Make deterministic the default

Signed-off-by: Mattias Andrée <m@maandree.se>

1 files changed, 27 insertions, 10 deletions

diff --git a/README b/README
index 37b2701..23d231b 100644
--- a/README
+++ b/README
@@ -76,11 +76,14 @@ EXTENDED DESCRIPTION
 		_low             = character | integer;
 		_high            = character | integer;
 
+		nondeterministic = "?";
+
+		committed        = "+", _, _operand;
 		rejection        = "!", _, _operand;
 		concatenation    = _operand, {_, ",", _, _operand};
-		alternation      = concatenation, {_, "|", _, concatenation};
-		optional         = "[", _, _expression, _, "]";
-		repeated         = "{", _, _expression, _, "}";
+		alternation      = concatenation, {_, [nondeterministic], "|", _, concatenation};
+		optional         = [nondeterministic], "[", _, _expression, _, "]";
+		repeated         = [nondeterministic], "{", _, _expression, _, "}";
 		group            = "(", _, _expression, _, ")";
 		char-range       = "<", _, _low, _, ",", _, _high, "_", ">";
 		exception        = "-";
@@ -88,7 +91,7 @@ EXTENDED DESCRIPTION
 
 		_literal         = char-range | exception | string;
 		_group           = optional | repeated | group | embedded-rule;
-		_operand         = _group | _literal | rejection;
+		_operand         = _group | _literal | rejection | committed;
 
 		_expression      = alternation;
 
@@ -109,12 +112,24 @@ EXTENDED DESCRIPTION
 	reached, the parser will terminate there.
 
 	Repeated symbols may occur any number of times, including
-	zero. The compiler is able to backtrack if it takes too much.
+	zero. The parser will try to take as much as as possible.
+
+	Optional symbols are taken whenever possible.
+
+	Concatenation has higher precedence than alternation, groups
+	("(", ..., ")") have no semantic meaning and are useful only
+	for including alternations inside concatenations or put
+	alternations or concatenations inside a commitment or
+	rejection.
 
-	Concatenation has higher precedence than alternation,
-	groups ("(", ..., ")") have no semantic meaning and are useful
-	only to put a alternation inside a concatenation without
-	creating a new rule for that.
+	The parser has the ability to be non-deterministic, which
+	can make it really slow. The speed up parsing, you can add
+	commits. Once the committed sentence has been matched, the
+	branching-points inside it are unrecorded, so when the parser
+	backtracks, it will not try different choices inside the
+	committed sentence. Commit sentences are undone by the parser
+	when it backtracks to a branching-point outside the committed
+	sentence.
 
 	In character ranges, the _high and _low values must be at
 	least 0 and at most 255, and _high must be greater than _low.
@@ -125,7 +140,9 @@ EXTENDED DESCRIPTION
 
 	Left recursion is illegal (it will cause stack overflow at
 	runtime as the empty condition before the recursion is always
-	met).
+	met). Likewise, repeated optional sentences are illegal; a
+	repeated sentence must always consume input, otherwise it
+	gets stuck.
 
     Right-context-sensitive grammar
 	libparser originally used context-free grammar, but with