diff options
Diffstat (limited to 'README')
| -rw-r--r-- | README | 37 |
1 files changed, 27 insertions, 10 deletions
@@ -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 |