mirror of https://github.com/t1meshift/js.git
107 lines
3.7 KiB
Python
107 lines
3.7 KiB
Python
|
from antlr4 import *
|
||
|
|
||
|
relativeImport = False
|
||
|
if __name__ is not None and "." in __name__:
|
||
|
relativeImport = True
|
||
|
|
||
|
|
||
|
class JavaScriptBaseLexer(Lexer):
|
||
|
def __init__(self, *args, **kwargs):
|
||
|
print("JavaScriptBaseLexerInit")
|
||
|
super(JavaScriptBaseLexer, self).__init__(*args, **kwargs)
|
||
|
|
||
|
"""Stores values of nested modes. By default mode is strict or
|
||
|
defined externally (useStrictDefault)"""
|
||
|
self.scopeStrictModes = []
|
||
|
self.lastToken: Token = None
|
||
|
|
||
|
"""Default value of strict mode
|
||
|
Can be defined externally by setUseStrictDefault"""
|
||
|
self.useStrictDefault = False
|
||
|
|
||
|
"""Current value of strict mode
|
||
|
Can be defined during parsing, see StringFunctions.js and StringGlobal.js samples"""
|
||
|
self.useStrictCurrent = False
|
||
|
|
||
|
def getStrictDefault(self) -> bool:
|
||
|
return self.useStrictDefault
|
||
|
|
||
|
def setUseStrictDefault(self, value: bool):
|
||
|
self.useStrictDefault = value
|
||
|
self.useStrictCurrent = value
|
||
|
|
||
|
def isStrictMode(self):
|
||
|
return self.useStrictCurrent
|
||
|
|
||
|
def isStartOfFile(self):
|
||
|
return self.lastToken is None
|
||
|
|
||
|
def nextToken(self) -> Token:
|
||
|
"""Return the next token from the character stream and records this last
|
||
|
token in case it resides on the default channel. This recorded token
|
||
|
is used to determine when the lexer could possibly match a regex
|
||
|
literal. Also changes scopeStrictModes stack if tokenize special
|
||
|
string 'use strict';
|
||
|
|
||
|
:return the next token from the character stream."""
|
||
|
next_token: Token = super(JavaScriptBaseLexer, self).nextToken()
|
||
|
|
||
|
if next_token.channel == Token.DEFAULT_CHANNEL:
|
||
|
self.lastToken = next_token
|
||
|
|
||
|
return next_token
|
||
|
|
||
|
def processOpenBrace(self):
|
||
|
self.useStrictCurrent = bool(self.scopeStrictModes) and (
|
||
|
True if self.scopeStrictModes[-1] else self.useStrictDefault
|
||
|
)
|
||
|
self.scopeStrictModes.append(self.useStrictCurrent)
|
||
|
|
||
|
def processCloseBrace(self):
|
||
|
self.useStrictCurrent = bool(self.scopeStrictModes) and (
|
||
|
True if self.scopeStrictModes.pop(-1) else self.useStrictDefault
|
||
|
)
|
||
|
|
||
|
def processStringLiteral(self):
|
||
|
if relativeImport:
|
||
|
from .JavaScriptLexer import JavaScriptLexer
|
||
|
else:
|
||
|
from JavaScriptLexer import JavaScriptLexer
|
||
|
if not self.lastToken or self.lastToken.type == JavaScriptLexer.OpenBrace:
|
||
|
text = self.text
|
||
|
if text == '"use strict"' or text == "'use strict'":
|
||
|
if self.scopeStrictModes:
|
||
|
self.scopeStrictModes.pop(-1)
|
||
|
self.useStrictCurrent = True
|
||
|
self.scopeStrictModes.append(self.useStrictCurrent)
|
||
|
|
||
|
def isRegexPossible(self) -> bool:
|
||
|
"""Returns {@code true} if the lexer can match a regex literal. """
|
||
|
if relativeImport:
|
||
|
from .JavaScriptLexer import JavaScriptLexer
|
||
|
else:
|
||
|
from JavaScriptLexer import JavaScriptLexer
|
||
|
|
||
|
if not self.lastToken:
|
||
|
# No token has been produced yet: at the start of the input,
|
||
|
# no division is possible, so a regex literal _is_ possible.
|
||
|
return True
|
||
|
|
||
|
if self.lastToken.type in [
|
||
|
JavaScriptLexer.Identifier,
|
||
|
JavaScriptLexer.NullLiteral,
|
||
|
JavaScriptLexer.BooleanLiteral,
|
||
|
JavaScriptLexer.This,
|
||
|
JavaScriptLexer.CloseBracket,
|
||
|
JavaScriptLexer.CloseParen,
|
||
|
JavaScriptLexer.OctalIntegerLiteral,
|
||
|
JavaScriptLexer.DecimalLiteral,
|
||
|
JavaScriptLexer.HexIntegerLiteral,
|
||
|
JavaScriptLexer.StringLiteral,
|
||
|
JavaScriptLexer.PlusPlus,
|
||
|
JavaScriptLexer.MinusMinus,
|
||
|
]:
|
||
|
return False
|
||
|
|
||
|
return True
|