/** EOT (end of token) is a label that indicates when the DFA conversion
* algorithm would "fall off the end of a lexer rule". It normally
* means the default clause. So for ('a'..'z')+ you would see a DFA
* with a state that has a..z and EOT emanating from it. a..z would
* jump to a state predicting alt 1 and EOT would jump to a state
* predicting alt 2 (the exit loop branch). EOT implies anything other
* than a..z. If for some reason, the set is "all char" such as with
* the wildcard '.', then EOT cannot match anything. For example,
*
* BLOCK : '{' (.)* '}'
*
* consumes all char until EOF when greedy=true. When all edges are
* combined for the DFA state after matching '}', you will find that
* it is all char. The EOT transition has nothing to match and is
* unreachable. The findNewDFAStatesAndAddDFATransitions() method
* must know to ignore the EOT, so we simply remove it from the
* reachable labels. Later analysis will find that the exit branch
* is not predicted by anything. For greedy=false, we leave only
* the EOT label indicating that the DFA should stop immediately
* and predict the exit branch. The reachable labels are often a
* set of disjoint values like: [<EOT>, 42, {0..41, 43..65534}]
* due to DFA conversion so must construct a pure set to see if
* it is same as Label.ALLCHAR.
*
* Only do this for Lexers.
*
* If EOT coexists with ALLCHAR:
* 1. If not greedy, modify the labels parameter to be EOT
* 2. If greedy, remove EOT from the labels set
*/
protected boolean ReachableLabelsEOTCoexistsWithAllChar(OrderedHashSet labels)
{
Label eot = new Label(Label.EOT);
if ( !labels.containsKey(eot) ) {
return false;
}
[email protected]("### contains EOT");
bool containsAllChar = false;
IntervalSet completeVocab = new IntervalSet();
int n = labels.size();
for (int i=0; i<n; i++) {
Label rl = (Label)labels.get(i);
if ( !rl.Equals(eot) ) {
completeVocab.addAll(rl.Set());
}
}
[email protected]("completeVocab="+completeVocab);
if ( completeVocab.Equals(Label.ALLCHAR) ) {
[email protected]("all char");
containsAllChar = true;
}
return containsAllChar;
}
