4

我创建了一个语法,其精简版本如下所示:

(0) exp1: ternary;
(1) exp1: exp2;
(2) ternary: exp2 "?" exp1 ":" exp1;
(3) exp2: exp2 "+" exp3;
(4) exp2: exp3;
(5) exp3: maybe;
(6) exp3: "1";
(7) maybe: exp3 "?";

我相信这种语言是明确的,应该是 LR 可解析的。(如果我错了,请告诉我!)

但是,当我尝试为此语法生成 LR(1) 解析器时,我会遇到移位/归约冲突,因为当解析器看到exp3前瞻?时,它不知道是移位还是归约:

Conflicts in state 3:
    Reduction using rule 4: exp2:  exp3 · | "?"
    Shift to state 6

Conflicts in state 9:
    Reduction using rule 3: exp2:  exp2 "+" exp3 · | "?"
    Shift to state 6

Conflicts in state 13:
    Reduction using rule 4: exp2:  exp3 · | "?"
    Shift to state 16

Conflicts in state 20:
    Reduction using rule 4: exp2:  exp3 · | "?"
    Shift to state 23

Conflicts in state 25:
    Reduction using rule 3: exp2:  exp2 "+" exp3 · | "?"
    Shift to state 23

Conflicts in state 28:
    Reduction using rule 3: exp2:  exp2 "+" exp3 · | "?"
    Shift to state 16

有没有一种合理的方法可以使这种语言 LR(1) 可解析(没有冲突)?

还是 GLR(或者可能是 LR(2)?)对于这样的语言,我唯一现实的选择是什么?
(或者我认为语言一开始是明确的,我什至是错误的?)


作为参考,我生成的模棱两可的状态机如下(其中 ♦ 是 EOF):

State 0:
    exp1:  · ternary | {♦} → shift 1
    ternary:  · exp2 "?" exp1 ":" exp1 | {♦} → shift 2
    exp2:  · exp2 "+" exp3 | {"?", "+"} → shift 2
    exp2:  · exp3 | {"?", "+"} → shift 3
    exp3:  · maybe | {"?", "+"} → shift 4
    exp3:  · "1" | {"?", "+"} → shift 5
    maybe:  · exp3 "?" | {"?", "+"} → shift 3

State 1:
    exp1:  ternary · | {♦} → reduce 0

State 2:
    ternary:  exp2 · "?" exp1 ":" exp1 | {♦} → shift 7
    exp2:  exp2 · "+" exp3 | {"?", "+"} → shift 8

State 3:
    exp2:  exp3 · | {"+"} → reduce 4
    exp2:  exp3 · | {"?"} → reduce 4 shift 6
    maybe:  exp3 · "?" | {"?", "+"} → reduce 4 shift 6

State 4:
    exp3:  maybe · | {"?", "+"} → reduce 5

State 5:
    exp3:  "1" · | {"?", "+"} → reduce 6

State 6:
    maybe:  exp3 "?" · | {"?", "+"} → reduce 7

State 7:
    exp1:  · ternary | {":"} → shift 10
    exp1:  · exp2 | {":"} → shift 11
    ternary:  · exp2 "?" exp1 ":" exp1 | {":"} → shift 11
    ternary:  exp2 "?" · exp1 ":" exp1 | {♦} → shift 12
    exp2:  · exp2 "+" exp3 | {"?", ":", "+"} → shift 11
    exp2:  · exp3 | {"?", ":", "+"} → shift 13
    exp3:  · maybe | {"?", ":", "+"} → shift 14
    exp3:  · "1" | {"?", ":", "+"} → shift 15
    maybe:  · exp3 "?" | {"?", ":", "+"} → shift 13

State 8:
    exp2:  exp2 "+" · exp3 | {"?", "+"} → shift 9
    exp3:  · maybe | {"?", "+"} → shift 4
    exp3:  · "1" | {"?", "+"} → shift 5
    maybe:  · exp3 "?" | {"?", "+"} → shift 9

State 9:
    exp2:  exp2 "+" exp3 · | {"+"} → reduce 3
    exp2:  exp2 "+" exp3 · | {"?"} → reduce 3 shift 6
    maybe:  exp3 · "?" | {"?", "+"} → reduce 3 shift 6

State 10:
    exp1:  ternary · | {":"} → reduce 0

State 11:
    exp1:  exp2 · | {":"} → reduce 1
    ternary:  exp2 · "?" exp1 ":" exp1 | {":"} → shift 26
    exp2:  exp2 · "+" exp3 | {"?", ":", "+"} → shift 27

State 12:
    ternary:  exp2 "?" exp1 · ":" exp1 | {♦} → shift 17

State 13:
    exp2:  exp3 · | {":", "+"} → reduce 4
    exp2:  exp3 · | {"?"} → reduce 4 shift 16
    maybe:  exp3 · "?" | {"?", ":", "+"} → reduce 4 shift 16

State 14:
    exp3:  maybe · | {"?", ":", "+"} → reduce 5

State 15:
    exp3:  "1" · | {"?", ":", "+"} → reduce 6

State 16:
    maybe:  exp3 "?" · | {"?", ":", "+"} → reduce 7

State 17:
    exp1:  · ternary | {♦} → shift 1
    exp1:  · exp2 | {♦} → shift 18
    ternary:  · exp2 "?" exp1 ":" exp1 | {♦} → shift 18
    ternary:  exp2 "?" exp1 ":" · exp1 | {♦} → shift 19
    exp2:  · exp2 "+" exp3 | {♦, "?", "+"} → shift 18
    exp2:  · exp3 | {♦, "?", "+"} → shift 20
    exp3:  · maybe | {♦, "?", "+"} → shift 21
    exp3:  · "1" | {♦, "?", "+"} → shift 22
    maybe:  · exp3 "?" | {♦, "?", "+"} → shift 20

State 18:
    exp1:  exp2 · | {♦} → reduce 1
    ternary:  exp2 · "?" exp1 ":" exp1 | {♦} → shift 7
    exp2:  exp2 · "+" exp3 | {♦, "?", "+"} → shift 24

State 19:
    ternary:  exp2 "?" exp1 ":" exp1 · | {♦} → reduce 2

State 20:
    exp2:  exp3 · | {♦, "+"} → reduce 4
    exp2:  exp3 · | {"?"} → reduce 4 shift 23
    maybe:  exp3 · "?" | {♦, "?", "+"} → reduce 4 shift 23

State 21:
    exp3:  maybe · | {♦, "?", "+"} → reduce 5

State 22:
    exp3:  "1" · | {♦, "?", "+"} → reduce 6

State 23:
    maybe:  exp3 "?" · | {♦, "?", "+"} → reduce 7

State 24:
    exp2:  exp2 "+" · exp3 | {♦, "?", "+"} → shift 25
    exp3:  · maybe | {♦, "?", "+"} → shift 21
    exp3:  · "1" | {♦, "?", "+"} → shift 22
    maybe:  · exp3 "?" | {♦, "?", "+"} → shift 25

State 25:
    exp2:  exp2 "+" exp3 · | {♦, "+"} → reduce 3
    exp2:  exp2 "+" exp3 · | {"?"} → reduce 3 shift 23
    maybe:  exp3 · "?" | {♦, "?", "+"} → reduce 3 shift 23

State 26:
    exp1:  · ternary | {":"} → shift 10
    exp1:  · exp2 | {":"} → shift 11
    ternary:  · exp2 "?" exp1 ":" exp1 | {":"} → shift 11
    ternary:  exp2 "?" · exp1 ":" exp1 | {":"} → shift 29
    exp2:  · exp2 "+" exp3 | {"?", ":", "+"} → shift 11
    exp2:  · exp3 | {"?", ":", "+"} → shift 13
    exp3:  · maybe | {"?", ":", "+"} → shift 14
    exp3:  · "1" | {"?", ":", "+"} → shift 15
    maybe:  · exp3 "?" | {"?", ":", "+"} → shift 13

State 27:
    exp2:  exp2 "+" · exp3 | {"?", ":", "+"} → shift 28
    exp3:  · maybe | {"?", ":", "+"} → shift 14
    exp3:  · "1" | {"?", ":", "+"} → shift 15
    maybe:  · exp3 "?" | {"?", ":", "+"} → shift 28

State 28:
    exp2:  exp2 "+" exp3 · | {":", "+"} → reduce 3
    exp2:  exp2 "+" exp3 · | {"?"} → reduce 3 shift 16
    maybe:  exp3 · "?" | {"?", ":", "+"} → reduce 3 shift 16

State 29:
    ternary:  exp2 "?" exp1 · ":" exp1 | {":"} → shift 30

State 30:
    exp1:  · ternary | {":"} → shift 10
    exp1:  · exp2 | {":"} → shift 11
    ternary:  · exp2 "?" exp1 ":" exp1 | {":"} → shift 11
    ternary:  exp2 "?" exp1 ":" · exp1 | {":"} → shift 31
    exp2:  · exp2 "+" exp3 | {"?", ":", "+"} → shift 11
    exp2:  · exp3 | {"?", ":", "+"} → shift 13
    exp3:  · maybe | {"?", ":", "+"} → shift 14
    exp3:  · "1" | {"?", ":", "+"} → shift 15
    maybe:  · exp3 "?" | {"?", ":", "+"} → shift 13

State 31:
    ternary:  exp2 "?" exp1 ":" exp1 · | {":"} → reduce 2
4

1 回答 1

1

我认为这可能是一个优先问题。当解析器查看以下内容时,您会发生冲突:

 a + b ? c : d

在解析器已经看到a + b ?和正在查看的时候c,它无法决定是否需要

  1. Reduce b?,以便它解析表单的表达式,a + (b?)然后从那里继续,或者

  2. Reduce a + b,以便它解析表单的表达式(a + b) ? c : d

我认为这里的挑战是,在一种情况下,?优先级非常低(用作三元运算符时),而在另一种情况下,它具有非常高的优先级(用作一元运算符时)。但是,如果您确实以这种方式分配了优先级,我认为解析器可能能够消除这些情况的歧义。

希望这可以帮助!

于 2013-03-03T20:27:58.337 回答