ANTLR 重用令牌子集
在我的 ANTLR 语法中,我有一组操作(OP):
- OP: 'I1', 'I2'....'I9', 'I10' (一组 TOKENS);
每当我发现 TOKEN:
- CASE : 有效操作是 'I1','I2','I3' (move1);
- SWITCH :有效操作为“I2”、“I4”、“I5”(move2);
其余操作由其他指令使用。
当然,在我的 Lexer 中,我不能将两个标记定义为:
OP_MOVE1 : 'I1' | 'I2' | 'I3';
OP_MOVE2 : 'I2' | 'I4' | 'I5';
因为我会得到::
OP_MOVE2 值无法访问。 I2 总是与代币 OP_MOVE1 重叠
因此,想象一下交易不仅是从 I1 到 I10,而且是从 I1 到 I5000。
一种可能的解决方案可能是:
LEXER.G4
lexer grammar LexerComment;
CASE : 'CASE' -> pushMode(CASE_MODE);
SWITCH : 'SWITCH' -> pushMode(CASE_SWITCH);
WS : [ \t] -> skip ;
EOL : [\r\n]+;
// ------------ Everything INSIDE a CASE ------------
mode CASE_MODE;
CASE_MODE_MOVE1 : 'I1' | 'I2' | 'I3';
CASE_MODE_WS : [ \t] -> channel(HIDDEN) ;
CASE_MODE_EOL : EOL -> type(EOL),popMode;
// ------------ Everything INSIDE a SWITCH ------------
mode CASE_SWITCH;
CASE_SWITCH_MOVE2: 'I2' | 'I4' | 'I5';
CASE_SWITCH_WS : [ \t] -> channel(HIDDEN) ;
CASE_SWITCHT_EOL : EOL -> type(EOL),popMode;
PARSER.g4:
parser grammar ParserComment;
options {
tokenVocab = LexerComment;
}
prog : (line? EOL)+;
line : instruction;
instruction: CASE CASE_MODE_MOVE1
|SWITCH CASE_SWITCH_MOVE2;
inputFile:
CASE I1
CASE I2
CASE I3
SWITCH I2
SWITCH I4
SWITCH I5
语法似乎工作正常,尽管我对该解决方案不满意,因为它需要大量代码,每种情况 1 种模式,并且在 中重复标记模式中常见。
还因为如果我想识别除 CASE 和 SWITCH 之外的以 MOVE1 OR MOVE2 开头的行,如下所示:
instruction: CASE CASE_MODE_MOVE1
|SWITCH CASE_SWITCH_MOVE2
| MOVE1 ;
我还没有找到解决问题的最佳解决方案:
- 我无法重用之前的 MOVE1 (CASE_MODE_MOVE1);
- 我无法定义新模式。
有没有办法正确处理类似的情况? 给定一组令牌,我想要一个可以根据上下文使用的子集。
可能试图避免将每个操作定义为基本令牌:
fragment I1: 'I1';
fragment I2: 'I2';
etc
编辑: 在我的语法中,相同的 TOKEN 可以有不同的含义。
例如,对于以下语法,TOKEN I1 具有不同的含义。
解析器:
parser grammar ParserComment;
options {
tokenVocab = LexerComment;
}
prog : (line? EOL)+;
line : instruction;
instruction: CONTEXT case_instruction;
case_instruction
: I1
| I2
| I3
;
词法分析器:
lexer grammar LexerComment;
// I1 IS A CONTEXT
CONTEXT: I1 | CASE;
CASE : 'CASE';
SWITCH : 'SWITCH';
//OPERATIONS
I1 : 'I1';
I2 : 'I2';
I3 : 'I3';
I4 : 'I4';
I5 : 'I5';
WS : [ \t] -> skip ;
EOL : [\r\n]+;
尽管名称相同,但 I1 有两个完全不同的含义(CONTEXT 和 OP)。我想认识这两种情况,并避免出现常见的 I1:
prog : (line? EOL)+;
line : instruction;
instruction: context case_instruction;
context: I1 | CASE;
case_instruction
: I1
| I2
| I3
;
出于这个原因,我曾尝试使用词法分析器中的模式进行管理。
In my ANTLR grammar I have a set of operations(OP):
- OP: 'I1', 'I2'....'I9', 'I10' (a set of TOKENS);
whenever I find TOKEN:
- CASE : valid operations are 'I1','I2','I3' (move1);
- SWITCH : valid operations are 'I2','I4','I5' (move2);
the remaining operations are used by other instructions.
Certainly in my Lexer I can't define two tokens as:
OP_MOVE1 : 'I1' | 'I2' | 'I3';
OP_MOVE2 : 'I2' | 'I4' | 'I5';
because I would get::
OP_MOVE2 values unreachable. I2 is always overlapped by token OP_MOVE1
Consequently, imagine that the transactions are not just from I1 to I10 but from I1 to I5000.
One possible solution might be:
LEXER.G4
lexer grammar LexerComment;
CASE : 'CASE' -> pushMode(CASE_MODE);
SWITCH : 'SWITCH' -> pushMode(CASE_SWITCH);
WS : [ \t] -> skip ;
EOL : [\r\n]+;
// ------------ Everything INSIDE a CASE ------------
mode CASE_MODE;
CASE_MODE_MOVE1 : 'I1' | 'I2' | 'I3';
CASE_MODE_WS : [ \t] -> channel(HIDDEN) ;
CASE_MODE_EOL : EOL -> type(EOL),popMode;
// ------------ Everything INSIDE a SWITCH ------------
mode CASE_SWITCH;
CASE_SWITCH_MOVE2: 'I2' | 'I4' | 'I5';
CASE_SWITCH_WS : [ \t] -> channel(HIDDEN) ;
CASE_SWITCHT_EOL : EOL -> type(EOL),popMode;
PARSER.g4:
parser grammar ParserComment;
options {
tokenVocab = LexerComment;
}
prog : (line? EOL)+;
line : instruction;
instruction: CASE CASE_MODE_MOVE1
|SWITCH CASE_SWITCH_MOVE2;
inputFile:
CASE I1
CASE I2
CASE I3
SWITCH I2
SWITCH I4
SWITCH I5
The grammar seems to work correctly, although I'm not satisfied with the solution as it requires a lot of code, 1 mode for each case, and repetition of tokens in common in the modes.
Also because if I wanted to recognize, in addition to CASE and SWITCH, a line that begins with MOVE1 OR MOVE2 as:
instruction: CASE CASE_MODE_MOVE1
|SWITCH CASE_SWITCH_MOVE2
| MOVE1 ;
I have not found an optimal solution to solve the problem:
- I cannot reuse the previous MOVE1 (CASE_MODE_MOVE1);
- I cannot define a new mode.
Is there a way to correctly handle similar cases?
Given a set of TOKENs I would like a subset that can be used depending on the context.
possibly trying to avoid having to define every single operation as a basic TOKEN:
fragment I1: 'I1';
fragment I2: 'I2';
etc
EDIT:
In my grammar equal TOKEN can have different meanings.
for example with the following grammar the TOKEN I1 has a different meaning.
Parser:
parser grammar ParserComment;
options {
tokenVocab = LexerComment;
}
prog : (line? EOL)+;
line : instruction;
instruction: CONTEXT case_instruction;
case_instruction
: I1
| I2
| I3
;
Lexer:
lexer grammar LexerComment;
// I1 IS A CONTEXT
CONTEXT: I1 | CASE;
CASE : 'CASE';
SWITCH : 'SWITCH';
//OPERATIONS
I1 : 'I1';
I2 : 'I2';
I3 : 'I3';
I4 : 'I4';
I5 : 'I5';
WS : [ \t] -> skip ;
EOL : [\r\n]+;
despite having the same name, I1 has two completely different meanings (CONTEXT and OP). I would like to recognize these two cases and avoid having a common I1:
prog : (line? EOL)+;
line : instruction;
instruction: context case_instruction;
context: I1 | CASE;
case_instruction
: I1
| I2
| I3
;
I had tried to manage with modes in the lexer for this reason.
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
IMO,您不应该让词法分析器决定何时应创建某些标记。让解析器决定标记在某个位置何时是正确的。
像这样的事情:
IMO, you should not let the lexer decide when certain tokens should be created. Let the parser decide when a token is correct in a certain spot.
Something like this: