);
break;
case 'W':
expr = new CharacterClass(ecma ? Category.EcmaWord : Category.Word, true);
break;
case 'S':
expr = new CharacterClass(ecma ? Category.EcmaWhiteSpace : Category.WhiteSpace, true);
break;
case 'P':
expr = new CharacterClass(ParseUnicodeCategory(), true);
break;
// positions
case 'A': expr = new PositionAssertion(Position.StartOfString); break;
case 'Z': expr = new PositionAssertion(Position.End); break;
case 'z': expr = new PositionAssertion(Position.EndOfString); break;
case 'G': expr = new PositionAssertion(Position.StartOfScan); break;
case 'b': expr = new PositionAssertion(Position.Boundary); break;
case 'B': expr = new PositionAssertion(Position.NonBoundary); break;
// references
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7':
case '8':
case '9': {
ptr--;
int n = ParseNumber(10, 1, 0);
if (n < 0)
{
ptr = p;
return(null);
}
// FIXME test if number is within number of assigned groups
// this may present a problem for right-to-left matching
Reference reference = new BackslashNumber(IsIgnoreCase(options), ecma);
refs.Add(reference, n.ToString());
expr = reference;
break;
}
case 'k': {
char delim = pattern[ptr++];
if (delim == '<')
{
delim = '>';
}
else if (delim != '\'')
{
throw NewParseException("Malformed \\k<...> named backreference.");
}
string name = ParseName();
if (name == null || pattern[ptr] != delim)
{
throw NewParseException("Malformed \\k<...> named backreference.");
}
++ptr;
Reference reference = new Reference(IsIgnoreCase(options));
refs.Add(reference, name);
expr = reference;
break;
}
default:
expr = null;
break;
}
if (expr == null)
{
ptr = p;
}
return(expr);
}